Muhammadreza Haghiri: I am enthusiastic about everything new, innovative and even game-changing. I had more use-cases for my generative AI in my mind but I needed a little motivation to bring them to the real world.
One of my friends, who’s a talented web developer, once asked me about vector outputs in Mann-E. I told her it’s not possible, but with a little research and development, we did it. We could combine different models and then, create the breakthrough platform.

devmio: What are some of the biggest lessons you’ve learned throughout your journey as an AI engineer?

Muhammadreza Haghiri: This was quite a journey for me and people who joined me. Learned a lot, and the most important one is that infrastructure is all you need. living in a country where infrastructure isn’t as powerful and humongous as USA or China, we usually stop at certain points.
Although I personally made efforts to get past those points and make my business bigger and better, even with the limited infrastructure we have here.

devmio: What excites you most about the future of AI, beyond just the art generation aspects?

Muhammadreza Haghiri: AI is way more than the generative field we know and love. I wrote a lot of AI apps way before Mann-E and Vecentor. Such as ALPNR (Automated License Plate Number Recognition) proof-of-concept for Iranian license plates, American and Persian sign language translators, OSS Persian OCR, etc.
But in this new advanced field, I see a lot of potentials. Specially with these new methods such as function calling, we easily can do a lot of stuff such as making personal/home assistants, AI powered handhelds, etc.

Updates on Mann-E

devmio: Since our last conversation, what kind of updates and upgrades for Mann-E have you been working on?

Muhammadreza Haghiri: Mann-E is now having a new model (no SD anymore, but heavily influenced by SD), generating better images and we’re getting closer to Midjourney. To be honest, in eyes of most of our users, our outputs were much better than Dall-E 3 and Midjourney.
We have one more rival to fight (according to the feedback from users) and that is Ideogram. One thing we’ve done is that we’ve added an LLM improvement system for user prompts!

devmio: How does Mann-E handle complex or nuanced prompts compared to other AI models?
Are there any plans to incorporate user feedback into the training process to improve Mann-E’s generation accuracy?

Muhammadreza Haghiri: As I said in the previous answer, we now have an LLM in the middle of user and model (you have to check its checkbox by the way) and it takes your prompt, processes it, gives it to the model and boom, you have results even better that Midjourney!

P.S: I mention Midjourney a lot, since most of Iranian investors expected us to be exactly like current version of midjourney when even SD 1.5 was a new thing, this is why Midjourney became our benchmark and biggest rival at the same time!

Explore Generative AI Innovation

Generative AI and LLMs

Learn more

Questions about vecentor:

devmio: Can you please tell our readers more about the model underneath vecentor?

Muhammadreza Haghiri: It’s more like a combination of models or a pipeline of models. It uses an image generation model (like Mann-E’s model), then a pattern recognition model (or a vision model if you mind) and then, a code generation model generates the resulting SVG code.

This is the best way of creating SVG’s using AI, specially complex SVG’s like what we have on our platform!

devmio: Why did you choose a mixture of Mistral and Stable Diffusion?

Muhammadreza Haghiri: The code generation is done by Mistral (a finetuned version), but image generation and pattern recognition aren’t exactly done by SD.
Although at the time of our initial talks, we were still using SD, but we just switched to Mann-E’s proprietary models and trained a vector style on top of that.
Then, we just moved to OpenAI’s vision models in order to get the information about the image and the patterns.
At the end, we use our LLM in order to create the SVG code.
It’s a fun and complex task of generation of SVG images!

devmio: How does Vecentor’s approach to SVG generation differ from traditional image generation methods (like pixel-based models)?

Muhammadreza Haghiri: As I mentioned, SVG generation is being treated as code generation because vector images are more like guidelines of how lines and dots are drawn and colored on the user’s screen. Also there are some information of scales and the scales aren’t literal (hence the name “scalable”).
So we can claim that we achieved code generation in our company and it opens the doors for us to make new products for developers and people who need to code.

devmio: What are the advantages and limitations of using SVGs for image creation compared to other formats?

Muhammadreza Haghiri: For a lot of applications such as desktop publications or web development, SVG’s are better choice.
They can be easily modified and their quality will be the same. This is why SVG’s matter. The limitations on the other hand are that you just can’t expect a photo realistic image be a good SVG, since they’re made very very geometric.

devmio: Can you elaborate on specific applications where Vecentor’s SVG generation would be particularly beneficial (e.g., web design, animation, data visualization)?

Muhammadreza Haghiri: Of course. Our initial target market was for frontend developers and UI/UX designers. But it can also be spread to the other industries and professions as well. 

The Future of AI Art Generation

devmio: With the rise of AI art generators, how do you see the role of human artists evolving?

Muhammadreza Haghiri: Unlike what a lot of people think, humans are always ahead of machines. Although an intelligent machine is not without its own dangers, but we still can be far ahead of what a machine can do. Human artists will evolve and will become better of course, and we can take a page from their books and make better intelligent machines!

devmio: Do you foresee any ethical considerations specific to AI-generated art, such as copyright or plagiarism concerns?

Muhammadreza Haghiri: This is a valid concern and debate. Artists want to protect their rights and we also want more data. I guess the best way of getting rid of copyright disasters is that not being like OpenAI and if we use copyrighted material, we pay the owners of the art!
This is why both Mann-E and Vecentor are trained of AI generated and also royalty free material.

devmio: What potential applications do you see for AI art generation beyond creative endeavors?

Muhammadreza Haghiri: AI image, video and music generation is a tool for marketers in my opinion. You will have a world to create without any concerns about copyrights and what’s better than this? I personally think this is the future in those areas.
Also, I personally look at AI art as a form of entertainment. We used to listen to the music other people made, nowadays we can produce the music ourselves just by typing what we have in our mind!

Personal Future and Projects

devmio: Are you currently planning new projects or would you like to continue working on your existing projects?

Muhammadreza Haghiri: Yes. I’m planning for some projects, specially in Hardware and Code Generation areas. I guess they’ll be surprises for next quarters

devmio: Are there any areas in the field of AI or ML that you would like to explore further in the near future?

Muhammadreza Haghiri: I like the hardware and OS integrations. Something like a self operating computer or similar stuff. Also, I like to see more AI usage in our day to day lives.

devmio: Thank you very much for taking the time to answer our questions.

The post Art and creativity with AI appeared first on ML Conference.

Katleen Gabriels: My name is Katleen Gabriels, I am an Assistant Professor in Ethics and Philosophy of Technology at the University of Maastricht in the Netherlands, but I was born and raised in Belgium. I studied linguistics, literature, and philosophy. My research career started as an avatar in Second Life and the social virtual world. Back then I was a master student in moral philosophy and I was really intrigued by this social virtual world that promised that you could be whoever you want to be. 

That became the research of my master thesis and evolved into a PhD project which was on the ontological and moral status of virtual worlds. Since then, all my research revolves around the relation between morality and new technologies. In my current research, I look at the mutual shaping of morality and technology. 

Some years ago, I held a chair at the Faculty of Engineering Sciences in Brussels and I gave lectures to engineering and mathematics students and I’ve also worked at the Technical University of Eindhoven.

devmio: Where exactly do philosophy and AI overlap?

Katleen Gabriels: That’s a very good but also very broad question. What is really important is that an engineer does not just make functional decisions, but also decisions that have a moral impact. Whenever you talk to engineers, they very often want to make the world a better place through their technology. The idea that things can be designed for the better already has moral implications.

Way too often, people believe in the stereotype that technology is neutral. We have many examples around us today, and I think machine learning is a very good one, that a technology’s impact is highly dependent on design choices. For example, the data set and the quality of the data: If you train your algorithms with just even numbers, it will not know what an uneven number is. But there are older examples that have nothing to do with AI or computer technology. For instance, a revolving door does not include people who need a walking cane or a wheelchair.

In my talks, I always share a video of an automatic soap dispenser that does not recognize black people’s hands to show why it is so important to take into consideration a broad variety of end users. Morality and technology are not separate domains. Each technological object is human-made and humans are moral beings and therefore make moral decisions. 

Also, the philosophy of the mind is very much dealing with questions concerning intelligence, but with breakthroughs in generative AI like DALL-E, also, with what is creativity. Another important question that we’re constantly debating with new evolutions in technology is where the boundary between humans and machines is. Can we be replaced by a machine and to what extent?

Explore Generative AI Innovation

Generative AI and LLMs

Learn more

devmio: In your book “Conscientious AI: Machines Learning Morals”, you write a lot about design as a moral choice. How can engineers or developers make good moral choices in their design?

Katleen Gabriels: It’s not only about moral choices, but also about making choices that have ethical impact. My most practical hands-on answer would be that education for future engineers and developers should focus much more on these conceptual and philosophical aspects. Very often, engineers or developers are indeed thinking about values, but it’s difficult to operationalize them, especially in a business context where it’s often about “act now, apologize later”. Today we see a lot of attempts of collaboration between philosophers and developers, but that is very often just a theoretical idea.

First and foremost, it’s about awareness that design choices are not just neutral choices that developers make. We have had many designers with regrets about their designs years later. Chris Wetherell is a nice example: He designed the retweet button and initially thought that the effects of it would only be positive because it can increase how much the voices of minorities are heard. And that’s true in a way, but of course, it has also contributed to fake news to polarization.

Often, people tend to underestimate how complex ethics is. I exaggerate a little bit, but very often when teaching engineers, they have a very binary approach to things. There are always some students who want to make a decision tree out of ethical decisions. But often values clash with each other, so you need to find a trade-off. You need to incorporate the messiness of stakeholders’ voices, you need time for reflection, debate, and good arguments. That complexity of ethics cannot be transferred into a decision tree. 

If we really want to think about better and more ethical technology, we have to reserve a lot of time for these discussions. I know that when working for a highly commercial company, there is not a lot of time reserved for this.

devmio: What is your take on biases in training data? Is it something that we can get rid of? Can we know all possible biases?

Katleen Gabriels: We should be aware of the dynamics of society, our norms, and our values. They’re not static. Ideas and opinions, for example, about in vitro fertilization have changed tremendously over time, as well as our relation with animal rights, women’s rights, awareness for minorities, sustainability, and so on. It’s really important to realize that whatever machine you’re training, you must always keep it updated with how society evolves, within certain limits, of course. 

With biases, it’s important to be aware of your own blind spots and biases. That’s a very tricky one. ChatGPT, for example, is still being designed by white men and this also affects some of the design decisions. OpenAI has often been criticized for being naive and overly idealistic, which might be because the designers do not usually have to deal with the kind of problems they may produce. They do not have to deal with hate speech online because they have a very high societal status, a good job, a good degree, and so on.

devmio: In the case of ChatGPT, training the model is also problematic. In what way?

Katleen Gabriels: There’s a lot of issues with ChatGPT. Not just with the technology itself, but things revolving around it. You might already have read that a lot of the labeling and filtering of the data has been outsourced, for instance, to clickworkers in Africa. This is highly problematic. Sustainability is also a big issue because of the enormous amounts of power that the servers and GPUs require. 

Another issue with ChatGPT has to do with copyright. There have already been very good articles about the arrogance of Big Tech because their technology is very much based on the creative works of other people. We should not just be very critical about the interaction with ChatGPT, but also about the broader context of how these models have been trained, who the company and the people behind it are, what their arguments and values are, and so on. This also makes the ethical analysis much more complex.

The paradox is that on the Internet, with all our interactions, we become very transparent for Big Tech companies, but they in turn remain very opaque about their decisions. I’ve also been amazed and but annoyed about how a lot of people dealt with the open letter demanding a six-month ban on AI development. People didn’t look critically at people like Elon Musk signing it and then announcing the start of a new AI company to compete with OpenAI.

This letter focuses on existential threats and yet completely ignores the political and economic situation of Big Tech today. 

devmio: In your book, you wrote that language still represents an enormous challenge for AI. The book was published in 2020 – before ChatGPT’s advent. Do you still hold that belief today?

Katleen Gabriels: That is one of the parts that I will revise and extend significantly in the new edition. Even though the results are amazing in terms of language and spelling, ChatGPT still is not magic. One of the challenges of language is that it’s context specific and that’s still a problem for algorithms, which has not been solved with ChatGPT. It’s still a calculation, a prediction.

The breakthrough in NLP and LLMs indeed came sooner than I would have expected, but some of the major challenges are not being solved. 

devmio: Language plays a big role in how we think and how we argue and reason. How far do you think we are from artificial general intelligence? In your book, you wrote that it might be entirely possible, that consciousness might be an emergent property of our physiology and therefore not achievable outside of the human body. Is AGI even achievable?

Katleen Gabriels: Consciousness is a very tricky one. For AGI, first of all, from a semantic point of view, we need to know what intelligence is. That in itself is a very philosophical and multidimensional question because intelligence is not just about being good in mathematics. The term is very broad. There is also emotional and different kinds of intelligence, for instance. 

We could take a look at the term superintelligence, as the Swedish philosopher Nick Bostrom defines it: Superintelligence means that a computer is much better than a human being and each facet of intelligence, including emotional intelligence. We’re very far away from that. It also has to do with bodily intelligence. It’s one thing to make a good calculation, but it’s another thing to teach a robot to become a good waiter and balance glasses filled with champagne through a crowd. 

AGI or strong AI means a form of consciousness or self-consciousness and includes the very difficult concept of free will and being accountable for your actions. I don’t see this happening. 

The concept of AGI is often coupled with the fear of the singularity, which is basically a threshold: The final thing we as humans do, is develop a very smart computer and then we are done for as we cannot compete with these computers. Ray Kurzweil predicted that this is going to happen in 2045. But depending on the definition of superintelligence and the definition of singularity, I don’t believe that 2045 will be the time when this happens. Very few people actually believe that.

devmio: We regularly talk to our expert Christoph Henkelmann. He raised an interesting point about AGI. If we are able to build a self-conscious AI, we have a responsibility to that being and cannot just treat it as a simple machine.

Katleen Gabriels: I’m not the only person who made the joke, but maybe the true Turing Test is that if a machine gains self-consciousness and commits suicide, maybe that is a sign of true intelligence. If you look at the history of science fiction, people have been really intrigued by all these questions and in a way, it very much fits the quote that “to philosophize is to learn how to die.”

I can relate that quote to this, especially the singularity is all about overcoming death and becoming immortal. In a way, we could make sense of our lives if we create something that outlives us, maybe even permanently. It might make our lives worth living. 

At the academic conferences that I attend, the consensus seems to be that the singularity is bullshit, the existential threat is not that big of a deal. There are big problems and very real threats in the future regarding AI, such as drones and warfare. But a number of impactful people only tell us about those existential threats. 

devmio: We recently talked to Matthias Uhl who worked on a study about ChatGPT as a moral advisor. His study concluded that people do take moral advice from a LLM, even though it cannot give it. Is that something you are concerned with?

Katleen Gabriels: I am familiar with the study and if I remember correctly, they required a five minute attention span of their participants. So in a way, they have a big data set but very little has been studied. If you want to ask the question of to what extent would you accept moral advice from a machine, then you really need a much more in-depth inquiry. 

In a way, this is also not new. The study even echoes some of the same ideas from the 1970s with ELIZA. ELIZA was something like an early chatbot and its founder, Joseph Weizenbaum, was shocked when he found out that people anthropomorphized it. He knew what it was capable of and in his book “Computer Power and Human Reason: From Judgment to Calculation” he recalls anecdotes where his secretary asked him to leave the room so she could interact with ELIZA in private. People were also contemplating to what extent ELIZA could replace human therapists. In a way, this says more about human stupidity than about artificial intelligence. 

In order to have a much better understanding of how people would take or not take moral advice from a chatbot, you need a very intense study and not a very short questionnaire.

devmio:  It also shows that people long for answers, right? That we want clear and concise answers to complex questions.

Katleen Gabriels: Of course, people long for a manual. If we were given a manual by birth, people would use it. It’s also about moral disengagement, it’s about delegating or distributing responsibility. But you don’t need this study to conclude that.

It’s not directly related, but it’s also a common problem on dating apps. People are being tricked into talking to chatbots. Usually, the longer you talk to a chatbot, the more obvious it might become, so there might be a lot of projection and wishful thinking. See also the media equation study. We simply tend to treat technology as human beings.

devmio: We use technology to get closer to ourselves, to get a better understanding of ourselves. Would you agree?

Katleen Gabriels: I teach a course about AI and there’s always students saying, “This is not a course about AI, this is a course about us!” because it’s so much about what intelligence is, where the boundary between humans and machines is, and so on. 

This would also be an interesting study for the future of people who believe in a fatal singularity in the future. What does it say about them and what they think of us humans?

devmio: Thank you for your answers!

The post Building Ethical AI: A Guide for Developers on Avoiding Bias and Designing Responsible Systems appeared first on ML Conference.

Embeddings are essentially a technique for converting unstructured data into a structure that can be easily processed by software. They are used to transform complex data such as words, sentences, or even entire documents into a vector space, with similar elements close to each other. These vector representations allow machines to recognize and exploit nuances and relationships in the data. Which is essential for a variety of applications such as natural language processing (NLP), image recognition, and recommendation systems.

OpenAI, the company behind ChatGPT, offers models for creating embeddings for texts, among other things. At the end of January 2024, OpenAI presented new versions of these embeddings models, which are more powerful and cost-effective than their predecessors. In this article, after a brief introduction to embeddings, we’ll take a closer look at the OpenAI embeddings and the recently introduced innovations, discuss how they work, and examine how they can be used in various software development projects.

Embeddings briefly explained

Imagine you’re in a room full of people and your task is to group these people based on their personality. To do this, you could start asking questions about different personality traits. For example, you could ask how open someone is to new experiences and rate the answer on a scale from 0 to 1. Each person is then assigned a number that represents their openness.

Next, you could ask about another personality trait, such as the level of sense of duty, and again give a score between 0 and 1. Now each person has two numbers that together form a vector in a two-dimensional space. By asking more questions about different personality traits and rating them in a similar way, you can create a multidimensional vector for each person. In this vector space, people who have similar vectors can then be considered similar in terms of their personality.

In the world of artificial intelligence, we use embeddings to transform unstructured data into an n-dimensional vector space. Similarly how a person’s personality traits are represented in the vector space, each point in this vector space represents an element of the original data (such as a word or phrase) in a way that is understandable and processable by computers.

OpenAI Embeddings

OpenAI embeddings extend this basic concept. Instead of using simple features like personality traits, OpenAI models use advanced algorithms and big data to achieve a much deeper and more nuanced representation of the data. The model not only analyzes individual words, but also looks at the context in which those words are used, resulting in more accurate and meaningful vector representations.

Another important difference is that OpenAI embeddings are based on sophisticated machine learning models that can learn from a huge amount of data. This means that they can recognize subtle patterns and relationships in the data that go far beyond what could be achieved by simple scaling and dimensioning, as in the initial analogy. This leads to a significantly improved ability to recognize and exploit similarities and differences in the data.

Explore Generative AI Innovation

Generative AI and LLMs

Learn more

Individual values are not meaningful

While in the personality trait analogy, each individual value of a vector can be directly related to a specific characteristic – for example openness to new experiences or a sense of duty – this direct relationship no longer exists with OpenAI embeddings. In these embeddings, you cannot simply look at a single value of the vector in isolation and draw conclusions about specific properties of the input data. For example, a specific value in the embedding vector of a sentence cannot be used to directly deduce how friendly or not this sentence is.

The reason for this lies in the way machine learning models, especially those used to create embeddings, encode information. These models work with complex, multi-dimensional representations where the meaning of a single element (such as a word in a sentence) is determined by the interaction of many dimensions in vector space. Each aspect of the original data – be it the tone of a text, the mood of an image, or the intent behind a spoken utterance – is captured by the entire spectrum of the vector rather than by individual values within that vector.

Therefore, when working with OpenAI embeddings, it’s important to understand that the interpretation of these vectors is not intuitive or direct. You need algorithms and analysis to draw meaningful conclusions from these high-dimensional and densely coded vectors.

Comparison of vectors with cosine similarity

A central element in dealing with embeddings is measuring the similarity between different vectors. One of the most common methods for this is cosine similarity. This measure is used to determine how similar two vectors are and therefore the data they represent.

To illustrate the concept, let’s start with a simple example in two dimensions. Imagine two vectors in a plane, each represented by a point in the coordinate system. The cosine similarity between these two vectors is determined by the cosine of the angle between them. If the vectors point in the same direction, the angle between them is 0 degrees and the cosine of this angle is 1, indicating maximum similarity. If the vectors are orthogonal (i.e. the angle is 90 degrees), the cosine is 0, indicating no similarity. If they are opposite (180 degrees), the cosine is -1, indicating maximum dissimilarity.

Figure 1 -Cosine similarity

[gdlr_box title=”A Python Notebook to try out” ]
Accompanying this article is a Google Colab Python Notebook which you can use to try out many of the examples shown here. Colab, short for Colaboratory, is a free cloud service offered by Google. Colab makes it possible to write and execute Python code in the browser. It’s based on Jupyter Notebooks, a popular open-source web application that makes it possible to combine code, equations, visualizations, and text in a single document-like format. The Colab service is well suited for exploring and experimenting with the OpenAI API using Python.[/gdlr_box]

In practice, especially when working with embeddings, we are dealing with n-dimensional vectors. The calculation of the cosine similarity remains conceptually the same, even if the calculation is more complex in higher dimensions. Formally, the cosine similarity of two vectors A and B in an n-dimensional space is calculated by the scalar product (dot product) of these vectors divided by the product of their lengths:

Figure 2 – Calculation of cosine similarity

The normalization of vectors plays an important role in the calculation of cosine similarity. If a vector is normalized, this means that its length (norm) is set to 1. For normalized vectors, the scalar product of two vectors is directly equal to the cosine similarity since the denominators in the formula from Figure 2 are both 1. OpenAI embeddings are normalized, which means that to calculate the similarity between two embeddings, only their scalar product needs to be calculated. This not only simplifies the calculation, but also increases efficiency when processing large quantities of embeddings.

OpenAI Embeddings API

OpenAI offers a web API for creating embeddings. The exact structure of this API, including code examples for curl, Python and Node.js, can be found in the OpenAI reference documentation.

OpenAI does not use the LLM from ChatGPT to create embeddings, but rather specialized models. They were developed specifically for the creation of embeddings and are optimized for this task. Their development was geared towards generating high-dimensional vectors that represent the input data as well as possible. In contrast, ChatGPT is primarily optimized for generating and processing text in a conversational form. The embedding models are also more efficient in terms of memory and computing requirements than more extensive language models such as ChatGPT. As a result, they are not only faster but much more cost-effective.

New embedding models from OpenAI

Until recently, OpenAI recommended the use of the text-embedding-ada-002 model for creating embeddings. This model converts text into a sequence of floating point numbers (vectors) that represent the concepts within the content. The ada v2 model generated embeddings with a size of 1536 dimensions and delivered solid performance in benchmarks such as MIRACL and MTEB, which are used to evaluate model performance in different languages and tasks.

At the end of January 2024, OpenAI presented new, improved models for embeddings:

text-embedding-3-small: A smaller, more efficient model with improved performance compared to its predecessor. It performs better in benchmarks and is significantly cheaper.
text-embedding-3-large: A larger model that is more powerful and creates embeddings with up to 3072 dimensions. It shows the best performance in the benchmarks but is slightly more expensive than ada v2.

A new function of the two new models allows developers to adjust the size of the embeddings when generating them without significantly losing their concept-representing properties. This enables flexible adaptation, especially for applications that are limited in terms of available memory and computing power.

Readers who are interested in the details of the new models can find them in the announcement on the OpenAI blog. The exact costs of the various embedding models can be found here.

Create embeddings with Python

In the following section, the use of embeddings is demonstrated using a few code examples with Python. The code examples are designed so that they can be tried out in Python Notebooks. They are also available in a similar form in the previously mentioned accompanying Google Colab notebook mentioned above.
Listing 1 shows how to create embeddings with the Python SDK from OpenAI. In addition, numpy is used to show that the embeddings generated by OpenAI are normalized.

Listing 1

from openai import OpenAI
from google.colab import userdata
import numpy as np

# Create OpenAI client
client = OpenAI(
    api_key=userdata.get('openaiKey'),
)

# Define a helper function to calculate embeddings
def get_embedding_vec(input):
  """Returns the embeddings vector for a given input"""
  return client.embeddings.create(
        input=input,
        model="text-embedding-3-large", # We use the new embeddings model here (announced end of Jan 2024)
        # dimensions=... # You could limit the number of output dimensions with the new embeddings models
    ).data[0].embedding

# Calculate the embedding vector for a sample sentence
vec = get_embedding_vec("King")
print(vec[:10])

# Calculate the magnitude of the vector. I should be 1 as
# embedding vectors from OpenAI are always normalized.
magnitude = np.linalg.norm(vec)
magnitude

Similarity analysis with embeddings

In practice, OpenAI embeddings are often used for similarity analysis of texts (e.g. searching for duplicates, finding relevant text sections in relation to a customer query, and grouping text). Embeddings are very well suited for this, as they work in a fundamentally different way to comparison methods based on characters, such as Levenshtein distance. While it measures the similarity between texts by counting the minimum number of single-character operations (insert, delete, replace) required to transform one text into another, embeddings capture the meaning and context of words or sentences. They consider the semantic and contextual relationships between words, going far beyond a simple character-based level of comparison.

As a first example, let’s look at the following three sentences (the following examples are in English, but embeddings work analogously for other languages and cross-language comparisons are also possible without any problems):

I enjoy playing soccer on weekends.
Football is my favorite sport. Playing it on weekends with friends helps me to relax.
In Austria, people often watch soccer on TV on weekends.

In the first and second sentence, two different words are used for the same topic: Soccer and football. The third sentence contains the original soccer, but it has a fundamentally different meaning from the first two sentences. If you calculate the similarity of sentence 1 to 2, you get 0.75. The similarity of sentence 1 to 3 is only 0.51. The embeddings have therefore reflected the meaning of the sentence and not the choice of words.

Here is another example that requires an understanding of the context in which words are used:
He is interested in Java programming.
He visited Java last summer.
He recently started learning Python programming.

In sentence 2, Java refers to a place, while sentences 1 and 3 have something to do with software development. The similarity of sentence 1 to 2 is 0.536, but that of 1 to 3 is 0.587. As expected, the different meaning of the word Java has an effect on the similarity.

The next example deals with the treatment of negations:
I like going to the gym.
I don’t like going to the gym.
I don’t dislike going to the gym.

Sentences 1 and 2 say the opposite, while sentence 3 expresses something similar to sentence 1. This content is reflected in the similarities of the embeddings. Sentence 1 to sentence 2 yields a cosine similarity of 0.714 while sentence 1 compared to sentence 3 yields 0.773. It is perhaps surprising that there is no major difference between the embeddings. However, it’s important to remember that all three sets are about the same topic: The question of whether you like going to the gym to work out.

The last example shows that the OpenAI embeddings models, just like ChatGPT, have built in a certain “knowledge” of concepts and contexts through training with texts about the real world.

I need to get better slicing skills to make the most of my Voron.
3D printing is a worthwhile hobby.
Can I have a slice of bread?

In order to compare these sentences in a meaningful way, it’s important to know that Voron is the name of a well-known open-source project in the field of 3D printing. It’s also important to note that slicing is a term that plays an important role in 3D printing. The third sentence also mentions slicing, but in a completely different context to sentence 1. Sentence 2 mentions neither slicing nor Voron. However, the trained knowledge enables the OpenAI Embeddings model to recognize that sentences 1 and 2 have a thematic connection, but sentence 3 means something completely different. The similarity of sentence 1 and 2 is 0.333 while the comparison of sentence 1 and 3 is only 0.263.

Similarity values are not percentages

The similarity values from the comparisons shown above are the cosine similarity of the respective embeddings. Although the cosine similarity values range from -1 to 1, with 1 being the maximum similarity and -1 the maximum dissimilarity, they are not to be interpreted directly as percentages of agreement. Instead, these values should be considered in the context of their relative comparisons. In applications such as searching text sections in a knowledge base, the cosine similarity values are used to sort the text sections in terms of their similarity to a given query. It is important to see the values in relation to each other. A higher value indicates a greater similarity, but the exact meaning of the value can only be determined by comparing it with other similarity values. This relative approach makes it possible to effectively identify and prioritize the most relevant and similar text sections.

Embeddings and RAG solutions

Embeddings play a crucial role in Retrieval Augmented Generation (RAG) solutions, an approach in artificial intelligence that combines the capabilities of information retrieval and text generation. Embeddings are used in RAG systems to retrieve relevant information from large data sets or knowledge databases. It is not necessary for these databases to have been included in the original training of the embedding models. They can be internal databases that are not available on the public Internet.
With RAG solutions, queries or input texts are converted into embeddings. The cosine similarity to the existing document embeddings in the database is then calculated to identify the most relevant text sections from the database. This retrieved information is then used by a text generation model such as ChatGPT to generate contextually relevant responses or content.

Vector databases play a central role in the functioning of RAG systems. They are designed to efficiently store, index and query high-dimensional vectors. In the context of RAG solutions and similar systems, vector databases serve as storage for the embeddings of documents or pieces of data that originate from a large amount of information. When a user makes a request, this request is first transformed into an embedding vector. The vector database is then used to quickly find the vectors that correspond most closely to this query vector – i.e. those documents or pieces of information that have the highest similarity. This process of quickly finding similar vectors in large data sets is known as Nearest Neighbor Search.

Challenge: Splitting documents

A detailed explanation of how RAG solutions work is beyond the scope of this article. However, the explanations regarding embeddings are hopefully helpful for getting started with further research on the topic of RAGs.

However, one specific point should be pointed out at the end of this article: A particular and often underestimated challenge in the development of RAG systems that go beyond Hello World prototypes is the splitting of longer texts. Splitting is necessary because the OpenAI embeddings models are limited to just over 8,000 tokens. One token corresponds to approximately 4 characters in the English language (see also).

It’s not easy finding a good strategy for splitting documents. Naive approaches such as splitting after a certain number of characters can lead to the context of text sections being lost or distorted. Anaphoric links are a typical example of this. The following two sentences are an example:

VX-2000 requires regular lubrication to maintain its smooth operation.
The machine requires the DX97 oil, as specified in the maintenance section of this manual.

The machine in the second sentence is an anaphoric link to the first sentence. If the text were to be split up after the first sentence, the essential context would be lost, namely that the DX97 oil is necessary for the VX-2000 machine.

There are various approaches to solving this problem, which will not be discussed here to keep this article concise. However, it is essential for developers of such software systems to be aware of the problem and understand how splitting large texts affects embeddings.

Summary

Embeddings play a fundamental role in the modern AI landscape, especially in the field of natural language processing. By transforming complex, unstructured data into high-dimensional vector spaces, embeddings enable in-depth understanding and efficient processing of information. They form the basis for advanced technologies such as RAG systems and facilitate tasks such as information retrieval, context analysis, and data-driven decision-making.

OpenAI’s latest innovations in the field of embeddings, introduced at the end of January 2024, mark a significant advance in this technology. With the introduction of the new text-embedding-3-small and text-embedding-3-large models, OpenAI now offers more powerful and cost-efficient options for developers. These models not only show improved performance in standardized benchmarks, but also offer the ability to find the right balance between performance and memory requirements on a project-specific basis through customizable embedding sizes.

Embeddings are a key component in the development of intelligent systems that aim to achieve useful processing of speech information.

Links and Literature:

1. https://colab.research.google.com/gist/rstropek/f3d4521ed9831ae5305a10df84a42ecc/embeddings.ipynb
2. https://platform.openai.com/docs/api-reference/embeddings/create
3. https://openai.com/blog/new-embedding-models-and-api-updates
4. https://openai.com/pricing
5. https://platform.openai.com/tokenizer

The post OpenAI Embeddings appeared first on ML Conference.

1. Geospatial Analysis: Accurate address matching forms the foundation of geospatial analysis, allowing organizations to make informed decisions about locations, market trends, and resource allocation across various industries like retail and media.
2. Real Estate Data Management: In the real estate industry, precise address matching facilitates property valuation, market analysis, and portfolio management.
3. Logistics and Navigation: Efficient routing and delivery depend on accurate address matching.
4. Compliance and Regulation: Many regulatory requirements mandate precise address data, such as tax reporting and census data collection.

Cherre is the leading real estate data management company, we specialize in accurate address matching for the second use case. Whether you’re an asset manager, portfolio manager, or real estate investor, a building represents the atomic unit of all financial, legal, and operating information. However, real estate data lives in many silos, which makes having a unified view of properties difficult. Address matching is an important step in breaking down data silos in real estate. By joining disparate datasets on address, we can unlock many opportunities for further portfolio analysis.

Data Silos in Real Estate

Real estate data usually fall into the following categories: public, third party, and internal. Public data is collected by governmental agencies and made available publicly, such as land registers. The quality of public data is generally not spectacular and the data update frequency is usually delayed, but it provides the most comprehensive coverage geographically. Don’t be surprised if addresses from public data sources are misaligned and misspelled.

Third party data usually come from data vendors, whose business models focus on extracting information as datasets and monetizing those datasets. These datasets usually have good data quality and are much more timely, but limited in geographical coverage. Addresses from data vendors are usually fairly clean compared to public data, but may not be the same address designation across different vendors. For large commercial buildings with multiple entrances and addresses, this creates an additional layer of complexity.

Lastly, internal data is information that is collected by the information technology (I.T.) systems of property owners and asset managers. These can incorporate various functions, from leasing to financial reporting, and are often set up to represent the business organizational structures and functions. Depending on the governance standards, and data practices, the quality of these datasets can vary and data coverage only encompasses the properties in the owner’s portfolio. Addresses in these systems can vary widely, some systems are designated at the unit-level, while others designate the entire property. These systems also may not standardize addresses inherently, which makes it difficult to match property records across multiple systems.

With all these variations in data quality, coverage, and address formats, we can see the need for having standardized addresses to do basic property-level analysis.

Address Matching Using the Parse-Clean-Match Strategy

In order to match records across multiple datasets, the address parse-clean-match strategy works very well regardless of region. By breaking down addresses into their constituent pieces, we have many more options for associating properties with each other. Many of the approaches for this strategy use simple natural language processing (NLP) techniques.

NEW & PRACTICAL ENDEAVORS FOR ML

Machine Learning Principles

Learn more

Address Parsing

Before we can associate addresses with each other, we must first parse the address. Address parsing is the process of breaking down each address string into its constituent components. Components in addresses will vary by country.

In the United States and Canada, addresses are generally formatted as the following:

{street_number} {street_name}

{city}, {state_or_province} {postal_code}

{country}

In the United Kingdom, addresses are formatted very similarly as in the U.S. and Canada, with an additional optional locality designation:

{building_number} {street_name}

{locality (optional)}

{city_or_town}

{postal_code}

{country}

French addresses vary slightly from U.K. addresses with the order of postal code and city:

{building_number} {street_name}

{postal_code} {city}

{country}

German addresses take the changes in French addresses and then swap the order of street name and building number:

{street_name} {building_number} {postal_code} {city} {country}

Despite the slight variations across countries’ address formats, addresses generally have the same components, which makes this an easily digestible NLP problem. We can break down the process into the following steps:

1. Tokenization: Split the address into its constituent words. This step segments the address into manageable units.
2. Named Entity Recognition (NER): Identify entities within the address, such as street numbers, street names, cities, postal codes, and countries. This involves training or using pre-trained NER models to label the relevant parts of the address.
3. Sequence Labeling: Use sequence labeling techniques to tag each token with its corresponding entity

Let’s demonstrate address parsing with a sample Python code snippet using the spaCy library. SpaCy is an open-source software library containing many neural network models for NLP functions. SpaCy supports models across 23 different languages and allows for data scientists to train custom models for their own datasets. We will demonstrate address parsing using one of SpaCy’s out-of-the-box models for the address of a historical landmark: David Bowie’s Berlin apartment.

import spacy

# Load the NER spaCy model
model = spacy.load("en_core_web_sm")

# Address to be parsed
address = "Hauptstraße 155, 10827 Berlin"

# Tokenize and run NER
doc = model(address)

# Extract address components
street_number = ""
street_name = ""
city = ""
state = ""
postal_code = ""

for token in doc:
    if token.ent_type_ == "GPE":  # Geopolitical Entity (City)
        city = token.text
    elif token.ent_type_ == "LOC":  # Location (State/Province)
        state = token.text
    elif token.ent_type_ == "DATE":  # Postal Code
        postal_code = token.text
    else:
        if token.is_digit:
            street_number = token.text
        else:
            street_name += token.text + " "

# Print the parsed address components
print("Street Number:", street_number)
print("Street Name:", street_name)
print("City:", city)
print("State:", state)
print("Postal Code:", postal_code)

Now that we have a parsed address, we can now clean each address component.

Address Cleaning

Address cleaning is the process of converting parsed address components into a consistent and uniform format. This is particularly important for any public data with misspelled, misformatted, or mistyped addresses. We want to have addresses follow a consistent structure and notation, which will make further data processing much easier.

To standardize addresses, we need to standardize each component, and how the components are joined. This usually entails a lot of string manipulation. There are many open source libraries (such as libpostal) and APIs that can automate this step, but we will demonstrate the basic premise using simple regular expressions in Python.

import pandas as pd
import re

# Sample dataset with tagged address components
data = {
    'Street Name': ['Hauptstraße', 'Schloß Nymphenburg', 'Mozartweg'],
    'Building Number': ['155', '1A', '78'],
    'Postal Code': ['10827', '80638', '54321'],
    'City': ['Berlin', ' München', 'Hamburg'],
}

df = pd.DataFrame(data)

# Functions with typical necessary steps for each address component

# We uppercase all text for easier matching in the next step

def standardize_street_name(street_name):

    # Remove special characters and abbreviations, uppercase names
    standardized_name = re.sub(r'[^\w\s]', '', street_name)
    return standardized_name.upper()

def standardize_building_number(building_number):
    # Remove any non-alphanumeric characters (although exceptions exist)
    standardized_number = re.sub(r'\W', '', building_number)
    return standardized_number

def standardize_postal_code(postal_code):

    # Make sure we have consistent formatting (i.e. leading zeros)
    return postal_code.zfill(5)

def standardize_city(city):

    # Upper case the city, normalize spacing between words
    return ' '.join(word.upper() for word in city.split())

# Apply standardization functions to our DataFrame
df['Street Name'] = df['Street Name'].apply(standardize_street_name)
df['Building Number'] = df['Building Number'].apply(standardize_building_number)
df['Postal Code'] = df['Postal Code'].apply(standardize_postal_code)
df['City'] = df['City'].apply(standardize_city)

# Finally create a standardized full address (without commas)

df[‘Full Address’] = df['Street Name'] + ' ' + df['Building Number'] + ' ' + df['Postal Code'] + ' ' + df['City']

Address Matching

Now that our addresses are standardized into a consistent format, we can finally match addresses from one dataset to address in another dataset. Address matching involves identifying and associating similar or identical addresses from different datasets. When two full addresses match exactly, we can easily associate the two together through a direct string match.

When addresses don’t match, we will need to apply fuzzy matching on each address component. Below is an example of how to do fuzzy matching on one of the standardized address components for street names. We can apply the same logic to city and state as well.

from fuzzywuzzy import fuzz

# Sample list of street names from another dataset
street_addresses = [
    "Hauptstraße",
    "Schlossallee",
    "Mozartweg",
    "Bergstraße",
    "Wilhelmstraße",
    "Goetheplatz",
]

# Target address component (we are using street name)
target_street_name = "Hauptstrasse " # Note the different spelling and space 

# Similarity threshold

# Increase this number if too many false positives

# Decrease this number if not enough matches

threshold = 80

# Perform fuzzy matching
matches = []

for address in street_addresses:
    similarity_score = fuzz.partial_ratio(address, target_street_name)
    if similarity_score >= threshold:
        matches.append((address, similarity_score))

matches.sort(key=lambda x: x[1], reverse=True)

# Display matched street name
print("Target Street Name:", target_street_name)
print("Matched Street Names:")
for match in matches:
    print(f"{match[0]} (Similarity: {match[1]}%)")

Up to here, we have solved the problem for properties with the same address identifiers. But what about the large commercial buildings with multiple addresses?

Other Geospatial Identifiers

Addresses are not the only geospatial identifiers in the world of real estate. An address typically refers to the location of a structure or property, often denoting a street name and house number.  There are actually four other geographic identifiers in real estate:

1. A “lot” represents a portion of land designated for specific use or ownership.
2. A “parcel” extends this notion to a legally defined piece of land with boundaries, often associated with property ownership and taxation.
3. A “building” encompasses the physical structures erected on these parcels, ranging from residential homes to commercial complexes.
4. A “unit” is a sub-division within a building, typically used in multi-unit complexes or condominiums. These can be commercial complexes (like office buildings) or residential complexes (like apartments).

What this means is that we actually have multiple ways of identifying real estate objects, depending on the specific persona and use case. For example, leasing agents focus on the units within a building for tenants, while asset managers optimize for the financial performance of entire buildings. The nuances of these details are also codified in many real estate software systems (found in internal data), in the databases of governments (found in public data), and across databases of data vendors (found in third party data). In public data, we often encounter lots and parcels. In vendor data, we often find addresses (with or without units). In real estate enterprise resource planning systems, we often find buildings, addresses, units, and everything else in between.

In the case of large commercial properties with multiple addresses, we need to associate various addresses with each physical building. In this case, we can use geocoding and point-in-polygon searches.

Geocoding Addresses

Geocoding is the process of converting addresses into geographic coordinates. The most common form is latitude and longitude. European address geocoding requires a robust understanding of local address formats, postal codes, and administrative regions. Luckily, we have already standardized our addresses into an easily geocodable format.

Many commercial APIs exist for geocoding addresses in bulk, but we will demonstrate geocoding using a popular Python library, Geopy, to geocode addresses.

from geopy.geocoders import Nominatim

geolocator = Nominatim(user_agent="my_geocoder")
location = geolocator.geocode("1 Canada Square, London")
print(location.latitude, location.longitude)

Now that we’ve converted our addresses into latitude and longitude, we can use point-in-polygon searches to associate addresses with buildings.

Point-in-Polygon Search

A point-in-polygon search is a technique to determine if a point is located within the boundaries of a given polygon.

The “point” in a point-in-polygon search refers to a specific geographical location defined by its latitude and longitude coordinates. We have already obtained our points by geocoding our addresses.

The “polygon” is a closed geometric shape with three or more sides, which is usually characterized by a set of vertices (points) connected by edges, forming a closed loop. Building polygons can be downloaded from open source sites like OpenStreetMap or from specific data vendors. The quality and detail of the OpenStreetMap building data may vary, and the accuracy of the point-in-polygon search depends on the precision of the building geometries.

While the concept seems complex, the code for creating this lookup is quite simple. We demonstrate a simplified example using our previous example of 1 Canada Square in London.

import json
from shapely.geometry import shape, Point

# Load the GeoJSON data
with open('building_data.geojson') as geojson_file:
    building_data = json.load(geojson_file)

# Latitude and Longitude of 1 Canada Square in Canary Wharf
lat, lon = 51.5049, 0.0195

# Create a Point geometry for 1 Canada Square
point_1_canada = Point(lon, lat)

# See if point is within any of the polygons
for feature in building_data['features']:
    building_geometry = shape(feature['geometry'])

    if point_1_canada.within(building_geometry):
        print(f"Point is within this building polygon: {feature}")
        break
else:
    print("Point is not within any building polygon in the dataset.")

Using this technique, we can properly identify all addresses associated with this property.

Summary

Addresses in real life are confusing because they are the physical manifestation of many disparate decisions in city planning throughout the centuries-long life of a city. But using addresses to match across different datasets doesn’t have to be confusing.

Using some basic NLP and geocoding techniques, we can easily associate property-level records across various datasets from different systems. Only through breaking down data silos can we have more holistic views of property behaviors in real estate.

Author Biography

Alyce Ge is data scientist at Cherre, the industry-leading real estate data management and analytics platform. Prior to joining Cherre, Alyce held data science and analytics roles for a variety of technology companies focusing on real estate and business intelligence solutions. Alyce is a Google Cloud-certified machine learning engineer, Google Cloud-certified data engineer, and Triplebyte certified data scientist. She earned her Bachelor of Science in Applied Mathematics from Columbia University in New York.

The post Address Matching with NLP in Python appeared first on ML Conference.

Open AI’s release of ChatGPT and GPT-4 has sparked a Cambrian explosion of new products and projects, shifting the landscape of artificial intelligence significantly. These models have both quantitatively and qualitatively advanced beyond their language modeling predecessors. Similarly to how the deep learning model called AlexNet significantly improved on the ImageNet benchmark for computer vision back in 2012. More importantly, these models exhibit a capability, the ability to perform many different tasks such as machine translation or when given a few examples of the task: few-shot learning. Unlike humans, most language models require large supervised datasets before they can be expected to perform a specific task. This plasticity of “intelligence” that GPT-3 was capable of opened up new possibilities in the field of AI. It is a system capable of problem-solving which enables the implementation of many long-imagined AI applications.

Even the successor model to GPT-3, GPT-4, is still just a language model at the end of the day and still quite far from Artificial General Intelligence. In general, the ”prompt to single response“ formulation of language models is much too limited to perform complex multi-step tasks. For an AI to be generally intelligent, it must seek out information, remember, learn, and interact with the world in steps. There have recently been many projects on GitHub that have essentially created self-talking loops and prompting structures on top of OpenAI’s APIs for the GPT-3.5 and GPT-4. These are models that form a system that can plan, generate code, debug, and execute programs. These systems in theory have the potential to be much more general and approach what many people think of when they hear “AI”.

The concept of systems that intelligently interact in their environment is not completely new, and has been heavily researched in a field of AI called reinforcement learning. The influential textbook “Artificial Intelligence: A Modern Approach” by Russell and Norvig covers many different structures for how to build intelligent “agents” – entities capable of perceiving their environment and acting to achieve specific objectives. While I don’t believe Russel and Norvig imagined that these agent structures would be mostly language model-based. They did describe how they would perform their various steps with plain English sentences and questions as they were mostly for illustrative purposes. Since we now have language models capable of functionally understanding the steps and questions they use, it is much easier to implement many of these structures as real programs today.

While I haven’t seen any projects using prompts inspired by the AI: AMA textbook for their agents, the open-source community has been leveraging GPT 3.5 and GPT-4 to develop agent or agent-like programs using similar ideas. Examples of such programs include Baby AGI, AutoGPT, and MetaGPT. While these agents are not designed to interact with a game or simulated environment like traditional RL agents, They do typically generate code, detect errors, and alter their behavior accordingly.  So in a sense, they are interacting with and perceiving the “environment” of programming, and are significantly more capable than anything before. 

This article aims to delve into the capabilities and limitations of OpenAI’s models, examine the functionalities of agents like Baby AGI, and discuss potential future advancements in this rapidly evolving field.

Understanding the Capabilities of GPT-3.5 and GPT-4:

GPT-3.5 and GPT-4 are important milestones not only in natural language processing but also in the field of AI. Their ability to generate contextually appropriate, coherent responses to a myriad of prompts has reshaped our expectations of what a language model can achieve. However, to fully appreciate their potential and constraints, it’s necessary to delve deeper into their implementation.

One significant challenge these models face is the problem of hallucination. Hallucination refers to instances where a language model generates outputs that seem plausible but are entirely fabricated or not grounded in the input data. Hallucination is a challenge in Chat GPT as these models are fundamentally outputting the probability distribution of the next word, and that probability distribution is sampled in a weighted random fashion. This leads to the generation of responses that are statistically likely but not necessarily accurate or truthful. The limitation of relying on maximum likelihood sampling in language models is that it prioritizes coherence over veracity, leading to creative but potentially misleading outputs. This essentially limits the ability of the model to reason and make logical deductions when the output pattern is very unlikely. While they can exhibit some degree of reasoning and common sense, they don’t yet match human-level reasoning capabilities. This is because they are limited to statistical patterns present in their training data, rather than a thorough understanding of the underlying concepts.

To quantitatively assess these models’ reasoning capabilities, researchers use a range of tasks including logical puzzles, mathematical operations, and exercises that require understanding causal relationships. [https://arxiv.org/abs/2304.03439] While OpenAI does boast about GPT-4’s ability to pass many aptitude tests including the Bar exam. The model struggles to show the same capabilities with out-of-distribution logical puzzles, which can be expected when you consider the statistical nature of the models.

To be fair to these models, the role of language in human reasoning is underappreciated by the general public. Humans also use language generation as a form of reasoning, making connections, and drawing inferences through linguistic patterns. If the brain area that is responsible for language is damaged, research has shown that reasoning is impaired: [https://www.frontiersin.org/articles/10.3389/fpsyg.2015.01523/full]. Therefore, just because language models are mostly statistical next-word generators, we shouldn’t disregard their reasoning capabilities entirely. While it has limitations, it is something that can be taken advantage of in systems. A genuine potential of language models exists to replicate certain reasoning processes and this theory of the link between reasoning and language explains their capabilities.

While GPT-3.5 and GPT-4 have made significant strides in natural language processing, there is still work to do. Ongoing research is focused on enhancing these abilities and tackling these challenges. It is important for systems today to work around these limitations and take advantage of language models’ strengths as we explore their potential applications and continue to push AI’s boundaries.

THE PECULIARITIES OF ML SYSTEMS

Machine Learning Advanced Developments

Learn more

Exploring Collaborative Agent Systems: BabyAGI, HuggingFace, and MetaGPT:

BabyAGI, created by Yohei Nakajima, serves as an interesting proof-of-concept in the domain of agents. The main idea behind it consists of creating three “sub-agents”: the Task Creator, Task Prioritizer, and Task Executor.  By making the sub-agents have specific roles and collaborating by way of a task management system, BabyAGI can reason better and achieve many more tasks than a single prompt alone, hence creating the ”collaborative agent system” concept.  While I do not believe the collaborative agent strategy BabyAGI implements is a completely novel concept.  It is one of the early successful experiments built on top of GPT-4 with code we can easily understand. In BabyAGI, the Task Creator initiates the process by setting the goal and formulating the task list. The Task Prioritizer then rearranges the tasks based on their significance in achieving the goal, and finally, the Task Executor carries out the tasks one by one. The output of each task is stored in a vector database, which can look up data by similarity, for future reference serving as a type of memory for the Task Executor.

Fig 1. A high-level description of the BabyAGI framework

HuggingFace’s Transformers Agents, is another substantial agent framework. It has gained popularity for its ability to leverage the library of pre-trained models on HuggingFace. By leveraging the StarCoder model, the Transformers Agent can string together many different models available on HuggingFace to accomplish various tasks. It can solve a range of visual, audio, and natural language processing functionalities. However, HuggingFace agents lack error recovery mechanisms, often requiring external intervention to troubleshoot issues and continue with the task.

Fig 2. Example of HuggingFace’s Transformers Agent

MetaGPT adopts a unique approach by emulating a virtual company where different agents play specific roles. Each virtual agent within MetaGPT has its own thoughts, allowing them to contribute their perspectives and expertise to the collaborative process. This approach recognizes the collective intelligence of human communities and seeks to replicate it in AI systems.

Fig. 3. The Software Company structure of MetaGPT

BabyAGI, Transformers, and MetaGPT, with their own strengths and limitations, collectively exemplify the evolution of collaborative agent systems. Although many feel that their capabilities are underwhelming, by integrating the principles of intelligent agent frameworks with advanced language models, their authors have made significant progress in creating AI systems that can collaborate, reason, and solve complex tasks.

A Deeper Dive into the Original BabyAGI:

BabyAGI presents an intuitive collaborative agent system operating within a loop, comprising three key agents: the Task Creator, Task Prioritizer, and Task Executor, each playing a unique role in the collaborative process. Let’s examine the prompts of each sub-agent.

Fig.4 Original task creator agent prompt

The process initiates with the Task Creator, responsible for defining the goal and initiating the task list. This agent in essence sets the direction for the collaborative system. It generates a list of tasks, providing a roadmap outlining the essential steps for goal attainment.

Fig 5. Original task prioritizer agent prompt

Once the tasks are established, they are passed on to the Task Prioritizer. This agent reorders tasks based on their importance for goal attainment, optimizing the system’s approach by focusing on the most critical steps. Ensuring the system maintains efficiency by directing its attention to the most consequential tasks.

Fig 6. Original task executor agent prompt

The Task Executor then takes over following task prioritization. This agent executes tasks one by one according to the prioritized order. As you may notice in the prompt, it is only just hallucinating and performing the tasks. The output of this prompt, the result of completing the task, is appended to the task object being completed and stored in a vector database.

An intriguing aspect of BabyAGI is the incorporation of a vector database, where the task object, including the Task Executor’s output, is stored. The reason this is important is that language models are static. They can’t learn from anything other than the prompt. Using a vector database to look up similar tasks allows the system to maintain a type of memory of its experiences, both problems and solutions, which helps improve the agent’s performance when confronted with similar tasks in the future.

Vector databases work by efficiently indexing the internal state of the language model.  For OpenAI’s text-embedding-ada-002 model, this internal state is a vector of length 1536. It is trained to produce similar vectors for semantically similar inputs, even if they use completely different words. In the BabyAGI system, the ability to look up similar tasks and append them to the context of the prompt is used as a way for the model to have memories of its previous experiences performing similar tasks.

As mentioned above, the vanilla version of BabyAGI operates predominantly in a hallucinating mode as it lacks external interaction. Additional tools, such as functions for saving text, interacting with databases, executing Python scripts, or even searching the web, were later integrated into the system, extending BabyAGI’s capabilities.

While BabyAGI is capable of breaking down large goals into small tasks and essentially working forever on them, it still has many limitations. Unless the task creator explicitly adds a check if a task is done, the system will tend to generate an endless stream of tasks, even after achieving the initial goal. Moreover, BabyAGI executes tasks sequentially, which slows it down significantly. Future iterations of BabyAGI, such as BabyDeerAGI, have implemented features to address these limitations, exploring parallel execution capabilities for independent tasks and more tools.

In essence, BabyAGI serves as a great introduction and starting point in the realm of collaborative agent systems. Its architecture enables planning, prioritization, and execution. It lays the groundwork for many other developers to create new systems to address the limitations and expand what’s possible.

The Rise of Role-Playing Collaborative Agent Systems:

While not every project claims BabyAGI as its inspiration, many similar multi-role agent systems exist in projects such as MetaGPT and AutoGen. These projects are bringing a new wave of innovation into this space. Much like how BabyAGI used multiple “Agents” to manage tasks, these frameworks go a step further. This is by trying to make many different agents with distinct roles that work together to accomplish the goal. In MetaGPT the agents are working together inside a virtual company, complete with a CEO, CTO, designers, testers, and programmers. People experimenting with this framework today can get this virtual company to create various types of simple utility software and simple games successfully. Though I would say they are rarely visually pleasing.

AutoGen is going about things slightly differently but in a similar vein to the framework I’ve been working on over at my company Xpress AI. 

AutoGen has a user proxy agent that interacts with the user and can create tasks for one or more assistant agents. The tool is more of a library than a standalone project so you will have to create a configuration of user proxies and assistants to accomplish the tasks you may have. I think that this is the future of how we will interact with agents. We will need those many conversation threads to interact with each other to expand the capabilities of the base model.

Why Collaborative Agents Systems are more effective

A language model is intelligent enough only by necessity. To predict the next work accurately, it has had to learn how to be rudimentarily intelligent. There is only a fixed amount of computation that can happen inside the various transformer layers inside the particular model. By giving the model a different starting point, it can put more computation and therefore thinking into its original response. Giving different roles to these specific agents helps them get out of the specific rut of wanting to be self-consistent. You can imagine how we can possibly go to an even larger scale on this idea to create AI systems closer to AGI.

Even in human society, it can be argued that we currently have various Superhuman intelligences in place. The stock market, for example, can allocate resources better than any one person could ever hope to. Take the scientific community, the paper review and publishing process are also helping humanity reach new levels of intelligence.

Even these systems need time to think or process the information. LLMs unfortunately only have a fixed amount of processing power. The future AI systems will have to include ways for the agent to think for itself, similar to how they can leverage functions today, but internally to give them the ability to apply an arbitrary amount of computation to achieve a task. Roles are one way to approach this, but it would be more effective if each agent in these simulated virtual organizations were able to individually apply arbitrary amounts of computation to their responses. Also, a system where each agent could learn from their mistakes, similar to humans, is required to really escape the cognitive limitations of the underlying language model. Without these capabilities, which have been known to the AI community as fundamental capabilities for a long time, we can’t reasonably expect these systems to be the foundation of an AGI.

Addressing Limitations and Envisioning Future Prospects:

Collaborative agent systems exhibit promising potential. However, they are still far from being truly general intelligence. Learning about these limitations can give clues to possible solutions that can pave the way for more sophisticated and capable systems. 

One limitation of BabyAGI in particular lies in the lack of active perception. The Executor Agent in BabyAGI nearly always assumes that the system is in the perfect state to accomplish the task, or that the previous task was completed successfully.  Since the world is not perfect it often fails to achieve the task. BabyAGI is not alone in this problem. The lack of perception greatly affects the practicality and efficacy of these systems for real-world tasks.

Error recovery mechanisms in these systems also need improvement. While a tool-enabled version of BabyAGI does often generate error-fixing tasks, the Task Prioritizer’s prioritization may not always be optimal. Causing the executor to miss the chance to easily fix the issue. Advanced prioritization algorithms, taking into account error severity and its impact on goal attainment are being worked on. The latest versions of BabyAGI have task dependency tracking which does help, but I don’t believe we have fully fixed this issue yet.

Task completion is another challenge in collaborative agent systems like BabyAGI. A robust review mechanism assessing the state of task completion and adjusting the task list accordingly could address the issue of endless task generation, enhancing the overall efficiency of the system. Since MetaGPT has managers that check the results of the individual contributors, they are more likely to detect that the task has been completed, although this way of working is quite inefficient.

Parallel execution of independent tasks offers another potential area of improvement. Leveraging multi-threading or distributed computing techniques could lead to significant speedups and more efficient resource utilization. BabyDeerAGI specifically uses dependency tracking to create independent threads of executors, while MetaGPT uses the company structure to perform work in parallel. Both are interesting approaches to the problem and, perhaps, the two approaches could be combined. 

The lack of the ability to learn from experience is another fundamental limitation. As far as I know, none of the current systems utilize fine-tuning of LLMs to form long-term memories. In theory, it isn’t a complicated process but in practice gathering the data necessary, in a way that doesn’t fundamentally make the model worse, is an open problem. Training models on model-generated outputs or training on already encountered data seems to cause the models to overfit quickly; often requiring careful hand-tuning of the training hyper-parameters. To make agents that can learn from experience, a sophisticated algorithm is required, not just to perform the training, but also to gather the correct data. This process is probably similar to the limbic system in our brains, for example.

While the current crop of agent systems has various limitations, there are still many open opportunities to address them with software and structure to create even more advanced applications. Enhancing active task execution, improving error recovery mechanisms, implementing efficient review mechanisms, and exploring parallel execution capabilities can boost the overall performance of these systems. 

Conclusion:

The emergence of open-source collaborative agent systems is creating a transformative era in AI. We are very close to a world where humans and AI can collaborate to solve the world’s problems. Similar to the idea of how companies or the market formed by many independent rational actors form a superhuman intelligence, the development of collaborative agent systems that have many independent sub-agents that communicate, collaborate, and reason together seems to enhance the capabilities of the language model alone to accomplish tasks, paving the way for the creation of more versatile applications.

Looking ahead, I think AI powered by collaborative agent systems has the potential to revolutionize industries such as healthcare, finance, education, and more. However, we must not forget the important sentence from an IBM manual: “A computer can never be held accountable”. In a future where we have human-level AIs that we can work hand-in-hand to tackle complex problems, it becomes increasingly important to ensure accountability measures are in place. The responsibility and accountability for their actions still ultimately lie with the humans who design, deploy, and use them. 

This journey towards AGI is thrilling, and collaborative agent systems play an integral role in this transformative era of artificial intelligence.

The post Talk of the AI Town: The Uprising of Collaborative Agents appeared first on ML Conference.

Mhairi Aitken: My name is Mhairi Aitken, I’m an ethics fellow at the Alan Turing Institute. The Alan Turing Institute is the UK’s National Institute for AI and data science and as an ethics fellow, I look at the ethical and social considerations around AI and data science. I work in the public policy program where our work is mostly focused on uses of AI within public policy and government, but also in relation to policy and government responses to AI as in regulation of AI and data science. 

devmio: For our readers who may be unfamiliar with the Alan Turing Institute, can you tell us a little bit about it? 

Mhairi Aitken: The national institute is publicly funded, but our research is independent. We have three main aims of our work. First, advancing world-class research and applying that to national and global challenges. 

Second, building skills for the future. That’s both going to technical skills and training the next generation of AI and data scientists, but also to developing skills around ethical and social considerations and regulation. 

Third, part of our mission is to drive an informed public conversation. We have a role in engaging with the public, as well as policymakers and a wide range of stakeholders to ensure that there’s an informed public conversation around AI and the complex issues surrounding it and clear up some misunderstandings often present in public conversations around AI.

NEW & PRACTICAL ENDEAVORS FOR ML

Machine Learning Principles

Learn more

devmio: In your talk at Devoxx UK, you said that it’s important to demystify AI. What exactly is the myth surrounding AI?

Mhairi Aitken: There’s quite a few different misconceptions. Maybe one of the biggest ones is that AI is something that is technically super complex and not something everyday people can engage with. That’s a really important myth to debunk because often there’s a sense that AI isn’t something people can easily engage with or discuss. 

As AI is already embedded in all our individual lives and is having impacts across society, it’s really important that people feel able to engage in those discussions and that they have a say and influence the way AI shapes their lives. 

On the other hand, there are unfounded and unrealistic fears about what risks it might bring into our lives. There’s lots of imagery around AI that gets repeated, of shiny robots with glowing brains and this idea of superintelligence. These widespread narratives around AI come back again and again, and are very present within the public discourse. 

That’s a distraction and it creates challenges for public engagement and having an informed public discussion to feed into policy and regulation. We need to focus on the realities of what AI is and in most cases, it’s a lot less exciting than superintelligence and shiny robots.

devmio: You said that AI is not just a complex technical topic, but something we are all concerned with. However, many of these misconceptions stem from the problem that the core technology is often not well understood by laymen. Isn’t that a problem?

Mhairi Aitken: Most of the players in big tech are pushing this idea of AI being something about superintelligence, something far-fetched, that’s closing down the discussions. It’s creating that sense that AI is something more difficult to explain, or more difficult to grasp, then it actually is, in order to have an informed conversation. We need to do a lot more work in that space and give people the confidence to engage in meaningful discussions around AI. 

And yes, it’s important to enable enough of a technical understanding of what these systems are, how they’re built and how they operate. But it’s also important to note that people don’t need to have a technical understanding to engage in discussions around how systems are designed, how they’re developed, in what contexts they’re deployed, or what purposes they are used for. 

Those are political, economic, and cultural decisions made by people and organizations. Those are all things that should be open for public debate. That’s why, when we talk about AI, it’s really important to talk about it as a human endeavor. It’s something which is created by people and is shaped by decisions of organizations and people. 

That’s important because it means that everyone’s voices need to be heard within those discussions, particularly communities who are potentially impacted by these technologies. But if we present it as something very complex which requires a deep technical understanding to engage with, then we are shutting down those discussions. That’s a real worry for me.

devmio: If the topic of superintelligence as an existential threat to humanity is a distraction from the real problems of AI that is being pushed by Big Tech, then what are those problems?

Mhairi Aitken: A lot of the AI systems that we interact with on a daily basis are opaque systems that make decisions about people’s lives, in everything from policing to immigration, social care and housing, or algorithms that make decisions about what information we see on social media. 

Those systems rely on or are trained on data sets, which contain biases. This often leads to biased or discriminatory outcomes and impacts. Because the systems are often not transparent in the ways that they’re used or have been developed, it makes it very difficult for people to contest decisions that are having meaningful impacts on their lives. 

In particular, marginalized communities, who are typically underrepresented within development processes, are most likely to be impacted by the ways these systems are deployed. This is a really, really big concern. We need to find ways of increasing diversity and inclusiveness within design and development processes to ensure that a diverse set of voices and experiences are reflected, so that we’re not just identifying harms when they occur in the real world, but anticipating them earlier in the process and finding ways to mitigate and address them.

At the moment, there are also particular concerns and risks that we really need to focus on concerning generative AI. For example, misinformation, disinformation, and the ways generative AI can lead to increasingly realistic images, as well as deep fake videos and synthetic voices or clone voices. These technologies are leading to the creation of very convincing fake content, raising real concerns for potential spread of misinformation that might impact political processes. 

It’s not just becoming increasingly hard to spot that something is fake. It’s also a widespread concern that it is increasingly difficult to know what is real. But we need to have access to trustworthy and accurate information about the world for a functioning democracy. When we start to question everything as potentially fake, it’s a very dangerous place in terms of interference in political and democratic processes.

I could go on, but there are very real concrete examples of how AI is already having presented harms today and they disproportionately impact marginalized groups. A lot of the narratives of existential risk we currently see are coming from Big Tech and are mostly being pushed by privileged or affluent people. When we think about AI or how we address the risks around AI, it’s important that we shouldn’t center around the voices of Big Tech, but the voices of impacted communities. 

devmio: A lot of misinformation is already on the internet and social media without the addition of AI and generative AI. So potential misuse on a large scale is of a big concern for democracies. How can western societies regulate AI, either on an EU-level or a global scale? How do we regulate a new technology while also allowing for innovation?

Mhairi Aitken: There definitely needs to be clear and effective regulation around AI. But I think that the dichotomy between regulation and innovation is false. For a start, we don’t just want any innovation. We want responsible and safe innovation that leads to societal benefits. Regulation is needed to make sure that happens and that we’re not allowing or enabling dangerous and harmful innovation practices.

Also, regulation provides the conditions for certainty and confidence for innovation. The industry needs to have confidence in the regulatory environment and needs to know what the limitations and boundaries are. I don’t think that regulation should be seen as a barrier to innovation. It provides the guardrails, clarity, and certainty that is needed. 

Regulation is really important and there are some big conversations around that at the moment. The EU AI Act is likely to set an international standard of what regulation will look like in this regard. It’s going to have a big impact in the same way that GDPR had with data protection. Soon, any organization that’s operating in the EU, or that may export an AI product to the EU, is going to have to comply with the EU AI Act. 

We need international collaboration on this.

devmio: The EU AI Act was drafted before ChatGPT and other LLMs became publicly available. Is the regulation still up to date? How is an institution like the EU supposed to catch up to the incredible advancements in AI?

Mhairi Aitken: It’s interesting that over the last few months, developments with large language models have forced us to reconsider some elements of what was being proposed and developed, particularly around general purpose AI. Foundation models like large language models that aren’t designed for a particular purpose can be deployed in a wide range of contexts. Different AI models or systems are built on top of them as a foundation.

That’s posed some specific challenges around regulation. Some of this is still being worked out. There are big challenges for the EU, not just in relation to foundation models. AI encompasses so many things and is used across all industries, across all sectors in all contexts, which poses a big challenge. 

The UK-approach to regulation of AI has been quite different to that proposed in the EU: The UK set out a pro-innovation approach to regulation, which was a set of principles intended to equip existing UK regulatory bodies to grapple the challenges of AI. It recognized that AI is already being used across all industries and sectors. That means that all regulators have to deal with how to regulate AI in their sectors. 

In recent weeks and months in the UK we have seen an increasing emphasis on regulation and AI, and increased attention at the importance of developing effective regulation. But I have some concerns that this change of emphasis has, at least in part, come from Big Tech. We’ve seen this in the likes of Sam Altman on his tour of Europe, speaking to European regulators and governments. Many voices talking about the existential risk AI poses come from Silicon Valley. This is now beginning to have an influence on policy discussions and regulatory discussions, which is worrying. It’s a positive thing that we’re having these discussions about regulation and AI, but we need those discussions to focus on real risks and impacts. 

devmio: The idea of existential threat posed by AI often comes from a vision of self-conscious AI, something often called strong AI or artificial general intelligence (AGI). Do you believe AGI will ever be possible?

Mhairi Aitken: No, I don’t believe AGI will ever be possible. And I don’t believe the claims being made about an existential threat. These claims are a deliberate distraction from the discussions of regulation of current AI practices. The claim is that the technology and AI itself poses a risk to humanity and therefore, needs regulation. At the same time, companies and organizations are making decisions about that technology. That’s why I think this narrative is being pushed, but it’s never going to be real. AGI belongs in the realm of sci-fi. 

There are huge advancements in AI technologies and what they’re going to be capable of doing in the near future is going to be increasingly significant. But they are still always technologies that do what they are programmed to do. We can program them to do an increasing number of things and they do it with an increasing degree of sophistication and complexity. But they’re still only doing what they’re programmed for, and I don’t think that will ever change. 

I don’t think it will ever happen that AI will develop its own intentions, have consciousness, or a sense of itself. That is not going to emerge or be developed in what is essentially a computer program. We’re not going to get to consciousness through statistics. There’s a leap there and I have never seen any compelling evidence to suggest that could ever happen.

We’re creating systems that act as though they have consciousness or intelligence, but this is an illusion. It fuels a narrative that’s convenient for Big Tech because it deflects away from their responsibility and suggests that this isn’t about a company’s decisions.

devmio: Sometimes it feels like the discussions around AI are a big playing field for societal discourse in general. It is a playing field for a modern society to discuss its general state, its relation to technology, its conception of what it means to be human, and even metaphysical questions about God-like AI. Is there some truth to this?

Mhairi Aitken: There’s lots of discussions about potential future scenarios and visions of the future. I think it’s incredibly healthy to have discussions about what kind of future we want and about the future of humanity. To a certain extent this is positive.

But the focus has to be on the decisions we make as societies, and not hypothetical far-fetched scenarios of super intelligent computers. These conversations that focus on future risks have a large platform. But we are only giving a voice to Big Tech players and very privileged voices with significant influence in these discussions. Whereas, these discussions should happen at a much wider societal level. 

The conversations we should be having are about how we harness the value of AI as a set of tools and technologies. How do we benefit from them to maximize value across society and minimize the risks of technologies? We should be having conversations with civil society groups and charities, members of the public, and particularly with impacted communities and marginalized communities.

We should be asking what their issues are, how AI can find creative solutions, and where we could use these technologies to bring benefit and advocate for the needs of community groups, rather than being driven by commercial for-profit business models. These models are creating new dependencies on exploitative data practices without really considering if this is the future we want.

devmio: In the Alan Turing Institute’s strategy document, it says that the institute will make great leaps in AI development in order to change the world for the better. How can AI improve the world?

Mhairi Aitken: There are lots of brilliant things that AI can do in the area of medicine and healthcare that would have positive impacts. For example, there are real opportunities for AI to be used in developing diagnostic tools. If the tools are designed responsibly and for inclusive practices, they can have a lot of benefits. There’s also opportunities for AI in relation to the environment and sustainability in terms of modeling or monitoring environments and finding creative solutions to problems.

One area that really excites me is where AI can be used by communities, civil society groups, and charities. At the moment, there’s an emphasis on large language models. But actually, when we think about smaller AI, there’s real opportunities if we see them as tools and technologies that we can harness to process complex information or automate mundane tasks. In the hands of community groups or charities, this can provide valuable tools to process information about communities, advocate for their needs, or find creative solutions.

devmio: Do you have examples of AI used in the community setting?

Mhairi Aitken: For example, community environment initiatives or sustainability initiatives can use AI to monitor local environments, or identify plant and animal species in their areas through image recognition technologies. It can also be used for processing complex information, finding patterns, classifying information, and making predictions or recommendations from information. It can be useful for community groups to process information about aspects of community life and develop evidence needed to advocate for their needs, better services, or for political responses.

A lot of big innovation is in commercially-driven development. This leads to commercial products instead of being about how these tools can be used for societal benefit on a smaller scale. This changes our framing and helps us think about who we’re developing these technologies for and how this relates to different kinds of visions of the future that benefit from this technology.

devmio: What do you think is needed to reach this point?

Mhairi Aitken: We need much more open public conversations and demands about transparency and accountability relating to AI. That’s why it’s important to counter the sensational unrealistic narrative and make sure that we focus on regulation, policy and public conversation. All of us must focus on the here and now and the decisions of companies leading the way in order to hold them accountable. We must ensure meaningful and honest dialogue as well as transparency about what’s actually happening.

devmio: Thank you for taking the time to talk with us and we hope you succeed with your mission to inform the public.

The post AI is a Human Endeavor appeared first on ML Conference.

I first encountered the CLIP model in early 2022 while experimenting with the AI drawing model. CLIP (Contrastive Language-Image Pre-Training) is a model proposed by OpenAI in 2021. CLIP can encode images and text into representations that can be compared in the same space. CLIP is the basis for many text-to-image models (e.g. Stable Diffusion) to calculate the distance between the generated image and the prompt during training.

Fig. 1: OpenAI’s CLIP model, source: https://openai.com/blog/clip/

As shown above, the CLIP model consists of two components: Text Encoder and Image Encoder. Let’s take the ViT-B-32 (different models have different output vector sizes) version as an example:

• Text Encoder can encode any text (length <77 tokens) into a 1×512 dimensional vector.
• Image Encoder can encode any image into a 1×512 dimensional vector.

By calculating the distance or cosine similarity between the two vectors, we can compare the similarity between a piece of text and an image.

Image Search on a Server

I found this to be quite fascinating, as it was the first time images and text could be compared in this way. Based on this principle, I quickly set up an image search tool on a server. First, process all images through CLIP, obtaining their image vectors, they should be a list of 1×512 vectors.

  # get all images list.  

img_lst = glob.glob(‘imgs/*jpg’) 

img_features = [] 

# calculate vector for every image.  

for img_path in img_lst: 

image = preprocess(Image.open(img_path)).unsqueeze(0).to(device) 

image_features = model.encode_image(image) 

img_features.append(image_features)

Then, given the search text query, calculate its text vector (with a size of 1×512) and compare similarity with each image vector in a for-loop.

  text_query = ‘lonely’  

# tokenize the query then put it into the CLIP model.  

text = clip.tokenize([text_query]).to(device)  

text_feature = model.encode_text(text)  

# compare vector similary with each image vector  

sims_lst = []  

for img_feature in img_features:  

sim = cosin_similarity(text_feature, img_feature)  

sims_lst.append(sim.item())  

Finally, display the top K results in order. Here I return the top3 ranked image files, and display the most relevant result.

  K = 3 

# sort by score with np.argsort  

sims_lst_np = np.array(sims_lst) 

idxs = np.argsort(sims_lst_np)[-K:] 

# display the most relevant result.  

imagedisplay(filename=img_lst[idxs[-1]])

I discovered that its image search results were far superior to those of Google, here are the top 3 results when I search for the keyword “lonely”:

Integrating CLIP into iOS with Swift

After marveling at the results, I wondered: Is there a way to bring CLIP to mobile devices? After all, the place where I store the most photos is neither my MacBook Air nor my server, but rather my iPhone.

To port a large GPU-based model to the iPhone, operator support and execution efficiency are the two most critical factors.

1. Operator Support

Fortunately, in December 2022, Apple demonstrated the feasibility of porting Stable Diffusion to iOS, proving that the deep learning operators needed for CLIP are supported in iOS 16.0.

Fig. 2: Pictures generated by Stable Diffusion

2. Execution Efficiency

Even with operator support, if the execution efficiency is extremely slow (for example, calculating vectors for 10,000 images takes half an hour, or searching takes 1 minute), porting CLIP to mobile devices would lose its meaning. These factors can only be determined through hands-on experimentation.

I exported the Text Encoder and Image Encoder to the CoreML model using the coremltools library. The final models has a total file size of 300MB. Then, I started writing Swift code.

I use Swift to load the Text/Image Encoder models and calculate all the image vectors. When users input a search keyword, the model first calculates the text vector and then computes its cosine similarity with each of the image vectors individually.

The core code is as follows:

  // load the Text/Image Encoder model. 

let text_encoder = try MLModel(contentsOf: TextEncoderURL, configuration: config) 

let image_encoder = try MLModel(contentsOf: ImageEncoderURL, configuration: config) 

// given a prompt/photo, calculate the CLIP vector for it. 

let text_feature = text_encoder.encode(“a dog”) 

let image_feature = image_encoder.encode(“a dog”) 

// compute the cosine similarity. 

let sim = cosin_similarity(img_feature, text_feature) 

As a SwiftUI beginner, I found that Swift doesn’t have a specific implementation for cosine similarity. Therefore, I used Accelerate to write one myself, the code below is a Swift translation of cosine similarity from Wikipedia.

  import Accelerate  

func cosine_similarity(A: MLShapedArray<Float32>, B: MLShapedArray<Float32>) -> Float {  

let magnitude = vDSP.rootMeanSquare(A.scalars) * vDSP.rootMeanSquare(B.scalars)  

let dotarray = vDSP.dot(A.scalars, B.scalars)  

return  dotarray / magnitude  

} 

The reason I split Text Encoder and Image Encoder into two models is because, when actually using this Photos search app, your input text will always change, but the content of the Photos library is fixed. So all your image vectors can be computed once and saved in advance. Then, the text vector is computed for each of your searches.

Furthermore, I implemented multi-core parallelism when calculating similarity, significantly increasing search speed: a single search for less than 10,000 images takes less than 1 second. Thus, real-time text searching from tens of thousands of Photos library becomes possible.

Below is a flowchart of how Queryable works:

Fig. 3: How the app works

Performance

But, compared to the search function of the iPhone Photos, how much does the CLIP-based album search capability improve? The answer is: overwhelmingly better. With CLIP, you can search for a scene in your mind, a tone, an object, or even an emotion conveyed by the image.

Fig. 4: Search for a scene, an object, a tone or the meaning related to the photo with Queryable.

To use Queryable, you need to first build the index, which will traverse your album, calculate all the image vectors and store them. This takes place only once, the total time required for building the index depends on the number of your photos, the speed of indexing is ~2000 photos per minute on the iPhone 12 mini. When you have new photos, you can manually update the index, which is very fast.

In the latest version, you have the option to grant the app access to the network in order to download photos stored on iCloud. This will only occur when the photo is included in your search results, the original version is stored on iCloud, and you have navigated to the details page and clicked the download icon. Once you grant the permissions, you can close the app, reopen it, and the photos will be automatically downloaded from iCloud.

3. Any requirements for the device?

• iOS 16.0 or above
• iPhone 11 (A13 chip) or later models

The time cost for a search also depends on your number of photos: for <10,000 photos it takes less than 1s. For me, an iPhone 12 mini user with 35,000 photos, each search takes about 2.8s.

Q&A on Queryable

1.On Privacy and security issues.

Queryable is designed as an OFFLINE app that does not require a network connection and will never request network access, thereby avoiding privacy issues.

2. What if my pictures are stored on iCloud?

Due to the inability to connect to a network, Queryable can only use the cache of the low-definition version of your local Photos album. However, the CLIP model itself resizes the input image to a very small size (e.g. ViT-B-32 is 224×224), so if your image is stored on iCloud, it actually does not affect search accuracy except that you cannot view its original image in search result.

The post Using OpenAI’S CLIP Model on the iPhone: Semantic Search For Your Own Pictures appeared first on ML Conference.

**Christoph Schuhmann:** LAION stands for Large-Scale Artificial Intelligence Open Network. First and foremost, it’s simply a huge community of people who share the dream of open-source AI models, research, and datasets. That’s what connects us all. We have a [Discord server](https://discord.com/invite/xBPBXfcFHd) where anyone can come in and share a bit about the latest research in the field. You can also propose a new project and find people to work on it with you. And if you ask the mods, me, or other people, you might even get a channel for your project. That’s basically the core.

When we had such surprising success with our first dataset called [LAION-400M](https://laion.ai/blog/laion-400-open-dataset/), we set up a small non-profit association that doesn’t actually do anything. We have a bank account with a bit of money coming into it from a few companies that support us. That’s primarily Hugging Face, but also StabilityAI, although we’re mostly supported not by money but by cloud compute.

StabilityAI, for example, has a huge cluster with 4000 or now 5600 GPUs, and there we or our members who are approved by the core team can use preemptable GPUs, for example, what is not being used at the moment and is idle.

**devmio: So we can just come to you and contribute? Propose our ideas and ask for help with our projects or help with ongoing projects?**

**Christoph Schuhmann:** Exactly! You can now come to our Discord server and say that you want to contribute to a project or help us with PR or whatever. You are most welcome!

**devmio: Is LAION based in Germany? And you are the chairman and co-founder?**

**Christoph Schuhmann:** Exactly. I am a physics and computer science teacher, I have been regularly involved with machine learning, and I also have a background in reform-oriented education. I made a Kickstarter documentary seven or eight years ago about schools where you can learn without grades and curriculum. After that took off, I did tutorials on how to start such an independent school. So I knew how to set up a grassroots non-profit organization. I am not paid for my work at LAION.

## The Beginnings of LAION

**devmio: How did LAION come to life? How did you get to know the other members?**

**Christoph Schuhmann:** I actually started LAION after reading a lot about deep learning and machine learning and doing online courses in my spare time over the last five to six years. When the first version of DALL-E was published at the beginning of 2021, I was totally shocked by how good it was. At that time, however, many non-computer scientists didn’t find it that impressive.

I then asked on a few Discord servers about machine learning and what we would need to replicate something similar and make it open-source. There was a well-known open-source programmer at the time called Philip Wang (his alias on GitHub is lucidrains) who is a legend in the community because whenever a new paper comes out he has the associated codebase implemented within a few days. He also built an implementation of the first version of DALL-E in Pytorch called [DALLE-pytorch](https://github.com/lucidrains/DALLE-pytorch). This model was then trained by a few people using small data sets on Discord, and that was proof of concept.

But the data was missing, and I suggested going to [Common Crawl](https://commoncrawl.org/), a non-profit from Seattle that scraps HTML code from the internet every two to three months and makes it available. A snapshot, so to speak, of the HTML code of all possible websites, which is 250 terabytes zip file. I then suggested downloading a gigabyte as a test and wrote a script that extracts image tags together with alt tags and then uses the CLIP model to see how well they fit together.

Then two “machine learning nerds”, who were much better at it than I was at the time, implemented it efficiently but didn’t finish it. That was a shame, but they were developing the GPT open-source variant [GPT-J](https://huggingface.co/docs/transformers/model_doc/gptj) and therefore didn’t have the time.

Then in the spring of 2021, I sat down and just wrote down a huge spaghetti code in a Google Colab and then asked around on Discord who wanted to help me with it. Someone got in touch, who later turned out to be only 15 at the time. And he wrote a tracker, basically a server that manages lots of colabs, each of which gets a small job, extracts a gigabyte, and then uploads the results. At that time, the first version was still using Google Drive.

## The Road to the LAION-400M Dataset

It was a complete disaster because Google Drive wasn’t suitable for it, but it was the easiest thing we could do quickly. Then I looked for some people on a Discord server, made some more accounts, and then we ended up with 50 Google Colabs working all the time.

But it worked, and then, within a few weeks, we had filtered 3 million image-text pairs, which at the time was more than Google’s [Conceptual Captions](https://ai.google.com/research/ConceptualCaptions/), a very well-known dataset of 2019. That little success got us so much attention on the Discord server that people just started supporting us and writing things like, “I have 50 little virtual machines here from my work, you could use them, I don’t need them right now,” or “I have another 3090 lying around here with me, I can share it with you.”

After three months, we had 413 million filtered image-text pairs. That was our LAION-400M dataset. At the time, it was by far the largest image-text dataset freely available, over 30 times larger than [Google’s Conceptual Caption 12M](https://github.com/google-research-datasets/conceptual-12m), with about 12 million pairs.

We then did a [blog post about our dataset](https://laion.ai/blog/laion-400-open-dataset/), and after less than an hour, I already had an email from the Hugging Face people wanting to support us. I had then posted on the Discord server that if we had $5,000, we could probably create a billion image-text pairs. Shortly after, someone already agreed to pay that: “If it’s so little, I’ll pay it.” At some point, it turned out that the person had his own startup in text-to-image generation, and later he became the chief engineer of Midjourney.

As you can see, it was simply a huge community, just 100 people who only knew each other from chat groups with aliases. At some point, I made the suggestion to create an association, with a banking account, etc. That’s how LAION was founded.

## Even Bigger: LAION-5B and LAION-Aesthetics

We then also got some financial support from Hugging Face and started working on LAION-5B, which is a dataset containing five billion image-text pairs. By the end of 2021, we were done with just under 70 percent of it, and then we were approached by someone who wanted to create a start-up that was like OpenAI but really open-source. He offered to support us with GPUs from AWS. This was someone who introduced himself as a former investment banker or hedge fund manager, which I didn’t quite believe at first. In the end, it was just some guy from Discord. But then the access data for the first pods came, and it turned out that the guy was Emad Mostaque, the founder of StabilityAI.

**devmio: What is the relationship between LAION and Stability AI?**

**Christoph Schuhmann:** Contrary to what some AI-art critics claim, we are not a satellite organisation of Stability AI. On the contrary, Stability AI came to us after the LAION-5B dataset was almost finished and wanted to support us unconditionally. They then did the same with LAION-Aesthetics.

**devmio: Could you explain what LAION-Aesthetics is?**

**Christoph Schuhmann:** I trained a model that uses the CLIP embeddings of the LAION images to estimate how pretty the images are on a scale of one to ten. It’s a very small model, a multilayer perceptron running on a CPU. At some point, I ran the model over a couple of 100,000 images, sorted them, and thought that the ones with the high scores looked really good. The next step was to run it on 2.3 billion CLIP embeddings.

## From LAION-Aesthetics to Stable Diffusion

**devmio: How did LAION-Aesthetics help with the development of Stable Diffusion?**

**Christoph Schuhmann:** I had already heard about Robin Rombach, who was still a student in Heidelberg at the time and had helped develop latent diffusion models at the CompVis Group. Emad Mostaque, the founder of StabilityAI, told me in May 2022 that he would like to support Robin Rombach with compute time, and that’s how I got in touch with Robin.

I then sent him the LAION-Aesthetics dataset. The dataset can be thought of as a huge Excel spreadsheet containing links to images and the associated alt text. In addition, each image is given a score, such as whether something contains a watermark or smut. Robin and his team later trained the first prototype of Stable Diffusion on this. However, the model only got the name Stable Diffusion through Stability AI, to whom the model then migrated.

LAION also got access to the Stability AI cluster. But we were also lucky enough to be able to use JUWELS, one of the largest European supercomputers, because one of our founding members, Jenia Jitsev, is the lab director at the Jülich Supercomputer Center for Deep Learning. We then applied for compute time to train our own OpenCLIP models. And now we have the largest CLIP models available in open source.

## LAION’s OpenCLIP

**devmio: What exactly do CLIP models do? And what makes LAION’s OpenCLIP so special?**

**Christoph Schuhmann:** On the Stability AI cluster, a Ph.D. student from UC Washington has trained a model called CLIP-ViT-G. This model can tell you how well an image matches a text, and this model has managed to crack the 80 percent zero-shot mark. This means that we have now built a general-purpose AI model that is better than the best state-of-the-art models from five years ago that were built and trained specifically for this purpose.

These CLIP models are in turn used as text encoders, as “text building blocks” by Stable Diffusion and by many other models. CLIP models have an incredible number of applications. For example, they can be used for zero-shot image segmentation, zero-shot object detection with bounding boxes, zero-shot classification, or even for text-to-image generation.

We have trained and further developed these models. We now have a variant that not only trains these CLIP models but also generates captions through a text decoder. This model is called [CoCa](https://laion.ai/blog/coca/) and is quite close to the state of the art.

We have many such projects running at the same time, sometimes so many that I almost lose track of them. Currently, we cooperate with Mila, an institute of excellence from Montreal, and together we have access to the second largest supercomputer in the US, Summit. We have been given 6 million GPU hours there and are training all kinds of models.

**devmio: You have already talked a lot about Stable Diffusion, and Robin Rombach, the inventor, is a member of your team. Is Stable Diffusion managed by you, is that “your” model?**

**Christoph Schuhmann:** No, we don’t have anything to do with that for now. But we have made the development and training of Stable Diffusion easier with LAION-Aesthetics and LAION-5B.

## Open Source as a Superpower

**devmio: LAION is committed to making the latest developments in AI freely available. Why is open source so important in AI?**

**Christoph Schuhmann:** Let’s take the sentence: “AI should be open source so that it is available to the general public.” Now let’s take that sentence and replace “AI” with “superpowers”: “Superpowers should be open source and available to the public.” In this case, it becomes much more obvious what I’m actually getting at.

Imagine if there was such a thing as superpowers, and only OpenAI, Microsoft, Google, maybe the Chinese and American governments, and five other companies, have control over it and can decide what to do with it. Now, you could say that governments only ever want what’s best for their citizens. That’s debatable, of course, but let’s assume that’s the case. But does that also apply to Microsoft? Do they also have our best interests at heart, or does Microsoft simply want to sell its products?

If you have a very dark view of the world, you might say that there are a lot of bad people out there, and if everyone had superpowers now, there would certainly be 10, 20, or 30 percent of all people who would do really bad things. That’s why we have to control such things, for example through the state. But if you have a rather positive and optimistic view of the world, like me, for example, then you could say that most people are relatively nice. No angels, no do-gooders, but most people don’t want to actively do something bad, or destroy something, but simply live their lives. There are some people who are do-gooders and also people who have something bad in mind. But the latter are probably clearly in the minority.

If we assume that everyone has superpowers, then everyone would also have the opportunity to take action against destructive behaviour and limit its effects. In such a world, there would be a lot of positive things. Things like superpower art, superpower music, superpower computer games, and superpower productivity of companies that simply produce goods for the public. If you now ask yourself what kind of world you would like to live in and assume that you have a rather positive worldview, then you will probably decide that it would be good to make superpowers available to the general public as open source. And once you understand that, it’s very easy to understand that AI should also be open source.

AI is not the same as superpowers, of course, but in a world in which the internet plays an ever greater role, in which every child grows up with YouTube, in which AI is getting better and better, in which more and more autonomous systems are finding their way into our everyday lives, AI is incredibly important. Software and computerised things are sort of superpowers. And that’s going to get much more blatant, especially with ChatGPT. In three to four years, ChatGPT will be much better than it is today.

Now imagine if the whole world used technologies like ChatGPT and only OpenAI and Microsoft, Google and maybe two or three other big companies controlled those technologies. They can cut you off at any time, or tell you “Sorry, but I can’t do this task, it’s unethical in my opinion”, “I have to block you for an hour now”, or “Sorry, your request might be in competition with a Microsoft product, now I have to block you forever. Bye.”

**devmio: We had also spoken to other experts, for example, Pieter Buteneers and Christoph Henkelkmann, who had similar concerns. But the question remains whether everyone should really have unrestricted access to such technologies, right?**

**Christoph Schuhmann:** A lot of criticism, not directed at LAION but at Stable Diffusion, goes in this direction. There is criticism that there are open-source models like Stable Diffusion that can be used to create negative content, circumvent copyright and create fakes, etc. Of course, it’s wrong to violate copyright, and it’s also wrong to create negative content and fakes. But imagine if these technologies were only in the hands of Microsoft, Google, and a few more large research labs. They would develop really well in the background, and at some point, you would be able to generate everything perfectly with them. And then they leak out or there is a replica, and society is not prepared at all. Small and medium-sized university labs wouldn’t be prepared at all to look at the source code and discover the problems.

We have something similar with LAION-5B. There are also some questionable images in the dataset that we were unable to filter. As a result, there is also a disclaimer that it is a research dataset that should be thoroughly filtered and examined before being used in production. You have to handle this set carefully and responsibly. But this also means that you can find things in the set that you would like to remove from the internet.

For example, there is an organisation of artists, [Have I Been Trained](https://haveibeentrained.com/), that provides a tool that artists can use to determine if their artwork is included in LAION-5B. This organisation has simply taken our open-source code and used it for their own purposes to organise the disappointed artists.

And that’s a great thing because now all those artists who have images on the internet that they don’t want there can find them and have them removed. And not only artists! For example, if I have a picture of myself on the internet that I don’t want there, I can find out through LAION-5B where it is being used. We don’t have the images stored in LAION-5B, we just have a table with the links, it’s just an index. But through that, you can find out which URL is linked to the image and then contact the owners of the site and have the image removed. By doing this, LAION generates transparency and gives security researchers an early opportunity to work with these technologies and figure out how to make them more secure. And that’s important because this technology is coming one way or another.

In probably a lot less than five years, you’re going to be able to generate pretty much anything in terms of images that you can describe in words, photo-realistically, so that a human being with the naked eye can’t tell whether it’s a photo or not.

## AI in Law, Politics, and Society

**devmio: Because you also mentioned copyright: The legal situation in Germany regarding AI, copyright, and other issues is probably not entirely clear. Are there sufficient mechanisms? Do you think that the new EU regulations that are coming will be sufficient while not hindering creativity and research?**

**Christoph Schuhmann:** I am not a lawyer, but we have good lawyers advising us. There is a Data Mining Law, an EU-wide exception to copyright. It allows non-profit institutions, such as universities, but also associations like ours, whose focus is on research and who make their results publicly available, to download and analyse things that are openly available on the internet.

We are allowed to temporarily store the links, texts, whatever, and when we no longer need them for research, we have to delete them. This law explicitly allows data mining for research, and that is very good. I don’t think all the details of what’s going to happen in the future, especially with ChatGPT and other generative AIs for text and images, were anticipated in these laws. The people who made the law probably had more statistical analysis of the internet in mind and less training data for AIs.

I would like to see more clarity from legislators in the future. But I think that the current legal situation in Germany is very good, at least for non-profit organisations like LAION. I’m a bit worried that when the [EU AI Act](https://digital-strategy.ec.europa.eu/de/policies/european-approach-artificial-intelligence), which is being drafted, comes, something like general purpose AI, like ChatGPT, would be classified as high risk. If that were to be the case, it would mean that if you as an organisation operate or train a ChatGPT-like service, you would have to constantly account for everything meticulously and tick off a great many compliance rules, catalogues, and checklists.

Even if this is certainly well-intentioned, it would also extremely restrict research and development, especially of open source, associations, and of grassroots movements, so that only Big Tech Corporate would be able to comply with all the rules. Whether this will happen is unclear so far. I don’t want high-risk applications like facial recognition to go unregulated either. And I don’t want to be monitored all day.

But if any lawmakers are reading this: Politicians should keep in mind that it is very important to continue to enable open-source AI. It would be very good if we could continue to practice as we have been doing. Not only for LAION but for Europe. I am sure that quite a lot of companies and private people, maybe even state institutions can benefit from such models as CLIP or from the datasets that we are making.

And I believe that this can generate a lot of value for citizens and companies in the EU. So I would even go so far as to call for politicians and donors to maybe think about building something similar to a CERN for AI. With a billion euros, you could probably build a great open-source supercomputer that all companies and universities, in fact, anyone, could use to do AI research under two conditions: First, the whole thing has to be reviewed by some smart people, maybe experts and people from the open-source community. Second, all results, research papers, checkpoints of models, and datasets must be released under a fully open-source licence.

Because then a lot of companies that can’t afford a supercomputer at the moment could open source their research there and only keep the fine-tuning or anything that is really sensitive to the business model on the companies’ own computers. But all the other stuff happens openly. That would be great for a lot of companies, that would be great for a lot of medium and small universities, and that would also be great for groups like LAION.

_**Editor’s note**: After the interview, LAION started a petition for a CERN-like project. Read more on [LAION’s blog](https://laion.ai/blog/petition/)._

## AI for a Better World

**Christoph Schuhmann:** Another application for AI would be a project close to my heart: Imagine there is an open-source ChatGPT. You would then take, say, 100 teachers and have them answer questions from students about all sorts of subjects. For these questions, you could make really nice step-by-step explanations that really make sense. And then, you would collect data from the 100 teachers for the school material up to the tenth grade. That’s at least similar everywhere in the Western world, except, of course, history, politics, etc. But suppose you were to simply break down the subject matter from 100 countries, from 100 teachers, from the largest Western countries, and use that to fine-tune a ChatGPT model.

You need a model that has maybe 20 to 30 billion parameters, and you could use it to give access to first-class education to billions of children in the Third World who don’t have schools but have an old mobile phone and internet access. You don’t need high-tech future technology, you can do that with today’s technology. And these are big problems of the world that could be addressed with it.

Or another application: My mum is 83 years old, she can’t handle a computer and is often lonely. Imagine if she had a Siri that she could have a sensible conversation with. Not as a substitute for human relationships, but as a supplement. How many lonely old people do you think would be happier if they could just ask what’s going on in the world. Or “Remember when I told you that story, Siri? Back in my second marriage 30 years ago?” That would make a lot of people happier. And I think things like that can have a lot of effect with relatively little financial outlay.

**devmio: And what do you see next in AI development?**

**Christoph Schuhmann:** What I just talked about could happen in the next five years. Everything that happens after that,  I can’t really predict. It’s going to be insane.

**devmio: Thank you very much for taking the time to talk to us!**

The post AI as a Superpower: LAION and the Role of Open Source in Artificial Intelligence appeared first on ML Conference.

On the flip side, ChatGPT has been roundly criticized for its inability to answer simple logic questions or work backwards from a desired solution to the steps needed to achieve it. Teachers and school administrators voiced fears that students would use the tool to cheat, while political conservatives complained that Chat generates answers with a liberal bias. Elon Musk, Apple co-founder Steve Wozniak, and others signed an open letter recommending a six-month pause in AI development, noting “Powerful AI systems should be developed only once we are confident that their effects will be positive and their risks will be manageable.”

The one factor missing from virtually all these comments – regardless of whether they regard ChatGPT as a huge step forward or a threat to humanity – is a recognition that no matter how impressive, ChatGPT merely gives the illusion of understanding. It is simply manipulating symbols and code samples which it has pulled from the Internet without any understanding of what they mean. And because it has no true understanding, it is neither good nor bad. It is simply a tool which can be manipulated by humans to achieve certain outcomes, depending on the intentions of the users.

It is that difference that distinguishes ChatGPT, and all other AI for that matter, from AGI – artificial general intelligence, defined as the ability of an intelligent agent to understand or learn any intellectual task that a human can. While ChatGPT undoubtedly represents a major advance in self-learning AI, it is important to recognize that it only seems to understand. Like all other AI to date, it is completely reliant on datasets and machine learning. ChatGPT simply appears more intelligent because it depends on bigger and more sophisticated datasets.

RETHINK YOUR APPROACHES

Business & Strategy

Learn more

While some experts continue to argue that at some point in the future, AI will morph into AGI, that outcome seems highly unlikely. Because today’s AI is entirely dependent on massive data sets, there is no way to create a dataset big enough for the resulting system to cope with completely unanticipated situations. In short, AI has no common sense and we simply can’t store enough examples to handle every possible situation. Further, AI, unlike humans, is unable to merge information from multiple senses. So while it might be possible to stitch language and image processing applications together, researchers have not found a way to integrate them in the same seamless way that a child integrates vision, language, and hearing.

For today’s AI to advance to something approaching real human-like intelligence, it must have three essential components of consciousness: an internal mental model of surroundings with the entity at the center; a perception of time which allows for a prediction of future outcome(s) based on current actions; and an imagination so that multiple potential actions can be considered and their outcomes evaluated and chosen. Just like the average three-year-old child, it must be able to explore, experiment, and learn about real objects, interpreting everything it knows in the context of everything else it knows.

To get there, researchers must shift their reliance on ever-expanding datasets to a more biologically plausible system modelled on the human brain, with algorithms that enable it to build abstract “things” with limitless connections and context.

While we know a fair amount about the brain’s structure, we still don’t know what fraction of our DNA defines the brain or even how much DNA defines the structure of its neocortex, the part of the brain we use to think. If we presume that generalized intelligence is a direct outgrowth of the structure defined by our DNA and that structure could be defined by as little as one percent of that DNA, though, it is clear that AGI emergence depends not on more computer power or larger data sets but on what to write as the fundamental AGI algorithms.

With that in mind, it seems highly likely that a broader context that is actually capable of understanding and learning gradually could emerge if all of today’s AI systems could be built on a common underlying data structure that allowed their algorithms to begin interacting with each other. As these systems become more advanced, they would slowly begin to work together to create a more general intelligence that approaches the threshold for human-level intelligence, enabling AGI to emerge. To make that happen, though, our approach must change. Bigger and better data sets don’t always win the day.

The post ChatGPT and Artificial General Intelligence: The Illusion of Understanding appeared first on ML Conference.

AI Alignment [1] is concerned with aligning AI to desired goals. The first challenge here is to agree on these goals in the first place. The next difficulty is that it is not (yet?) possible to give these goals directly and explicitly to an AI system.

For example, Amazon developed a system several years ago that helps select suitable applicants for open positions ([2], [3]). For this, resumes of accepted and unaccepted applicants were used to train an AI system. Although they contained no explicit information about gender, male applicants were systematically preferred. We will discuss how this came about in more detail later. But first, this raises several questions: Is this desirable, or at least acceptable? And if not, how do you align the AI system so that it behaves as you want it to? In other words, how do you successfully engage in AI alignment?

For some people, AI Alignment is an issue that will become more important in the future when machines are so intelligent and powerful that they might think the world would be better without humans [4]. Nuclear war provoked by supervillains is mentioned as another possibility of AI’s fatal importance. Whether these fears could ever become realistic remains speculation.

The claims being discussed as part of the EU’s emerging AI regulation are more realistic. Depending on what risk is realistically inherent in an AI system, different regulations may be applied here. This is shown in Figure 1, which is based on a presentation for the EU [5]. Four ranges from “no risk” to “unacceptable risk” are distinguished. In this context, a system with no significant risk only has the recommendation of a “Code of Conduct”, while a social credit system, as applied in China [6], is simply not allowed. However, this scheme only comes into effect if there is no specific law.

Fig. 1: Regulation based on outgoing risk, adapted from [5]

Alignment in Machine Learning Systems

A machine learning system is trained using sample data. It learns to mimic this sample data. In the best and most desirable case, the system can generalize beyond this sample data and recognizes an abstract pattern behind it. If this succeeds, the system can also react meaningfully to data that it has never seen before. Only then can we speak of learning or even a kind of understanding that goes beyond memorization.

This also happened in the example of Amazon’s applicant selection, as shown in a simplified form in Figure 2.

Fig. 2: How to learn from examples, also known as supervised learning

Here is another example. We use images of dogs and cats as sample data for a system, training it to distinguish between them. In the best case, after training, the system also recognizes cats that are not contained in the training data set. It has learned an abstract pattern of cats, which is still based on the given training data, however.

Therefore, this system can only reproduce what already exists. It is descriptive or representative, but hardly normative. In the Amazon example, it replicates past decisions. These decisions seemed to be that men simply had a better chance of being accepted. So, at least the abstract model would be accurate. Alternatively, perhaps there were just more examples of male applicants, or some other unfortunate circumstance caused the abstract model not to be a good generalization of the example data.

At its best, however, such an approach is analytical in nature. It shows the patterns of our sample data and their backgrounds, meaning that men performed better on job applications. If that matches our desired orientation, there is no further problem. But what if it doesn’t? That’s what we’re assuming and Amazon was of that opinion as well, since they scrapped the system.

MYRIAD OF TOOLS & FRAMEWORKS

Tools, APIs & Frameworks

Learn more

Pre-assumptions, aka: Priors

How to provide a machine learning system additional information about our desired alignment in addition to sample data has been commonly understood for a long time. This is used to provide world or domain knowledge to the system to guide and potentially simplify or accelerate training. You support the learning process by specifying which domain to look for abstract patterns in the data. Therefore, a good abstract pattern can be learned even if the sample data describes it inadequately. In machine learning, data being an inadequate description of the desired abstract model is the rule, rather than the exception. Yann LeCun, a celebrity on the scene, vividly elaborates on this in a Twitter thread [7].

  This kind of previous assumption is also called a prior. An illustrative example of a prior is linearity. As an explanation, let’s take another application example. For car insurance, estimating accident risk is crucial. For an estimation, characteristics of the drivers and vehicles to be insured are collected. These characteristics are correlated with existing data on accident frequency in a machine-learning model. The method used for this is called supervised learning, and it is  the same as described above.

For this purpose, let us assume that the accident frequency increases linearly with increased distance driven. The more one drives, the more accidents occur. This domain knowledge can be incorporated into the training process. This way, you can hope for a simpler model and potentially even less complex training. In the simplest case, linear regression [8] can be used here, which produces a reasonable model even with little training data or effort. Essentially, training consists of choosing the parameters for a straight line, slope, and displacement, to best fit the training data. Because of its simplicity, the advantage of this model is its good explainability and low resource requirement. This is because a linear relationship, “one-to-one”,  is intellectually easy, and a straight-line equation can be calculated on a modern computer with extremely little effort.

However, it is also possible to describe the pattern contained in the training data and correct it normatively. For this, let us assume that the relationship between age and driving ability is clearly over-linear. Driving ability does not decline in proportion to age, but at a much faster rate. Or, to put it another way, the risk of accidents increases disproportionately with age. That’s how it is in the world, and that’s what the data reflects. Let’s assume that we don’t want to give up on this important influence completely. However, we equally want to avoid excessive age discrimination. Therefore, we decide to allow a linear dependence at most. We can support the model and align it with our needs. This relationship is illustrated in Figure 3. The simplest way to implement this is the aforementioned linear regression.

Fig. 3: Normative alignment of training outcomes

Now, you could also argue that models usually have not only one input, but many, which act in combination on the prediction. Moreover, in our example, the linear relationship between distance driven and accident frequency does not need to be immediately plausible. Don’t drivers with little driving experience have a higher risk? In that case, you could imagine a partial linear relationship. In the beginning, the risk decreases in relation to the distance driven, but then it increases again after a certain point and remains linear. There are also tools for these kinds of complex correlations. In the deep learning field, TensorFlow Lattice [9] offers the possibility of specifying a separate set of alignments for each individual influencing factor. This is also possible in a nonlinear or only partially linear way. 

In addition to these relatively simple methods, there are other ways to influence. These include the learning algorithms you choose, the sample data selected, and, especially in deep learning, the neural network’s architecture and learning approach. These interventions in the training process are technically challenging and must be performed sparingly under supervision. Depending on the training data, otherwise, it may become impossible to train a good model with the desired priors. 

Is all this not enough? Causal Inference

The field of classical machine learning is often accused of falling short. People say that these techniques are suitable for fitting straight lines and curves to sample data, but not for producing intelligent systems that behave as we want them to. In a Twitter thread by Pedro Domingos [10], typical representatives of a more radical course such as Gary Marcus and Judea Pearl also come forward. They agree that without modeling causality (Causal Inference), there will be no really intelligent system or AI Alignment.

In general, this movement can be accused of criticizing existing approaches but not having any executable systems to show for themselves. Nevertheless, Causal Inference has been a hyped topic for a while now and you should at least be aware of this critical position.

THE PECULIARITIES OF ML SYSTEMS

Machine Learning Advanced Developments

Learn more

ChatGPT, or why 2023 is a special year for AI and AI Alignment.

Regardless of whether someone welcomes current developments in AI or is more fearful or dismissive of them, one thing seems certain: 2023 will be a special year in the history of AI. For the first time, an AI-based system, ChatGPT [11], managed to create a veritable boom of enthusiasm among a broad mass of the population. ChatGPT is a kind of chatbot that you can converse about any topic with, and not just in English. There are further articles for a general introduction to ChatGPT.

ChatGPT is simply the most prominent example of a variety of systems already in use in many places. They all share the same challenge: how do we ensure that the system does not issue inappropriate responses? One obvious approach is to check each response from the system for appropriateness. To do this, we can train a system using sample data. This data consists of pairs of texts and a categorization of whether they match our alignment or not. Operating this kind of system is shown in Figure 4. OpenAI, the producer of ChatGPT, offers this functionality already trained and directly usable as an API [12].

This approach can be applied to any AI setting. The system’s output is not directly returned, but first checked for your desired alignment. When in doubt, a new output can be generated by the same system, another system can be consulted, or the output can be denied completely. ChatGPT is a system that works with probabilities and is able to give any number of different answers to the same input. Most AI systems cannot do this and must choose one of the other options.

As mentioned at the beginning, we as a society still need to clarify which systems we consider risky. Where do we want to demand transparency or even regulation? Technically, this is already possible for a system like ChatGPT by inserting a kind of watermark [13] into generated text. This works by selecting words from a restricted list and assuming that a human making this specific combination has an extremely low probability. This can be used to establish the machine as the author. Additionally, the risk of plagiarism is greatly reduced because the machine – imperceivable to us – does not write exactly like a human. In fact, OpenAI is considering using these watermarks in ChatGPT [14]. There are other methods that work without watermarks to find out whether a text comes from a particular language model [15]. This only requires access to the model under suspicion. The obvious weakness is knowing or guessing the model under suspicion.

Fig. 4: A moderation system filters out undesirable categories

Conclusion

As AI systems become more intelligent, the areas where they can be used become more important and therefore, riskier. On the one hand, this is an issue that affects us directly today. On the other hand, an AI that wipes out humanity is just material for a science fiction movie.

However, targeting these systems for specific goals can only be achieved indirectly. This is done by selecting sample data and Priors that are introduced into these systems. Therefore, it may also be useful to subject the system’s results to further scrutiny. These are issues that are already being discussed at both the policy and technical levels. Neither group, those who see AI as a huge problem, and those who think no one cares, are correct.

Links & References

[1] https://en.wikipedia.org/wiki/AI_alignment 

[2] https://www.bbc.com/news/technology-45809919

[3] https://www.reuters.com/article/us-amazon-com-jobs-automation-insight-idUSKCN1MK08G

[4] https://www.derstandard.de/story/2000142763807/chatgpt-so-koennte-kuenstliche-intelligenz-die-menschheit-ausloeschen

[5] https://www.ceps.eu/wp-content/uploads/2021/04/AI-Presentation-CEPS-Webinar-L.-Sioli-23.4.21.pdf

[6] https://en.wikipedia.org/wiki/Social_Credit_System

[7] https://twitter.com/ylecun/status/1591463668612730880?t=eyUG-2osacHHE3fDMDgO3g

[8] https://en.wikipedia.org/wiki/Linear_regression

[9] https://www.tensorflow.org/lattice/overview

[10] https://twitter.com/pmddomingos/status/1576665689326116864

[11] https://openai.com/blog/chatgpt/

[12] https://openai.com/blog/new-and-improved-content-moderation-tooling/

[13] https://arxiv.org/abs/2301.10226 and https://twitter.com/tomgoldsteincs/status/1618287665006403585

[14] https://www.businessinsider.com/openai-chatgpt-ceo-sam-altman-responds-school-plagiarism-concerns-bans-2023-1

[15] https://arxiv.org/abs/2301.11305

The post AI Alignment appeared first on ML Conference.