Igor Kuvychko, PhD is a Principal Data Scientist at INFICON, where he builds ML and optimization systems for semiconductor manufacturing. He holds a PhD in Chemistry from Colorado State University, with over 2,000 academic citations. He was born in the Soviet Far East, grew up in Russia, and came to the US at 21.
I encountered him when he posted a piece, “Gregory Bateson & AI” on LinkedIn, connecting Bateson’s ideas about information, mind, and feedback to contemporary AI.
So as you may or may not know, but I start all these conversations with the same question, which I borrowed from a friend of mine. She helps people tell their story. And I really haven’t found a better way to get into a conversation than this question, but it’s really big.
So I over explain it the way that I’m doing now. So before I ask it, I want you to know that you’re in total control and you can answer or not answer any way that you want to. And the question is, where do you come from?
And again, you are in absolute control.
Oh, this is such a fun question. Well, I’m gonna start from the very beginning. So I paint the picture, paint the story.
So I was born and grew up in Vladivostok, in the Soviet Union. So Vladivostok is a coastal city, very close to North Korea and Japan. So to picture it on a map.
And it’s very much on the periphery of the country, so pretty far away from the center. So that’s where I grew up. And I was a very curious kid.
So curiosity was one of the big themes. And when I was pretty little, I was into many things, including mathematics. And I got really quite into it.
And then I just by happenstance, I switched to chemistry. And I was probably around 10 years old or something like that. And it was summer and I was bored.
And I found that my older sister’s chemistry textbook, school textbook. And back then in Soviet Union, we had four years of high school, mid school and high school chemistry. So pretty decent problem.
And I started reading it and it just clicked with me. It was a fun subject because it was abstract and it was applied at the same time. And I was bored.
So I read the textbook and I solved all the homework problems. And I just had a ton of fun. And I kept doing it.
And then eventually went through all four years of high school chemistry before high school chemistry started. And the school teachers noticed me and then they gave me a little bit of extra attention and they connected me with a chemistry professor at a local university. And she took me under her wing and she gave me free tutoring, amazingly enough.
So every week I would go to this university, and I would hang out with her and other professors in the lab. And she would tutor me on organic chemistry and it was just a blast. So that is one of the themes.
But of course, a lot was going on societally, politically at the same time. So when I was 11, Soviet Union collapsed. And so it was a massive change.
There was something that was supposed to be enduring and permanent just ceased to exist. So from many perspectives, that was a very turbulent time. And I think that was one of the early injections of just turbulence and change.
So that was one of the themes. And then going further, when I was 15, my mom passed away from cancer. And that was, of course, a very one of those forming events in my life.
And that pretty very early made me very aware of finitude of life. And it made me ask some existential questions and questions of meaning, specifically, where does meaning come from? So that’s another thing.
And fast forward, I was, yeah, Soviet Union collapsed. But one of the lessons was the world ends, and then the sun rises, and life goes on. And we adapt.
We are very adaptable creatures. So there was still state funded programs. And back then in Russia, there were these competitions, high school competitions in science.
So high school kids were competing. And I was, I went into this program and got a chance to travel on basically on a government dime across the country, compete with other kids. Eventually, it led me to a very good university program in Moscow.
So when I was 17, I left Vladivostok, and it’s an eight hour plane flight to Moscow. So it’s pretty far away, and got my undergrad in chemistry. Then I realized pretty early on that I wanted to leave the country.
I had a pretty clear sense that the future of Russia was not the future I wanted to be a part of. So I went to the US and got my PhD in chemistry. And at that point, I was, so I was living in Colorado.
And my life was just, it was just a straight shot leading me to a career as a chemistry professor. Just everything was making sense, I was going to be a chemistry professor, I was going to work in academia. But the longer I stayed in academia, the less excited I was feeling about it.
So at some point, I realized that this is not a path I wanted to take. So I started taking some, I took some business classes. And then I made the decision to say goodbye to chemistry.
And that was another big change, because a lot of my identity was tied with being a great chem. But I walked away from that, joined Intel, and started working in the semiconductor factory at Fab. So making chip.
And then I had a number of roles at Intel. They did for almost a year. But I was moving closer and closer to software.
And since I was a kid, I loved applied mathematics. So I was doing more and more at the intersection of applied mathematics and software. And then, three years ago, at this point, I made the decision to switch to software full time.
And I joined Inficon, where I’m currently at. And so currently, I’m a principal data scientist at Inficon. And yeah, I just kept going in the direction of, it’s not really specifically about applied math or software.
It is about solving difficult problems at intersections of the main. So, and that’s something that I’m very interested in, especially when the problem is practical and messy. And there is some interesting math and science, that’s always fun.
But also the problem that I’m dealing with, there is a human element, there is the politics of change, and how people interact with your systems. So, and I feel that this is where Bateson comes right in.
Yeah, hold on, I want to stop there. And because we’re going to get into the Bateson thing, but I still want to spend a little time in your past, when you were a kid, before you discovered, I mean, you told the story very well, of course. But do you have a recollection of what you wanted to be when you grew up?
You have a vision of young Igor, what you wanted to be when you grew up? It seems like you found chemistry as a passion. And so maybe that was the answer. But I’m just curious, what did you hope to be when you were a child?
I had a very clear vision. So, since I was very little, I wanted to be a scientist.
I wanted to be a scientist.
Yes. And what did that look like?
It is interesting, because when I try to think about what was the vision, there was a word scientist. But what did it mean for me, for little me? And I don’t have a clear answer.
But it was more with this sense of curiosity and adventure. And I would say playfulness, too, that you’re just wrestling with these difficult problems. And you’re solving puzzles.
And it’s very exciting. So I think at that stage, that was really the vision. And I think it worked out quite well. Because although my job is, I’m not like an official scientist, but I feel that this is a lot of what I do.
Yeah, that’s wonderful. And I was also curious, too, you told the story of, you were 11, when the Soviet Union fell. And I was just curious, how did you notice it?
What, can you tell a story of what did you know? How did you know that it had gone away? You talked about it, this enduring thing is not there anymore. How did you know?
So of course, I knew that something big was happening, because my parents were freaked out about it. And that was just a topic of the conversations, right? And everybody was anxious and noticeably scared, right?
And kids are very attuned to the state of their, to the mental state of their parents, right? So we know when something isn’t right, right? So that was certainly one aspect, but there was a lot of just purely practical aspects, I remember when I was little, and back then, central heating is typical, was typical in Soviet Union, probably in Russia now.
But heating was never a problem. And then after Soviet Union collapsed, it was a problem. And I remember, even after I moved to the US, during winter time, I would make a mental note that I’m warm, because I remember when I was a kid, it’s just, whatever you are, at home, at school, at university, outside, it’s always cold during winter, just this always background of cold.
And it’s slowly grinds on you, like, that’s one of the things. Another thing was food insecurity. So, I remember the long queues, and the long queues waiting for US humanitarian aid.
And have to, because the aid was given per headcount. So, I had to wait in these queues with my grandpa, and we would get the box. And I still remember the canned mandarins, they were amazing.
Yeah, I remember, still, there was this one, like, my grandpa grabbed me, he’s like, okay, there is a humanitarian aid is being given, we need to wait. And we waited for three hours in this queue. And they ran out of canned mandarins.
I still remember that. So, it was interesting. It was definitely, there was a lot of crime, and a lot of organized crime going on.
There was, it was a very chaotic time, we had to, I’m still an expert in subsistence farming. I mean, I still remember how to grow potatoes, because we would grow enough potatoes to last us through the winter. So, there was this aspect that there is just things that were fine, are not fine, and you have to deal with them.
But despite all of this, of course, I was very young, but despite all of this, I remember this, there was this atmosphere of change and possibilities. That despite the fact that things were pretty rough, there was this sense that now it’s possible to do stuff that was not possible before. So, I remember, actually, a lot of people say that 90s was a horrible time in Russia, and it’s true, but I remember it as a very hopeful time, too.
Yeah, and I know that you mentioned this already in your story, so to catch us up, where are you living now? What’s the work you’re doing?
So, right now, I live close to Seattle. So, I live in Everett, so Washington State in the US, and I’m a principal data scientist. I work on optimization of semiconductor factories, so basically making semiconductor manufacturing facilities run better, whatever better may mean.
And just a disclosure, I’m here as a private citizen, I’m not in my official capacity for sure.
Oh, yes, of course. Yeah, and what do you love about the work? Like, where’s the joy in your work for you?
There are a few aspects of my job that I really enjoy. First is, I have, I’m leading a small team, so I’m working with a few people, and that I find very rewarding. Working with people, guiding them in their career path, and solving problems together, that is great.
But at the same time, I have my own projects. So, I feel that I’m a bit of an unusual case, because I’m a manager, and I’m an individual contributor, rolled in one package. And I’m grateful for this opportunity, because it’s, usually people do one or the other.
But I feel pretty strongly that this is, at least for me, this is a perfect intersection, because I maintain my technical skills actively, so I never become stale. I can see that I can become, I can be very productive if I switch to being 100% manager, but then my technical skills will atrophy, and my usefulness will decline. But also, I think I’m not gonna have as much fun.
So, that’s the bigger picture. And as far as the work itself, the story is really making semiconductor fabs, semiconductor factories run better. And factories, any factory is a very interesting organism, because there is a lot of technical complexity, of course.
There are a lot of constraints. So, there is just, oh, we can make it work so much better, but the reality gets in the way. And also, these are very human enterprises.
So, there is a large number of people of different walks of life that make a factory work. So, I find it a fascinating aspect, because it’s not just a technical problem. It’s a human problem as well, because you, for example, would build some software, but the software interacts with people.
And the better is predicated on the whole system working well.
Yes. And so, first, I noticed you, because you had written this beautiful piece, Gregory Bateson and AI, early in March, right? And it just showed up in my feed.
And so, I’m curious, where did that, what’s the origin story of that piece, both your relationship with Gregory Bateson, but then also, I guess, AI and what your experience of that is, and what inspired you to write a piece about it?
Yeah, I first read Gregory Bateson during my PhD years. So, quite a while back. And, it had nothing to do with my PhD subject matter.
So, it was absolutely orthogonal. And I think I stumbled on the recommendation, basically, the idea of looking at the Bateson, just on social media. I remember there were some accounts that I was following, and they recommended, they cited some Batesonian ideas, and I was like, oh, this is really cool, I need to look into it.
So, I read his book, Steps to an Ecology of Mind. And it’s a brilliant book, and really quite unique. But many of these ideas, they stayed with me.
I think Bateson has the most beautiful definition of information, the differences that make a difference. I think it’s so insightful, right? Because information is not just a bunch of stuff, it’s not like notes somewhere.
It is really once somebody reads the notes, and they compel that individual to do something different. Or a system ingests the information, and there is a change in behavior. And that really struck me, because I think what Bateson does really well, he pulls us up to a higher level of discussion, right?
And the whole Batesonian trick, right, is to look at the whole system. To step back and take a look at the biggest system, and ask these biggest system questions. And I found that to be very insightful, and really beautiful, too.
So, that was my introduction to Gregory Bateson. And then, it was sort of low-level, simmering in my brain, and helping me along, because I think many, many problems become a lot clearer when you use this Batesonian lens to look at them. And I remember at Intel, my colleague, she used to say that my catchphrase is, what is the problem we’re trying to solve?
What is the real problem? Because usually people come to you, and it’s like, well, I want you to do this. And you’re like, well, okay, that’s great, but can you explain a little bit more?
Like, what do you actually want to achieve? Like, I want to achieve that. Okay, this is great, but why do you want to achieve it?
Like, what’s the biggest system where you’re embedded? What is the actual outcome that you desire to achieve, right? And that was a very fruitful line of questioning, because at the end, oftentimes it was like, oh, you know what?
Actually, we need to do something completely different. And with the origin story of the article on Batesonian AI, especially if you are in software, it is impossible to, the state of AI and LLMs. It is, I compare it actually, in terms of just the tectonic change, to how I felt during the collapse of the Soviet Union. You just realize that the, everything changed completely.
And there is also the sense of possibilities, of new possibilities that were not achievable before, but now they’re within reach. Of course, there is a downside to it, too. But me and my team and the company, we have been using AI and code, AI LLM-assisted coding for, very actively, I would say, for a year at this point.
So, we have been fairly early adopters. And it just struck me that the Bateson’s frame of mind and his ideas are so applicable to our age of AI. I mean, I think Bateson ideas are timeless, but they’re just so well suited to this new world that we’re living in.
And that gave rise to the paper. So, because I figured that probably not many people in the AI and software game are at Bateson. So, might as well try to introduce the thinker.
Yeah, yeah, yeah. And what was the response? I mean, you got on my radar through Simon Roberts at Stripe, he posted it and it was making the rounds in this anthropological thread. But what’s the experience been?
Oh, it’s been really interesting and wonderful, quite frankly, because I did not expect the anthropology folks to respond. And that was very heartening to me because anthropology is definitely not my field. I would say that I’m ignorant of the field.
I don’t know the latest debates and what drives the field and the conferences. Like, I’m completely unplugged from this slice of the universe. But I think it was Dawn and she’s the only anthropologist that I knew before I published the paper and we met at Intel.
And I think she posted the paper and complimented on it. And it was very encouraging to me because of course, when I publish anything, I’m like, well, I don’t know. Is it good enough?
Maybe I have maybe this is silly and but at the same time, I’m like if it’s silly, that’s OK. I can laugh at myself.
It’s beautiful. It’s beautiful. Yeah. Well, what so maybe just what happens when you look at AI through Bateson? What does it do for you?
I think for me, I mean, the power of that, there are several concrete things as a Bateson helps me. First, I think is really coming up with better questions. So I think that’s really important.
And I think oftentimes asking good questions is the crux almost. And I think Bateson for me is very effective at making me think about the bigger system. All right.
The AI is, it can be, you can think about AI just a model and you send something to the model and you receive the answer. But the reality is more complex is that the model is embedded in some decision system and it is helping people make a decision, for example, or it’s making their life easier in some way. And I think the thinking about that in a very explicit way, I think for me, that’s very helpful.
So that’s one aspect. Then I think probably the most famous Batesonian construct is the double bind. I think probably the best known popular culture.
And the double bind, that concept, I think it’s useful in designing AI applications because you want to make sure that, because the model may give you a strange answer and you can blame, you may start to blame the model. But the reason may be because the model is caught in a double bind. It has conflicting instructions, especially conflicting instructions at different levels of meaning.
So, for example, you’re asking the model to do something, design, help me design something, a nuclear bomb. And then there are instructions, they say, don’t help anybody with this request. And so the model may struggle, right?
And because it has conflicting information that is fed to it. So I think that is a very practical design guidelines that a model needs to have an explicit escape route that basically model if the model can detect such a conflict and raise a red flag and say, look, I cannot do it because there are conflicting, I’m encountering conflicting instructions and I don’t know what to do. Because another ingredient in Batesonian double bind is that this escape from conflicting instructions is impossible.
So that’s the classic Batesonian double bind. So I think that that’s amazing, right, because Bateson came up with it in the 60s. And this is just so directly applicable to our AI world.
What more can you say about that first one? It's funny — it's laying right on top of an example I give all the time. Are you familiar with Ursula Le Guin, the writer?
Yes.
Yeah. She has an essay that she wrote, and it was about communication. And she makes an argument that the way that we think about communication is, it seems like it’s the same thing that Bateson is saying, which is that we think about communication as like, I’m a box of information, you’re a box of information, and we transmit bits of information through a tube and we take turns, sender, receiver.
But she says anybody that actually has been in a conversation knows that that’s total, that’s not how it works. A conversation is a totally different, more complicated, fluid intersubjective and reciprocal space. And so she says instead of the metaphor of two boxes communicating information, and in fact, it’s not even about information, it’s about relationship.
And so she describes the better metaphor for a conversation is amoeba sex, the sex that amoebas have. And maybe as a chemist, you know more about this than anybody I’ve ever spoken to, but something happens when amoebas have sex, the boundaries between them go away and they become one thing. And her point being that if you really look at a conversation, listening and telling are the same thing. It’s one act in a way.
And does that get to the complicating that Bateson does and brings to the way that we think about AI, that it’s not just these units bouncing around in very finite ways, but there’s all these layers around it that make it just much more complicated.
Oh, absolutely, absolutely. And I think Bateson has a great phrase that I think fits really well, mind not limited by skin. And I think this is very apt here because the idea is that how we think and what we do is driven to the tools that we have and to our interactions and the relationships.
So, because certain types of thoughts are easier or harder depending on the tools that you have. And there is this, it’s a, you can call it the restatement of the old ideas that we create tools and then the tools that we create start to shape us. Right.
So, but tools is a very mechanistic term, but I think the same applies towards people, right? Because this is a cliched thing that you are the average of five people that you spend the most time with, right? And I think that’s very true.
And I think that’s, this is, I love this analogy of amoebas or, I don’t know, octopuses touching tentacles and it’s hard to, it’s difficult to draw a clear boundary, right? And I think this is more of a, it’s a feature, not a bug is that we have, we’re driven and defined by these complicated relationships. And I think there is, I mean, a lot of it is, I think science has been such a successful enterprise and the trick of science has been, you cut something into chunks and explain each chunk, right?
And then you build a puzzle out of explainable chunks. And this is a very powerful approach, right? But at some point, and I think we have already reached that point where we explain the chunks, but now we’re noticing that when you put a bunch of chunks together, the behavior is a lot more complex than we can infer from the structure of each individual chunk, right?
So we’re getting to the emergence phenomenon, right? When we have a complicated enough system, the behaviors that we’re seeing is really way more complex than we would expect based on constituent chunks. And I think this is immediately applicable to LLMs. These things are getting bigger and bigger and more and more capable. They’re amazingly capable right now. So, yes. Sorry, I’m blabbing on.
No, it’s wonderful. And I’m following you there. And I hadn’t, I appreciate how clear you were about that.
I hadn’t heard it explained that way. And I suppose my next question is really about, I’m curious how you imagine AI, like, because of the way that you think, the way that you’ve been educated and Bateson and chemistry and all this stuff. How do you think about AI in terms of what it is and how we interact with it?
Oh, it’s a difficult question because I think from, there is, there are several, I would say viewpoints that I have. One is more of a mechanistic point, what it is and how it works and the pieces. But then there is, and it’s interesting, of course, and in a way amazing that we can do it at the scale where behavior is interesting.
But there is another lens is, this is something that we grew in a Petri dish. Like we didn’t build it with intent. We build the system and then we trained it on basically all the human knowledge, a large chunk of it.
And then we created this artifact. And now we’re probing this artifact and we’re asking like, hey, can it do this? And it’s like, oh, wow, it can.
And can it do that? Like, wow, it can again. So I think there is, from my perspective, the more exciting piece is this discovery of what we can do now, because there is this mechanistic explanation of what it is.
And for me, this is less interesting. And right now I’m very excited about, oh, we’ll have this thing. And what can this thing help me do?
And already, in terms of doing research, studying something, digging into a new area. I mean, all of us basically gained a super intelligent, tireless assistant. I mean, it is an incredible capability.
I feel that many people haven’t, are not fully appreciating what it is and how powerful it is. But I do think that people, people are catching up for sure.
Yeah. Yeah. What’s your experience been with it? What do you enjoy about it most? How are you using it?
I think the succinct way to put it, for a long time I have a pretty diverse set of interests and there is a theme that unites them. But I felt that pre-AI, most of my ideas, they will blip in and out of existence, but I cannot act on them because in the old world, the overhead was too high, but now I actually can. And I have this wonderful feeling.
It’s what things that I really wanted to do and to learn and to try, now they’re actually achievable. I can do them and I can do them it’s feasible and I’m doing it. And I think that is very exciting to me.
Things that were just impossible before now become possible and I can try my ideas, test my ideas, dig into my ideas, dig into something complicated where I’m not an expert and get a good opinion. So that’s probably the most exciting piece for me is that it strikes me as a massive multiplier, a massive force multiplier. If you have an idea, it can really help you.
Of course, there is a danger there because it can amplify good and bad things. So every tool, every powerful tool can be misused. So there is the flip side too.
Yeah. I remember somebody, I don’t know where I heard it, but somebody said, well, two things are coming to me. The first is just I’ll share that.
Somebody said that the invention of the airplane was also the invention of the plane crash was the way that I had heard it. And then the second thing, I was just curious, when you say that you’ve, I completely identify, love the way that you said it, cause I’ve been experimenting on my own and that thing where these thoughts come where previously they would just come and then I wouldn’t be able to act on them. So they would evaporate away.
Now I have an opportunity to dig in them. There’s the, I had that same experience where the possibility to act on thoughts that previously I wouldn’t have been able to act on or to pursue is much easier now. And it’s quite something.
Are there any thoughts that you were, you’re willing, what’s an example of something that you’ve pursued or has been amplified by AI because AI is available to you?
There is of course the professional aspect, which I can’t get into details, but it is incredible in terms of what we can do, what we can try easily and cheaply. And before it was wouldn’t it be nice, but this is going to take a month and I don’t have the time. So can’t do it.
And now it’s okay, let’s spend half a day or day or whatever. So there is that aspect, but to make it more concrete, recently I published two papers on optimization and mathematics on my LinkedIn. And a lot of the research for the papers was done, I did it with the help of AI, right?
A lot of figures, for example, the figures are fully generated, right? I’m okay, I need this figure and everything quickly, quickly put it together. And this still took me, each paper took me days worth of work.
So it was not the case of make a paper for me. Okay. Post it.
No, this is, I value my voice and I really like the act of writing, right? So this is something very dear to me, but in the previous pre-AI, I would not be able to do it because it would have taken me probably a month or two of dedicated work to write these papers and all the support and information at the level of qualities that I find acceptable for sharing. And now in a matter of several weeks, or a month or whatever, working an hour here, an hour there, because I have a family and I like spending time with them too.
So now all of a sudden I can do it and I can explore these areas of mathematics that I didn’t have time to really dig into before, find interesting connections that I personally find beautiful. Maybe somebody else will, because I think for me also a good indicator, if it’s worth doing for me, I like using LinkedIn as a motivator to really push the idea and polish it and make sure that I give it justice. But I feel that I’m not doing it for LinkedIn or for engagement.
It’s more of an extra motivator. If LinkedIn didn’t exist, I would still do it.
Yeah. I love what you said about you value your voice and you like writing and that’s really important to you. And I’m curious, I know in my own experience, things get, maybe this is the whole question about, we’re talking about the boundaries that you can, where do I end?
Where does AI begin when it comes to the work and the creative stuff? How do you protect or what’s your workflow to protect your voice and to protect the writing and to use, how do you keep AI at a distance and away from that? Or is that a problem or a struggle at all that you need to do?
That is a really interesting question and I feel that it’s a moving target. At first I was, I would say very cautious in my use because I felt that I didn’t want to outsource too much of me and too much of my thinking. And as I use these tools more frequently and the tools are getting better, I feel that there is this it’s an interesting, I would say, blend almost.
And it is sometimes a little bit unsettling because it’s oh, is this too much or is it too little? And I feel that this is, again, always a moving target. But it’s a difficult question that I think all of us have to struggle with a little bit and find a good answer for ourselves.
And it’s also a moving target because models are constantly changing and they’re getting better. And I noticed something that half a year ago, I used to be explicit in my prompts and say don’t be sycophantic, tell it to me straight, all that stuff. And then I realized quite a while ago that I stopped doing it because the model got to a point where they push back, I would say, fairly effectively.
So, yeah, I don’t have an answer. I don’t have an answer because I think it is more of a process. But I do, and I think this is, for me, this is the definition that it is an active process and I need to maintain my judgment.
But also, I don’t want to be blind to the input because I do have these things that have every single book ever. And it would be, it probably wouldn’t be very smart to dismiss the feedback.
Yeah, yeah. I really appreciate the answer. And I know that in my own experience, I have the same challenges.
And I guess, is this what Bateson is talking about in a way that there’s, that things get blurry and complicated in ways that we don’t understand? Or am I misreading or over-reading Bateson into this part of the conversation?
Oh, I think that, for me, that’s very similar to how I think about it. I think things are, I think actual things, they’re not isolated. It is an ecology of things that interact.
And so trying to draw sharp boundaries, it is necessary and it is helpful because then, if you just well, everything is one, that’s not very useful or practical. You can’t really do anything with it. It’s okay, everything is one and then what?
Then it’s too complicated and you can’t really deal with the complexity. So, you have to cut and partition and think about things in isolation. But it has to be counteracted by this awareness that there is the ecology of things that interact.
And that’s very much how I think about my interactions with models because they’re definitely changing. I mean, they have changed the way I work. They have changed the way how my team works.
They have changed what is possible, what is worth doing versus what is not worth doing. I think that software is at the edge, at the leading edge of this change because the models are just so good in writing software right now. And of course, that creates, I see that many people have difficulty adjusting.
And they have a lot of empathy because there is the question of meaning, we do things, there is the practical aspect, we need a job to make a living and have shelter and food and all that stuff. But we also create meaning by what we do. So if your meaning was, well, I spent 20 years learning how to write awesome software and now this thing came along that can do it, frankly, better than I can and much faster and much cheaper.
What is my contribution to the universe? What is my meaning? What’s going on?
So, I have a lot of empathy because I changed careers, I said goodbye to chemistry and it was difficult. But sometimes, just this is the reality of it.
Yeah. I have two questions coming in at once, and maybe it’s really one. How are you feeling about all these changes?
I’ll give you the context. A guy invited me to write for his blog and asked, what would you say to a CMO? I do face-to-face qualitative research, ethnography — I’m a human being who goes and talks to other human beings on behalf of a company. And he said, why should a CMO pay for you when there’s synthetic data and AI all over the place?
It was an existential question for me. Because honestly, if I were a CMO, I don’t know that I would pay for a human to go out and talk to another human when AI can do so much of what it can do. It was a serious challenge to my sense of professional value.
So I’m wondering — did you have a moment like that? Where you asked yourself what AI does that you do? And more broadly, how are you feeling about all of it?
Yes. And so, full disclosure, right? I was very uneasy about these changes because I spent, I didn’t spend 20 years, but I spent years learning the craft of programming.
It was at least 10 years of dedicated study, and I enjoy it. And I pride myself at being good at it, right? And so, initially, I remember initially, I guess it was autumn, fall of 2025, the models were getting better.
And I was starting to experiment with those things like Copilot, and they were useful, but there was still these things that when I would try to push them even a little bit, hey, create this plot, Python, Matplotlib, whatever for me, give very detailed instructions, and it would fail. And basically, I was like, it’s faster for me to do it myself. So, I remember, I gave this talk, and it was October of 2025.
And I was pretty, sorry, October of 2024. So, October of 2024. Well, I was fairly skeptical because I’m like, yes, this is useful, but it’s very, I can see the limitations, and these are the limitations that I’m seeing.
And I didn’t believe that, I could see the limits. And I didn’t expect the models to just blow through these limits so quickly.
And so, when I definitely had that sensation of, oh, these things are getting better and better and better, but I think there is this mental protection that there is an uncomfortable reality, and you are explaining it away. And the models for instance are good also. So, it was easy to explain it away, to say, okay, yeah, this case is okay, but this case is not great.
And then, last year, I had this experience. It was in May of 2025. So, almost exactly a year ago.
And I had a very urgent project. And just ChatGPT basically sped me up by about a factor of two. And I was like, wow, this is legitimately amazing.
And then, a friend introduced me to QuoteCode and agentic coding. And I started on my personal time, I tried it. And I was like, wow, this is amazing.
And I remember, I’m usually not a command and control person in my management style. But I remember on Monday, I told my team, look, if your house is on fire, take out the fire, and then install it. And you can spend a week, just push your project, spend a week familiarizing yourself with the technology and start using it.
Because this is the future. I hate the phrase game changer, but I felt that not pushing my team on that would be a dereliction of duty. So, I guess my extended answer is that there was definitely a mental, some mental barrier to overcome.
Because it is, suddenly, a big part of your identity and your skill set became, the value that’s probably not the right word, but it’s not really where your value to the enterprise, to the economic system, I guess, this is not where your value is coming from, right? You need to figure something out. And I think this is a big, I see still, right now, and the models are staggeringly good now.
I still see a lot of discussions of people saying, well, but when I asked ChatGPT to count R and strawberries, it doesn’t do a good job. And I’m like, this is, to me, is more indicative of, this is more the window into psychology. Because, yes, it’s not perfect.
It’s not an oracle. But it’s an extremely powerful capability, right? And it’s, yes, it can’t count R and strawberries.
If you ask it to use Python to count R, it’s going to do a perfect job, by the way. But you need to, yes, there is this massive change, and it’s totally unexpected, and it’s very threatening. But what can we do?
We have to adapt. And we don’t choose the moment. We were born in this universe at this moment.
And this is a huge disruption, right? And we have to navigate it. And I think my experience is that it’s better to accept it and start figuring out a solution.
This is like dentistry. A toothache is not going to go away on its own. It’s only going to get worse.
I didn’t see a dentist analogy coming, but that’s perfect. The wonderful analogy. We’ve just got a few minutes left.
And I wanted to, because you’ve walked us right up to it, I wanted to bring it back to your piece about Bateson, in which you said you wanted to introduce Bateson to people who may not know Bateson. But also you wanted to introduce it into the conversation about AI because Bateson would, and you said it at the beginning of our conversation, it would lead to better questions. So, I guess I want to give you the chance.
What are the questions? What are the better questions that we should be asking or that you’re thinking about as it relates to AI?
I think it’s, yeah, the better questions to me is really there are some tactical, practical questions about, for example, double binds, schismogenesis. This is more around design and how we actually build these systems. But I think a bigger question is really thinking about the bigger system and the impacts that we want to have.
So, I think that’s a question that I find very compelling. And I think there is, of course, you can ask it on a very large level, right? Okay, this is a new technology, how it’s going to impact our lives.
And the answer is, I don’t know. I’m just a little guy doing some stuff. But there are things that I’m doing for our clients and I want to make sure that we’re successful as a company and that we’re making awesome software.
And so, I think asking these bigger questions, how people are going to interact with the software, how I can make their lives better, how I can make the whole operation more efficient. And, yeah, I think this is really the bigger question. And, of course, for me personally, it’s about having this awareness of, now there is this blend of model intelligence and my own intelligence, and it’s hard to separate, to draw a very sharp line.
And, it’s, I would say the adventure continues, it’s a very exciting time. And I didn’t expect the models to get as good so quickly. So, I don’t know where we’re going to be in a year from now, what capabilities we’re going to have available to us a year from now.
And this is very exciting and a little bit scary.
Yeah, for sure. Igor, thank you so much. I know that my invitation came from a, or you used the word earlier, orthogonal direction with regards to our professions. But I really just appreciate you accepting the invitation and sharing both the piece and just, and showing up and talking about it. So thank you so much.
Absolutely. It’s been a great pleasure. And Peter, thank you so much for reaching out. It’s been so much fun.
