The interviewer asks whether the "information agents" in her account of nature can be physically measured. She gives the example of people who once compared books by counting their words.
Yeah, I think absolutely. We are made of the same stuff as the universe, so just the subset of those atoms or molecules or higher complex structures. And this process of communication with the world or information processing, computation within all those subsystems is specific and local. So there are processes in my atoms which are constituent part of my molecules, which are constituent parts of my cells, which are constituents of tissues, organs, and the whole organism. And nervous system plays a very important role in this whole distributed computational info-computational system. So you see it like networks of networks of computational processes. And they start from bottom up and from top down. There is also top-down layer, which was very much feared in science. You would say Aristotle has the idea that things are searching their goals. They are searching for anthelegia or whatever it was the word. And this idea that there is some goal, there is some meaning in the whole. For example, in evolution, that there is a goal of evolution when we talk about cognition and biology. And this is a new view of evolution, extended evolutionary theory, and the new view of living organisms as agents. I think Levin says agential material, Darwin's agential material. And many people also, Deal Miller and Balushka, Frantysek Balushka, and Reber, and I think it was Sleepcevich and other people talking about how important it is that those agents, those things that are evolving, they do not evolve randomly. They have their own goals. They want to survive and they choose the way they behave. So it changes a lot how we see evolution as well. And we are also.
Can we measure these information agents in any way? Like, do they carry a physical measurement that we can record in some sense? Can we measure how much information an agent contains? You know, like, are we going to hit the limits of our physical instrumentation with studying these informational agents? Do they even carry physical units in the same way that we think of other things carrying physical units and measurements?
There are two answers to your question. First is that, as I told you, there is this informational physics that rephrases the whole of physics, basic physics, in terms of information. So that's one thing. And the other thing is our obsession by measurement is really dominating science today. And I think we must go back to the meaning, not only the quantity, because quantity can be irrelevant, totally irrelevant. I remember at some point people were counting the number of words in books, concordances. They would compare books according to number of words or how many words certain words appear in a book. You can measure whatever about the book. You will never know what this book means actually without focusing on that aspect. So I think when we talk about intelligence, there is really a good reason to think about meaning and to focus on meaning, not counting things, but meaning of the mechanism, the goals, the processes that are going on in nature. We can learn a lot from nature. That we can reconstruct from the history, from evolution, and from all we know about the nature. We can learn a lot how things are developing, how processes are going on, unfolding, and we can reason and try to see where are we going, where do we want to go. There is a lot of value reasoning which is needed to be done, especially today when we are developing this AI systems, which are getting more and more autonomous and both for cars and in war applications and anywhere in decision-making in the society. So, I think I would prefer in my approach, in my research, to focus on what is happening, why is it happening? Do we want to continue that way? Do we want to do something else? And how can we do that if we choose to do something differently? So, we don't know if artificial systems can be autopoietic in the way as living systems are totally self-sufficient in reproducing, in repair, in such things. So, we don't know. People are thinking about it, dreaming about it, but no one knows.
Let's take concretely an example from Michael Levin. You've brought up his work a bunch of times, and he's just incredible. I've spoken to him twice now, and his work is just awesome. If we look at his Xenobots research, people really interpret that experiment in two ways. There's two sort of polar opposite camps of how they interpret that work. When you take these cells and you take them out of their usual environmental and biological constraints, they start doing all this stuff. And the why is where these two camps really divide. So, in one camp, you have the these are cells problem solving, goal modeling, trying to survive, and that is why you get this emergent behavior because these cells are acting like agents, they have goals and they're trying to achieve their goals. But then you kind of have this other camp, which is this is just biochemistry, it's just biochemistry in a different under a different set of constraints, and that fully explains why you get different gene transcription, why you get random emergent properties. It really is random. It's it's it's there's no sense of agency, there's no sense of sentience or consciousness or experience, there's no goal modeling, there's no cognition, it's just biochemistry or just biophysics. How do you, what do you say to those, to the critic that says we can't use these terms, we're just blurring definitions? And I do struggle to come up with a really good answer here myself because I'm in the agents, Michael Levin, camp, but I find it difficult to fully articulate why that can't just be, say, dumb biochemistry and dumb biophysics. Like, why is that not just a biological system acting differently in a different scenario? It's kind of hard for me to really give a good answer for that.
Yeah, I think the idea of the framework I'm developing is to try to say there are so many different levels of abstraction and different levels of description of a system. So, you can describe a human as a bag of chemicals. When I worked in the safety field, the human was modeled as a bag of water, and it was enough for that model, for that approach. So, a bag of water, that's it. You can also be interested in chemistry of the human body, and then you can sort what chemicals we are made of. You can look at atoms, you can see us as at many different levels of organization. But if you are interested in the most fascinating abilities of those xenobodes, it is that they behave in a way you don't expect them. They are really solving problems which are problems of their current life. They are not behaving as in their old context of a lung or wherever they were. So, that shows that those biological cells are really cognitive systems. They try to live their life where they are, adapt to the environment. intelligently solve the problems they they go through the maze they they can even reproduce by collecting other other such cells and and so on so that's a very good uh example why we should if we are interested in in really the the highest uh competencies of of of a system we are studying that then then we cannot uh neglect the the the the fact that they behave they behave they behave in totally uh surprising ways we didn't know that will happen with with the cells outside the body of a human they they they they made the experiments even with the other i don't know if it was
trachea cells right the anthro bots or the