I met Marvin Minsky in a conference in Toronto in 2013. After his keynote, I wanted to ask him something I had recently thought about. I approached him, introduced myself, chit-chat a bit about Finland with him and his wife, who had some friends from Finland. Then I said something like this:
“As we humans are exposed all the time to huge amount of data through our senses — from visual to hearing and from pain in our back to our capacity to imagine, see dreams etc. — does he know anyone studying computers that are purposefully provided all nonsense data and noise, for instance from the Internet, to process it relatively randomly and to learn from it, like we humans do?”.
Marvin Minsky smiled friendly and replied: “what a silly idea” and turned to his wife next to us and continued “Did you hear that, this guy would like to build an irrational machine”.
Why am I thinking about this today? Let me give you a context. Firstly, to be honest, my personal relationship with the phenomenon of Artificial Intelligence (AI) is complicated. On some level, I follow the general discussion and also read research about it, but very selectively, as some proper AI researchers could say. I somehow know how large language models (e.g. ChatGPT) and machine learning work. I have some kind of idea how machine learning is used in search engines, social media platforms, online marketing and in mass surveillance. When I talk about these topics with real AI researchers, they often say that I, like many, think that AI is capable of much more than what the situation is today. They, however, say that it is just a matter of time when this will be real.
I personally like the geeky definition that “AI is a cool thing that computers can’t do yet”. The real challenge is to make sure that we know what “things” are “cool” and what “things” are “not cool” at all (e.g. mass surveillance is not cool). So, there is a good reason to follow AI research. Naturally, I am particularly interested in the role of AI in human learning and development, in human creativity, in art and design. I am not interested in AI per se, but I am interested in to study what it means for humanity at large and, therefore, for the entire planet and the universe.
More context. Lately, I have been thinking about my father’s childhood. Why? I guess it’s because he is sick, and I have been reading a book by Antti Järvi about the part of Finland where his and my family got to flee from when the Soviet Union attacked Finland in 1939. With Antti we are grandchildren of war refugees. Recently, I have also met with some artist friends and listened to their thoughts about world politics, Ukraine. Gaza, the United States, China, India . . . . Then I have also studied the free energy principle, something my doctoral researcher Avner Peled, has talked to me about. Furthermore, I recently went to see two theater plays that made an impact on me: Europeana written by Patrik Ouředník, directed by Minna Leino and Everstiina written by Rosa Liksom, directed by Susanna Airaksinen and performed by Heidi Herala. Both plays are about the history and the future of humanity, especially in the little corner of the globe where I spend most of my time. Finally, for some months, I have been listen to and studying contemporary classical music and been introduced to, and somehow been excited about, the music of John Zorn. A lot of irrational behavior, nonsense and noise.
If I’ll think of any rational reason for all this, there are at least two. Firstly, I am soon starting a new job. Secondly, in a year or so, I have started a new daily practice in my life: meditation. Believe or not, all this makes sense in my mind.
To be better prepared for my new job, I am aiming to have new experiences, to study new things. This helps me to get new ideas. With meditation practice, I can see that the ability to spend 5, 15, 20 or even 45 minutes every day without thinking anything clears my mind. The new data, all the noise, finds its place in my artificial model of the reality where I am operating. This makes me better to predict how the world works and how to exist in it: to alleviate suffering of all the living creatures, starting from myself. To handle all the noise, I need a peace of mind.
The complexity of humans — being at the same time irrational and rational and everything between — is fascinating. The most interesting thing about humans are those things that look at the first glance to be irrational. I love many highly rational people (I know several), but the character of a person comes from their level of irrationality. All people, however, do and say silly things. We are all irrational in different ways in a different situation.
What does all this have to do with Artificial Intelligence? Let’s get back to 2013 in Toronto. I argue that what I described then is very close to what state-of-the-art AI research is doing today. The people working with generative artificial intelligence and talking about singularity are precisely doing this: filling machines with datasets from the Internet and using machine learning to generate something new and sometimes surprising out of them. Is this cool? I am not convinced.
Related to my earlier story about the context, one day, we may also have AI with personal history and presence in a particular time and space, just like you and me. From the human learning and development point of view, I think this kind of AI would be much more interesting than the current forms of AI. I would like to collaborate with an AI that has a character, a life story and life situation, and therefore have a different level of rationality and to irrationality in various situations, similar way as we humans do. This kind of AI could be our peer from whom we could learn from and who could learn from us: be a creative and critical friend — a creative peer — helping us to develop as humans. I am lucky to have a colleagues, Jeongki Lim to work with the topic. Still, I can’t help to think what would Marvin Minsky say?