Talent Spotlight: Dr. Anton Kolonin, AI and Blockchain Architect
In this third part of our talent spotlight series, we speak with Dr. Anton Kolonin, an AI and Blockchain architect, one of SingularityNET뭩 senior AI researchers and the founder of Aigents, a social computing platform. Anton is playing a vital role in SingularityNET뭩 work on Unsupervised Language Learning and in the development of SingularityNET뭩 decentralized Reputation Management System. Anton regularly contributes to our blog and has represented SingularityNET in various events.
Hi Anton! Thank you for setting aside some time to share your thoughts with us. Let뭩 start with something not so technical. What would you say is your favorite movie of all time?
There is an excellent Hollywood movie that came out recently called Arrival. Though it is not actually about AI itself, it뭩 more about a specific aspect of AI that I work on ?#8202;the nature of natural language learning. In my opinion, Arrival represents a rare case of Hollywood capturing the right aspect of an AI problem. Since the movie perfectly captures the idea of natural language learning, I recommend people I work with to watch it.
I뭭e seen Arrival and I agree, it뭩 an amazing movie.
Let뭩 move on to the most frequently asked question when it comes to AI: when do you think we will create a human-level AI?
To be honest, I don뭪 know how close we are to creating an AI that has human-level intelligence. Currently, I think we are somewhere in the middle of a long path. There has been a lot of development, and we already have products making practical use and creating business value using elements of AI.
However, I think we should avoid using the term AI for some systems and perhaps call them ?EM class="markup--em markup--p-em" style="font-feature-settings: " liga: ; 1: ; salt: ">advanced math and statistics applications.?For instance, different sort of graph theory applications of neural networks such as car navigation, face recognition, optical character recognition, etc. use particular technologies associated with AI. Such systems can not learn new skills in an unsupervised way or even learn from teachers ?#8202;these systems are manually engineered to learn to operate in a given narrow domain, which is not true for human-level intelligence.
When viewed from such a perspective, I would say we are still far from AI being a complete reality.
Even today, we do not have the complete idea of how human intelligence works. So although the development of AI is more than half a century old, there is still no solid design of how AI is supposed to work.
What has changed over the past decades is the performance of our tools ?#8202;things like hardware capacity, availability of high-performance computing, cheap memory, easy availability of data and huge memory volumes. The availability of the hardware and the volumes of tagged data necessary to do machine learning have made the recent advances possible, but the ideas that are behind these advances are often 50 years or so old.
So at the moment, there is no agreement in the community of AI researches on what is to be done next to achieve human-level AI. At one end there are people saying that we have all the algorithms in place, we just to need to put them together in the right way and provide them with more memory and computing speed. At the other end, there are people saying that we still do not entirely understand the nature of intelligence and consciousness, and so there is a long way ahead.
How would you say the attitudes in the scientific community have changed towards AI over the past decades?
From the perspective of the scientific community, there remains a stable distribution of optimism and pessimism. There have always been optimists saying we need a few more years to fully develop true AI while the pessimists have also always been there, saying we will never get there. So though there is a relatively stable distribution, currently I뭗 say the distribution is slightly in favor of the optimists.
Those who are pessimistic about AI do serve a valuable role, but we often regret not doing something in hindsight. Do you look back and think that you could have done some things differently?
Fortunately, in my career whenever I decided to do something, it seems to have been the right decision ten years later. So I have very few regrets because, thankfully, I was right on most occasions. However, there were some moments where I wished I had more time, money or resources to work on certain projects.
It뭩 great that most of your decisions have worked out and I hope that luck continues to favor you. Can you share with us some of the projects that you are currently working on?
These days I am mostly working on three projects, and I would say they are complex enough to keep me busy.
The first project is called Aigents. It뭩 a different and unique project that aims to create AI agents which can help people find information on the internet. So rather than having a person manually searching for information on the internet, the aim is to have their AI agent recommend interesting websites and information to them based on their preferences or needs. Such an AI agent would also be able to monitor social networks so that it can detect changes and events in the social environment of a user ?#8202;providing them with insights on how to improve their social performance. Such insights would enable users to identify opinion leaders, track the sentiment of friends and better understand their audience.
My second project involves building a decentralized Reputation System for SingularityNET, such that it can be used in other networks too. For this project, I am currently developing a proof of concept implementation to combine data about financial transactions as well as explicit feedback made by users in forms of 뱇ikes,?votes,?or 뱑atings.?The data will be collected from blockchain and social media platforms like Steemit to see if reputation distribution from the community can reveal insights like discovering fair high-performance contributors and top influencers and distinguishing them from bots and spammers.
Finally, my last project involves working on Unsupervised Language Learning for SingularityNET and OpenCog. In that project, to put it simply, I am working with other researchers to try to create algorithms that can solve the problem that is depicted in the film Arrival. Basically, we want to find ways to make sense of languages that have not been studied by computational linguists ?#8202;or, in fact, by any linguist at all ?#8202;so currently such languages cannot be processed by computer systems. Potentially, this work on Unsupervised Language Learning would enable us to deal with alien languages as well. Through this work, for example, we hope to figure out something like circles of this type represent nouns, while circles of another type represent verbs. Once we unravel these findings, we can start figuring out how to communicate.
All of those projects are incredibly ambitious, no wonder that they are keeping you busy. You mentioned earlier that you regret not having enough money and resources to work on certain projects. The big tech corporations have seemingly infinite resources. I뭢 sure an experienced AI researcher like you has received hundreds of lucrative job offers. So what drew you towards SingularityNET?
It뭩 not like I have not worked for the big corporations, I have tried that lifestyle and work environment. I뭗 say working for those companies were the worst days of my life. I felt like I was being suffocated because there were so many meetings, so many phone calls, and so many emails ?#8202;with little actual progress on anything. Frankly, that was not fun at all. It뭩 not about being adventurous, I just personally enjoy working in a more dynamic and creative environment. So I would say working in a distributed environment on a project like SingularityNET was far more attractive to me than being part of a typical multinational corporation.
Another motivation for me was that I had started a project in 2014, called Aigents. The mission of the project was to create personal AI agents which could interact on behalf of their human owners and even become a community living on a network. Since this was an idea that was complementary to one of SingularityNET뭩 goals of creating a distributed network of inter-communicating AI, Ben and Cassio wanted me and my project to be part of SingularityNET. Also, they were interested in my work on reputation systems and unsupervised language learning, which I had been working on in collaboration with Ben earlier, almost 20 years ago.
It뭩 interesting that you had collaborated and worked with some of the team members in the past. After you joined, how was working with friends on a challenging project?
I뭗 say I know Ben, Cassio, and Matt very well ?#8202;having worked with them two decades ago. I am well aware of how they behave under challenging circumstances. I know how to communicate with them across many situations, and so because of that, they will always have my trust.
Actually, when you have a bunch of researchers experimenting around with AI, it can be a lot of fun. For example, last year for one of the consulting projects at Novamente ?#8202;Cassio, myself and a few other guys used machine learning to make predictions about the NHL Stanley Cup.
That does sound fun, did it work?
Let뭩 just say it was fairly accurate.
I guess trying to predict the future is one way to use AI. One thing about the future that we don뭪 even need to predict is that the use of AI will continue to increase and that it will have the greatest impact on today뭩 youth. What advice would you give to the youth who are reading this interview?
As with every new technology, there are a lot of opportunities. However, it also brings challenges. Hopefully, today뭩 youth can mitigate three major challenges from AI. The first challenge is the possibility of AI being used as a weapon. The second challenge is the risk of AI fueling a digital divide in the society. Lastly, there is the danger of AI making humans utterly dependent on technology. I think these risks should be mitigated sooner than later.
I advice the youth to be much more socially responsible so that they can strive to prevent the militarization of AI, stop the emergence of a new form of slavery of the 밾ave-nots?due to the digital divide that will further amplify the inequalities in our societies and finally to learn the necessary skills that empower them to do things themselves and not rely entirely on AI.
You are definitely in a position right now where you can shape the evolution of our society. What sort of future do you hope to create with SingularityNET?
The ideal future I would imagine is that the SingularityNET platform would become the equivalent of 밐TTP?protocol for AI. So when anyone wants to create AI, the first thing they think about is SingularityNET. This, to me, would be the most impressive goal for us to achieve.
That is an ambitious goal, and it makes me wonder why didn뭪 someone launched a project like SingularityNET earlier?
In terms of building a community like SingularityNET, it is a 밼irst of its kind?attempt. Regarding the technological and scientific aspects of the project, I think a similar attempt was made much earlier, around 20 years ago, by Ben and Cassio. Back then, communication was difficult. We only had email. There was no chat or tools like Skype and Slack. Ease of communication mattered a lot because we mostly worked in a distributed environment across different time zones. Despite the difficulties, we were still able to work on pretty much the same AI concepts or ideas that we are now advancing at SingularityNET. In essence, currently, it is still the same kind of organization except we have smarter ways of communication.
Anton, I know I can keep asking you questions for the next several hours, but I feel I have taken too much of your time. So one last question. If we were out having this chat at a pizza place, what pizza would you order?
To tell you the truth, I prefer Moussaka. I like Greek food in general. Actually, Moussaka is to Greece what pizza is to Italy. However, if I have to pick one type of pizza, it would be pizza with pineapples.
I think it is safe to say that you just broke the hearts of our Italian friends around the world. Now you have to create those AI Agents for them so that they can forgive you. Jokes aside, Anton it was great to talk to you. Thank you for your time.
We started the talent spotlight series to showcase career paths that can lead to a meaningful and rewarding career in AI. If this episode intrigued your interest, we recommend that you read the first episode of this series in which we interviewed Dr. Deborah Duong or read about our community manager Tim Richmond, who shared with us his journey in the second part of this series. If you would like to work at SingularityNET, please find our currently available jobs here.
In the future, we will interview more members of the SingularityNET team so that we can further highlight meaningful and rewarding careers in AI. If you have any further questions or would like to participate in the discussion about this interview, please visit our Community Forum.
For any additional information, please refer to our roadmaps and subscribe to our newsletter to stay informed about all of our developments