DruckFin

Chemify Transcript: Lee Cronin on Molecular Computation, Assembly Theory, and Genesis Platform

August 2026 In-Depth Interview with Chemify CEO Lee Cronin

Lee Cronin, DARPA, and Chemical Robots

Host: But Lee, it is great to have you back, man. Last time you were here, episode 289, it was like a year and a half ago, you closed like we literally stopped the cameras and then you are like, "Yeah, I was doing some work for DARPA as well." And you just laid that one on us. And I am like, well, that, I mean, you are supposed to say that at the beginning of the podcast, but now you are telling me like, ah, it is not that big a deal. How is DARPA not a big deal? I mean, I am like a big DARPA fan.

Lee Cronin: Yeah, I do not. So, I will, I have had maybe, let me get this right. So, how many grants have I had from DARPA? So, DARPA is a funder, right? They are a grant funder. They give people money. I think the reason they were established, you know why they were established? I think you can check this, their mandate was like to not, when Sputnik went up into space, the US were not expecting that satellite. The Soviets launched a satellite, I believe it was in the 1950s, I think that is right, and basically DARPA was then established to eliminate strategic surprise. So basically they wanted loads of money, fund people to do crazy things, and also to have a mandate to do stuff. So arguably DARPA might have had something to do with say the internet, right, self-driving cars, and a few years ago DARPA started looking at chemistry and AI, like many years ago, way before the hype now. And there was a program they started called Make-It, actually. And the person who started it, you know, they kind of used some terminology from one of my papers and I was kind of a bit grumpy. They did not offer me any funding, right? Because they gave money to MIT, to Stanford Research Institute.

Host: DARPA plagiarized Lee Cronin.

Lee Cronin: You used the P-word, I didn't. They were inspired. Anyway, I remember they got on a call with me and I was like complaining bitterly about that they used this paragraph of what I was trying to do. And they were like, "Dr. Cronin, we are phoning you up because we want to talk to you about that." I was like, "Okay." And basically what it turned out is they wanted me to put in a white paper. And I said, "Okay, I want this, you know, a zillion dollars." And they said, "Well, why not ask for a little bit less and do this thing that fits into our program. You know, it is US taxpayers and you are one of the few groups overseas doing this thing and we think it would be valuable for us."

Host: And what was the thing specifically?

Lee Cronin: So it was part of their Make-It program. Can you make molecules on demand, right in theater, or you know, on the moon, in Antarctica, whatever, right? How do we do chemistry robotically? And then when they ask you about that, and I have been building chemical robots for years.

Host: Hey guys, three quick things. Number one, if you haven't subscribed, please subscribe. It is a huge, huge help. Number two, if you would like to join my Patreon for early uncensored releases of the full episodes, you can join via the link in my description or in the pinned comment below. And number three, if you would like to join my clipping community for a chance to make content from the show and make money, you can join via the Discord link in my description below. You have been building chemical robots for years.

Lee Cronin: Yeah, yeah, yeah, yeah.

Host: That is a dense statement.

Lee Cronin: For since 2012, right? So about 14 years.

Host: How would you define a chemical robot for people out there who are unfamiliar with a specific type?

Lee Cronin: A robot that you give instructions to and it will do chemistry. So yeah, I mean it is out. So I made, I can tell you about it, but let me answer the DARPA question because you asked. So that was the first grant they gave me, Make-It, and they carried on giving me money there. And then they had one on computation using molecules to make molecular computers, and I got money from DARPA there. And that was in a US consortium, so there was a couple other groups in the US, right. And then there was another grant they gave me for using AI for discovery. So they were the three DARPA grants, I think there is three DARPA grants.

Host: What year did they give you the AI grant?

Lee Cronin: Oh, ballpark 2018, 2019, and then it carried on through the pandemic. It was a bit tougher in the pandemic, but DARPA are a brilliant organization, right? They go where the experts are. They do not give grants to people in countries that are probably anti-America, I would say, right? So they didn't fund Bin Laden. And also they want to start new fields, and they want to start fields that are going to obviously be of strategic importance, right? And the UK and the US have done a lot of hookups in different scientific disciplines for a long while. So I thought that was quite good. I made a molecular computer, chemicals to do computation as well, because I wanted—

Host: What does that mean?

Lee Cronin: Yeah, I don't know. Well, I know what it means now. So what does a computer do? We are going everywhere, but anyway, let's start with what is a chemical computer. A chemical computer is a chemistry set where you could put in inputs, the data. You input the data into the chemistry somehow, and the chemistry would then take that data and process it using chemical reactions rather than transistors on silicon. And then you would read it out using some method. And I built a chemical computer using a thing called the Belousov-Zhabotinsky reaction, which is a chemical clock. And all it does goes tick-tock, tick-tock, or red-blue, red-blue, red-blue. Okay, so basically pour a lot of chemicals in a pot. You could just take a beaker, which is like a glass, fill it up halfway, put in some chemicals, and without doing anything, just watching it, it would flash red-blue, red-blue, red-blue whilst you are stirring it. When you stop stirring it, it would basically stop, but then spontaneously flash one color or another, what is called an excitable media. It is a bit like thinking about how neurons flash. So I thought that is a cool idea. Why don't I just basically make a grid of stirrers? So I 3D printed a grid basically about the size of a small book, and in that grid it was a 7 by 7 grid, so 49 little wells. And I put a stirrer bar in each one and put a little motor under it and stirred them and put the chemical reaction in. And all the colors went across the grid, and I used a webcam to basically read out the colors. And so when the stirrers were on, that was a 1. When a stirrer was off, that was a 0. And I basically then printed in 1s and 0s to represent some input and then just read it out and then worked out what it is doing. We kind of just made it up to start with.

Host: Tell us when you are done.

Brain Gel and Whether AI Is Sentient

Lee Cronin: Well, hey, it is a DARPA project. It is supposed to be crazy. No. So what it was supposed to do is, could I somehow use the, because remember the BZ is a clock: tick-tock, tick-tock, tick-tock. If you put a load of clocks together and you allow them to synchronize, as they desynchronize you could process some information. So we actually used it to classify, and we used it to make a primitive neural net. And the idea was a neural net literally used no power and was quite good at error correction and could like refine images. The same way you would use ChatGPT or you would use Nvidia's architecture to do gradient descent. And what is gradient descent? Well, it is basically that it changes the weightings in the neural network to basically maximize its training capability, because obviously you have the weights and the activations, two different things. But basically this was a chemical neural net, and it was the idea, it was the first step. I am now making brain gels, but that is another discussion.

Host: You are making brain gels now?

Lee Cronin: Because we took that and put it in a gel, because it was all liquid. It is sloppy and the memory wasn't that good. But if you put it in a gel, then maybe you could basically use the polymer gel to switch in such a way you could teach the system to learn for a longer period of time.

Host: How does that work scientifically?

Lee Cronin: No idea. That is why I did it.

Host: So you are trying to figure it out.

Lee Cronin: Yeah. I mean, look, we are all over the place, but why not? I was inspired by the episode of Pickle Rick in Rick and Morty. And so I took a gherkin, attached the gherkin to the mains to 240 volts, and then put in—you know my workshop I have at home—and used a bunch of electrodes and tried to program the gherkin to recognize the difference between a circle and a square. But it kept exploding. It kept exploding because it kept heating up. I thought the gherkin could be a good—it seemed like a good idea. It is always a gherkin computer.

Host: Literally you could just be making this up, right?

Lee Cronin: Just anyway, gherkin, heat it up, put it in. It sounds ridiculous. You heat it up by putting it in 240 volts. What would happen? Sodium ions, potassium ions, because it is in salt. So you got this gherkin, you soak it in the salt, so it has this salt in it. It is conducting when you put electricity into it. When you put alternating current at high voltage, it heats up and basically the ions can move around. So it literally makes an excitable media, a bit like the BZ reaction flashing on or off, but here with the gherkin. And then when I put in the little electrodes, I put in a little grid, and I was then trying to basically train the gherkin to tell the difference between a square and a circle because it was made malleable. It is a bit like taking some cement and making it liquid again, and then it will learn something and you will set it. But the thing is, the gherkin kept exploding before I could actually train it.

Host: And this is in your lab at your house?

Lee Cronin: I did it at home, yeah. I don't think the university would be too happy with those exploding gherkins. So now fast-forward to the brain gel. I was like, well look, if I could take a gel and then make the gel conduct, so it conducts electricity a bit like a wire, and then put in electrodes on one face and the other and program one surface using a camera. So I could just take a visual feed and plug it into the gel, and then basically read out as I show the camera to different objects. Can the gel tell the difference between the edges on the objects, and train it in the same way you would use OpenCV or some kind of silicon-based system for edge detection, right? It is a bit like you use a Jetson Nano or something, right? And I was just playing around if I could use material to compute. Turns out you can. But now when you said, "Well, how does that work?" and I laughed, well, it is not because I am making it up, but because the way you program a material is the material responds in time according to its stimuli. So it is very much trial and error. So it is a bit like how the brain has evolved. The brain has evolved over billions of years, and so the brain has a number of different kind of programming periods. It was programmed by evolution. So your brain, let's say, is 3.8 billion years old. So that is time number one. Then you are created by your parents. So over a period of 9 months, your brain is growing in the womb, and so your brain is starting to program. When you are then born, obviously you go through massive developmental surges. So age 21, 22, your brain may be fully almost there. So then your brain is then able to process data in real time. So there is all these different developmental stages that the brain is able to evolve, grow, be programmed, activate, and then continue to work. And we have no idea how brains work. And what are unique to brains? Well, sentience, consciousness, the purpose, a living system that will tell you that they are an individual. You talk to all humans. Most humans you can communicate with will claim they are an individual.

Host: Do you desperately need a nicotine pouch every 15 minutes just to feel normal? So does Joey D. He is a complete addict. Which is why D started using Ultra's new energy pouches. They have been an elite upgrade for this fiend. Ultra pouches are instant energy in one pouch. Ain't that right, D? He has been using them obsessively for about a month now, and they have been giving him a huge, intense energy boost when he needs that extra kick in the morning, during work, or before working out. These are 100% nicotine-free pouches, and they give you energy in a way that is completely different from nicotine, coffee, or other pouches. Ultra Energy Pouches still contain some caffeine to give you that little lift. The real difference comes from their other ingredients, including L-theanine to smooth out the energy and curb the crash, Vitamin B3 to turn carbs, fat, and protein into brain fuel, Vitamins B6 and B12 for balanced mood and recovery. So unlike nicotine that might only help for 10 to 15 minutes, Ultra Pouches deliver instant energy that lasts for 30 to 60 minutes. It is why you see me passing so many of them under the table during podcasts to make sure he stays awake. The best part about them is these are clean, truly good-for-you pouches. Zero nicotine, zero calories, zero sugar, zero sketchy gas station chemicals. Ultra is the ultimate guilt-free pouch delivering instant energy, mental clarity, and boosted mood without nicotine, sugar, or calories. Ultra is trusted by pro athletes, entrepreneurs, and engineers around the world. It is for everybody. And today, our listeners can get the ultimate guilt-free pouch for 15% off using code JULIAN at takeultra.com. That link is in my description below. That is takeultra.com, link in my description below, for 15% off with code JULIAN. After you purchase, they are going to ask you where you heard about them. Please support our show and tell them I sent you. So you think the brain is the thing that is strictly just programming that in, and it might not necessarily mean that that is the case?

Lee Cronin: No, no. I think there is something very interesting. I don't think it stops with the brain. I think biological systems are very complicated, but I think humans are unique in that they will claim to be an individual, right? With words, if we will teach them language. And I think that is something that we don't understand, how chemistry gives rise to consciousness and how chemistry gives rise to the fact that I will claim to be me and I have my own particular personality and do stuff. This is very important in this time because we are viewing that AI are entities that are sentient, and they are indeed, they are not.

Host: Yeah, you have made a lot of arguments on this because you think that, for example, a lot of AI doomers are making leaps that the AI can exist without human beings in the future, and you don't agree that that is the case.

Lee Cronin: Yeah, clearly not. It is fantastical.

Host: Why do you say it is fantastical?

Lee Cronin: Well, we have not seen any evidence. If we think about the origin of AI, right, if you go all the way back, origin of life, the AI is produced by—well, let's say the origin of the term AI was at the Dartmouth meeting, which you can check. Whatever, let's build as many conspiracies as we can go. But I think DARPA has funded some AI work right at the beginning. So at Dartmouth College the term AI was coined. When was that again? In the 1950s, 1956, I don't know, a long time ago, maybe a bit later. And so, if you think about what a computer is, a computer is a deterministically built system. It is switches, right? And the way those switches work is a computer is built out of switches. Those switches have high and low. You put those together to make logic gates. The universal logic gate in a computer is a NAND gate.

Host: A NAND gate.

Lee Cronin: Yeah, NOT-AND, right? You can build everything with NOTs, right? You can build memories, you can build everything. The way you architect those together, you build the silicon infrastructure for everything we have today based on that. There may be some differences that electrical engineers listening to this might be shouting going, "No, that is wrong." There are other things called ASICs which basically might be able to do neuromorphic computing, where basically you can have an analog voltage. So in a transistor, typically you would have a voltage between 0 and 5. 5 is high, 1; 0 is low, 0; and something in between. If it is on the 0 side you call it a low, and if it is on the 5-volt side you call it a 1. But I am not a qualified electrical engineer, so I don't know how the ASICs are built today. I have designed some ASICs for fun.

Host: For fun.

Lee Cronin: Yeah, well, I want to build a brain. So if I want to build a device, a chemical brain, I have to design an analog electrical system for addressing the chemical brain. We will get there later.

Host: Yeah, we will come back to that.

Lee Cronin: So the most important thing with a computer is it is fairly deterministic in some regard, and let's say that it is digital. And so that means I can put the inputs in there and go on. So with AIs today, they are based upon digital computers and they use all this digital infrastructure. We take massive amounts of data and train a model, and the model can do splendid things. But the doomers say that these models could run away. And I agree the models could run away with bad humans running them, right? There is always a human in the loop. The human is turning on the power station. The AIs, although we use terms like recursively self-programming, I mean it is a bit like you and I could build an AI now that we could have a loop in a program: do this thing, and if this thing occurs, then do that thing, and carry on until it either hits the objective or someone turns off the computer. Now that has always been the case, right? Those loops have always existed.

Sam Altman, Anthropic, and OpenAI Compared

Lee Cronin: So the AI doomers are kind of stuck in this fantastical reality whereby they say that AI could overtake the world and cause a nuclear explosion or something else. What I think is more likely is that bad humans using the AI to do bad things will happen, and I do worry about that. So I am not dismissing AI doomers. I am dismissing AI doomers who are saying this thing will happen because magic, on its own.

Host: Yeah, got it.

Lee Cronin: And I think that is actually a major worry, because they are talking about magic and actually there is just some bad actor going, "I am going to use AI to crack into this system and do some damage," or "I am going to use an AI bot system that will basically do some social engineering on social media to cause some kind of effect," right? So I think that I haven't seen any evidence of sentience in any AI, right? The sentience comes from humans exclusively. Any evidence of sentience you get from the internet training on the model. If you say to Claude, "Are you alive?", Claude will probably say, "No, I am not alive, but..." and they will use "I". And this anthropomorphization is just to hook you to be addicted to talking to the entity.

Host: What if it is a technicality though, and it doesn't get the sentience, but because—and I am going to way oversimplify this for a minute—we are able to eventually create technology that isn't quite simply plugged in and unplugged to be able to do that. Because of that, a non-sentient but extremely intelligent artificial intelligence who has no ability to empathize, feel any sort of emotion or level of humanity, took actions that it doesn't understand the grave consequences of. Do they have an argument for something like that to possibly happen?

Lee Cronin: I don't quite understand the statement, but I know what you mean, so I am not going to be too obtuse. But AI, are they intelligent? We have to define intelligence. What really makes me very confused is there are people out there that say in the next years the AI will be more intelligent than any human. And what does that mean? What is intelligence? If you can't measure something, could the AI cause something to happen on its own? No. Could humans not understanding how to set up a system cause something bad to happen? Think about social engineering. Take any system. This is kind of hard because I am not qualified, but let's say we are going to create a type of bylaw in the city that says, for you to get planning permission in the city, every time you build a building you have to have a tree on every corner or something. So suddenly before you know it, there is just random trees everywhere. When humans make decisions and policies, they get propagated, and they get propagated by systems, whether it is in a corporation or in an institution or in a software algorithm, right? So those things get propagated, and there are unforeseen consequences of them. And I totally buy that an AI trying to achieve an objective could do bad things. But those bad things, humans are in the loop all the time. And I guess what I am trying to say is the AI is not bad; the human is bad, or the training data is bad.

Host: Yeah, and if I agreed with that point—and you might be right, honestly I hope you are right and it is in control of humanity, selfishly speaking here—but let's assume though there are some sociopaths who have their hands on the trigger with these things, which I think it is fair to say some of the technocrats we have in the world, not all of them, but some of them may fit that bill. What is to stop people, to put a name on it, like a Sam Altman who jokes about humanity ending, what is to stop him from being like, "Ah, you know, today I want to play World of Warcraft, but with AI, and let it go wild"?

Lee Cronin: Look, I have a lot of sympathy for people who are building companies and building technologies. I am building one myself right now, which we can talk about. I think we have to take a step back and say, how are we building our technologies? How are we regulating those technologies? And how do we then get feedback from the regulation of those technologies to make sure that good things are happening to humanity? Let's take an example where this has already happened: we got the internet, and then we got social media. What was the critical failure in social media that has given us—arguably I don't think anyone would argue that social media is particularly nice right now. What is the failure? I think there is one critical thing that policymakers should have done, which was say to social media that they should be treated like publishers, Section 230 or something like that, I don't know what the US version is. If you are treated like a publisher, anything you put on your platform, you have to be held accountable for. So when you start putting nonsense out there, you have to say, "No, it is clearly a lie, take it down." Basically Facebook and X and Instagram, all these entities, are not accountable for the stuff they put out. Now that for me is a real problem, because that entire decision—although they are a medium, it is a bit like you can't hold the person who built the printing press liable for the bad books printed, I get that. The same way we might argue when you talk about my chemical robots, am I going to be liable for anyone making any bad chemistry on them? But I do think there is something we can learn from social media and the fact we didn't regulate that, we didn't hold them accountable. And so it is very hard to know how it is going to evolve.

However, having said that, if someone invents a social media that is quite good, as in a nice place to be and not a hellscape, if we can work out why humans—you get anger for attention, right? If you create anger, you get the attention the system wants. If there was a social media system whereby—and I am not saying it should be like some Mary Poppins world where everyone is flying around and all happy—we have to understand the psychology of social media and the negative effects it has, particularly on teenagers. When I grew up, there was no social media, and I have seen that social media has been probably a negative for a lot of teenagers. But some might argue, in environments where they don't have access to social centers or youth clubs, people can play online, interact with one another, support one another, and we don't see all those positive things because only the negative things are amplified.

So when it comes to AI, we have a similar quandary, which is how do we let the free market and free humans dictate what is happening. Because there is one argument, if you look at the difference between say Anthropic and maybe OpenAI: Anthropic basically won't let you do anything on the models because it has got some kind of higher purpose wired in, whereas OpenAI is like, "No, no, I am going to serve you as the entity." And actually, if I choose to use the object to do bad things and I am breaking the law, then I should be held accountable.

Host: Sure.

Lee Cronin: It seems to me that there is a split right now. But that is kind of obscured by the doomers saying AI could run away. And I don't understand the mechanism for running away. I do understand the mechanism for tinkering, I do understand the mechanism for putting disinformation out there, or for doing conspiracy theories that aren't yet possible. Here is one crazy thing: there are no real conspiracy theories in history, not many, because humans couldn't keep them secret. But if we had good AIs, you could play around with the future with the AIs pretty well, and that might be one thing to be worried about. Is there one big—if we say that post-Second World War, when the winners talk about what happened, could that be one big glorified conspiracy theory because obviously the winners write the history books?

Host: Yeah, there is always a percentage that is written by victors, for sure. As far as when you look at a story, it is never 100% what was said. I take issue with when you then try to say, well, if 5% wasn't what we were told, therefore 100% is not. You got to be careful with where you go.

Lee Cronin: Absolutely. And I think that is really important nowadays. That critical reasoning to what can I verify and what do I find impactful to me is really important. So I think AI will accelerate the teaching of critical reasoning. Right now, I worry a lot about how young people are going to be allowed to get their first job, make mistakes, get the training, the mastery required, have the hard work emotion. Frustration is the best trainer. Whereas people now, if you don't want to get frustrated, just use ChatGPT. But I think this is an aberration. I think these tools, a lot of them are really good. The coding tools are amazing. And so I do wonder—and this is me actually shifting my emphasis a bit—if AI might actually save us all.

Host: Save us all.

Defining Intelligence Like Measuring Temperature

Lee Cronin: Save us all, because social media is such a cluster, right?

Host: If you listen to my show, you know I take my supplements every day religiously. You have also heard me talk about how the supplement industry is really crowded with a lot of fake products, too. On top of that, everybody in health disagrees with everybody. If you have ever Googled a supplement, you can read such a range of opinions on what the expected benefits are, somehow you leave more confused than when you started. But you are not going to have that problem with Blueprint's Longevity Mix. Blueprint is built on scientific evidence, so the guesswork is over. Every ingredient they use earns their place through population-level research and real-world testing, not what is trending on TikTok. With Blueprint, instead of a whole cabinet full of pills, it is one scoop, water, mix, done. That single scoop helps support energy, cognitive performance, mood, focus, sleep, cellular resistance, and healthy aging. It includes ingredients like magnesium, taurine, glycine, creatine, and more. Blueprint is also third-party tested for purity and toxins like heavy metals, and they publish their results on their site so you can see exactly what is in their supplements. Blueprint doesn't just offer Longevity Mix either; they also have essential capsules, high-polyphenol olive oil, omega-3, collagen peptides, and more. This is science-backed, precision-dosed, no BS. For a limited time only, new customers are going to get 20% off plus free shipping at blueprint.bryanjohnson.com, link in my description below, by using code JULIAN at checkout. That is code JULIAN at blueprint.bryanjohnson.com, link in my description below, for 20% off. After you purchase, they are going to ask you where you heard about them. Please support our show and tell them I sent you.

Lee Cronin: If by using AI tools we can help us do critical thinking, help us verify things. The current AIs as classically trained weren't very good at being correct, but actually they are getting better at being correct because you can say, "Hey, go out and check the literature and check if this thing actually happened, and help me strawman or steelman a particular argument, and then also help me understand as objectively as possible how I might frame that argument." Because if nothing else, these AIs might be able to allow you to do one-to-one teaching. They have got infinite time, infinite patience, if you use them correctly. If we can use anything incorrectly, so I am saying wildly contradictory things, strong opinions loosely held.

Host: Yeah, for sure.

Lee Cronin: The doomers serve a very good purpose, but I just wonder if the doomers could actually help us by shifting their doom to the real issues that we need to deal with, such as bad actors using the systems.

Host: Right. Well, that is the thing that I am thinking about with your world right there that you are painting. On paper that would make sense. But again, if these different platforms are going to actually be a positive and help teach critical thinking and things like that for a person to be able to do that themselves, then I think a requirement for humanity across the board would then be you need to open source all the code at all times. Because you want to make sure that there are not human beings slightly injecting their little opinions into coding the actual AI organism that is going to teach us what we do.

Lee Cronin: AI organisms are cybernetic organisms. They are the combination of human programmers and the model working together, right? So I am very happy to say the sentience in the AI comes from the humans. This is really the nice mystery that we can talk about. But look, I am uniquely unqualified to say what the models do because I don't build—well, I build models based on data and very basic learning algorithms and systems, because I need a first-principles understanding of it. And if I can't understand it, then I don't tend to implement it. And the reason for that is I need to understand what the model is doing. A lot of these models have got lots of meta levels. It is a bit like having a conversation with someone and you think they are understanding you, and then you have completely different images in your head, and then you leave the conversation, they go and do something else completely different to what you thought they would do, and you likewise. So what you have got to be able to do with these systems is somehow ground them and understand conceptually what is going on in science. That is really important. The only thing I have exception to in AI, my hard lines are: the AIs are not intelligent, right?

Host: They are not intelligent.

Lee Cronin: They are not intelligent. They are not intelligent.

Host: How do you define intelligence?

Lee Cronin: Yeah, that is the question that makes sense, because people talk about AI, AGI, and superintelligence. What is superintelligence, everybody? Is superintelligence like some kind of magic that I don't understand? What is intelligence and what is artificial general intelligence? Let's go all the way back. I have got very strong positions on this which I think will endure the time. Intelligence I would define to be a property of an entity to solve problems that will allow that entity to survive. So this is like biology, right? A general intelligence would be an entity that can solve almost any problem coming at it to survive. You could say your entity is really good at mathematics. Now people will say, "Well, the AIs are good at mathematics and they do solve problems." So now we are saying, well, the AIs are very good at taking a prompt, a problem that you give it, and saying, "Please, when you have got this dataset, help me interrogate this dataset and give me the outcomes." And so what I am playing with right now is trying to engage with people positively about what we mean by intelligence. Intelligence is a very interesting thing, and what I will say to you is that we do not know yet how to define or measure it. I am working on a problem, a shadow problem that I am uniquely unqualified for or maybe uniquely qualified for at the end, which will have a go at measuring it. And I am really inspired by going back in science looking at problems and discontinuities in problems. The one that I really like is the concept of temperature.

What LLMs Reveal About Human Cognition

Lee Cronin: We knew in the past—or gravity, but let's take temperature. We know that things are hot and cold, but what is that? "Oh, that is hot. Oh, that is cold." We didn't really know how to measure temperature. We knew some things were hot and some things were cold. It wasn't until scientists and technologists said, "Right, what do we need to be able to do? We need to measure temperature, right?" How do you measure temperature today? Well, in the good old days you would have a thermometer. The thermometer could have a number of different liquids in it. You want a liquid that, as the temperature goes up, it would expand and hopefully have kind of a linear relationship, whether that is mercury or alcohol. So for that expansion, let's go back: first we knew that things are hot and cold, but we didn't have a scale. How do we get a scale? Well, we realized that liquids expand as they get hot and they contract as they cool down. Great. So let's build a system whereby we can measure the expansion of liquid. What technology is required? Well, we need glass. And what property do we need of that glass? We need the glass to be uniform, because if you made it fat and thin and fanning out, maybe a triangle shape, that ain't going to work very well. So you needed to make a really nice tube where the sides of the tube were parallel, was a perfect cylinder, or as perfect as you can get. Well, back in earlier times, there were some glassblowers in Italy that perfected how to glassblow tubes. Suddenly they glassblow the tubes, and then you put the alcohol in. You could go around and measure the temperature everywhere. Suddenly, "Oh, it feels cold here, we will go Antarctica this temperature." And of course, the Americans, we are going to have Fahrenheit. I don't know where you guys got Fahrenheit.

Host: I like Fahrenheit a lot better. It has more units. A 1-degree Celsius jump is a jump, but if it goes like 82 to 83, I am wearing the same thing.

Lee Cronin: You got that wrong, just take the L, please. Well, there is this temperature scale called the absolute scale, the Kelvin scale. So 1 degree Kelvin is the same as 1 degree C, it just goes down to minus 273.something or other. So that means 0 Kelvin is the lowest temperature you can get. You can't get any colder than 0 K, right? In fact, you can't achieve 0 K anywhere on Earth. We have built some really good systems that can almost get to absolute zero, and what that would mean is the molecules at absolute zero wouldn't do anything. They wouldn't even vibrate, they would just be stationary. But anyway, brilliant thing about temperature: technology was built to measure it. We knew things were hot and cold before, but we didn't know quite what it meant. Now, with intelligence, the problem is human cognition and psychology and IQ tests and all this stuff. And we have people saying, "Well, this AI will be able to solve any problem from a human." They are misunderstanding what intelligence is. Giving an algorithm a series of queries to solve a mathematics test is not the same thing as a human solving an unseen problem. And intelligence is actually the ability to solve an unseen problem. Unseen is hard, right? Because if it is not in my training, I can't solve it. But when I actually have the problem and I have solved it, I can train on it.

Host: Am I way oversimplifying this if I put it in the bucket of almost what you are saying is logic versus the ability to be creative?

Lee Cronin: Computers are quite good at using logic if you encode them properly. There are some good programming languages to do it and they can do lots of things. No, what I am saying is something super concrete: that intelligence is a thing we don't know yet how to define or measure. And I think for a little hobby, because I am basically bored and I haven't got enough work—

Host: You are bored.

Lee Cronin: I am far from bored and I am too busy. But someone needs to have a go at this.

Host: Hey guys, if you haven't already subscribed, please hit that subscribe button. It is a huge, huge help. Thank you.

Lee Cronin: And of course, I am uniquely unqualified, so maybe that will make me—I have almost finished it, so it is not that hard.

Host: Yeah, you are so busy you don't change clothes.

Lee Cronin: I have 12 different pink shirts in play today.

Host: All right, I am going to need to see some evidence of that, because that looks like the same shirt as last time.

Lee Cronin: Probably. You can check. Maybe the other one was a wide-collar one; these are narrower collars. The nice thing is I don't have a cognitive load, right? When I dress, it is just pink. But it could be like Mark Zuckerberg and he just wears gray, I think.

Host: Isn't he kind of based with fashion now? He is always wearing different chains and—

Lee Cronin: I don't know, me no billionaire. Just poor person.

Host: Hey, you just raised like $70 million, I don't want to hear it.

Lee Cronin: Sure, but I raised it to change the world in chemistry, not sort out my wardrobe. That is right. But let's go back to intelligence measuring, because I want to linger on it. You can measure temperature and therefore we can have a scale. We got there. We don't know what intelligence is. We have got the IQ issue, we have also got the race issue that goes along with it that is out there, and also the genetics issue which was all kind of crazy because psychologists and cognitive psychologists got confused. But then we got people confusing the AI is getting good at stuff as being intelligent. And so Elon keeps saying, "Oh, the AI, Grok, is going to be a PhD level in all these subjects right now, and therefore no PhDs." And that is such a fundamental fail.

Why? Because a PhD—let me explain how the AIs work right now, and I am going to convince you and hopefully by extension anyone listening or watching this. AIs today do the following thing: you get a lot of data. You take that data and you train a model. That model will be able to tell you what is in that data and find relationships in that data. Great. If all the problems solved in the past are captured in that dataset, when you give the AI a problem that has been solved in the past, it can solve it, right? That is not intelligence; that is incredible retrieval and retention.

Host: Exactly.

Lee Cronin: And sometimes the AIs aren't that good at it because they hallucinate. Because of the way the structures of some of these models work, they hallucinate. And so what happens is, because they have to give you back an output, the top AI providers are very good at having a checking system. They will say it will come out with an answer, and we will check it. You go to ChatGPT and say do a calculation. It used to just retrieve it using the model, but of course the model would not get it right because the training data on the internet was put there by humans, and so it would just give you a random number back. But now it says, "Oh, I recognize this as a calculation. Let me bring up my calculator tool, press the buttons on the calculator, and get the number back." The AIs are trained on this huge amount of data, and all the problems that we have solved up till now in humanity are in there. Potentially that is awesome. But a PhD scientist, mathematician, it is not about understanding the past—they are trained on that—it is about using the scientific method to solve problems. So whenever you give a new problem that is not in the dataset to the AI, it can't solve it.

Host: But what if it has enough data and enough retention from all the things in the world that it can keep at one time that a human can't, such that it can find abnormalities or holes or something to be able to plug it in a faster way than a human could? How is that not intelligent?

Lee Cronin: Again, we have to define intelligence. I would call that very useful, right? My calculator is very useful. There are many tools that have been built that are useful that enable me as an entity with human rights to do stuff. But Einstein with a pencil was much more intelligent than Einstein without a pencil, because he was able to use a pencil and paper to think. It is almost like a cybernetic thing. So we get trapped in these circular arguments: are the AIs intelligent, do they have free will, and are they sentient? Intelligent entities can solve problems and want to persist in time; that is called biology. I would say intelligence is a property absolutely reserved for biological organisms. Right now, what are the AIs? We can call them maybe augmented informatics. That would be the word.

Host: Augmented informatics. That is a lot to say, Lee.

Lee Cronin: I know, it doesn't sound as good as AI, right? Artificial intelligence sounds great. The AIs aren't magic. They are incredible models. In fact, they are revealing things about human language and cognition that we didn't expect. If you train an AI in a certain language and then you train it in another language, and you look at the representations of those entities—let's say king or queen, or brick on cement—they occupy a space that is pre-lingual, right, which is conceptual. There is a mathematical representation beyond the language in a space, which is really interesting, but not that surprising because humans use language to represent concepts. So if I take Spanish and I distill Spanish to find love and hate, hot and cold, up and down, and plot them in some space, and I do the same in English, and you overlay them, they are very close. This is why AI translation tools are amazing, right?

Host: Yeah.

Lee Cronin: I learned German; I am not very good at it now, but there is a great word in German called gemütlich.

Host: Gemütlich.

Lee Cronin: Gemütlich, which loosely translates as cozy. You go into a pub in the UK and it is a windy winter night and it is cold outside and there is a nice fireplace and beer, you go, "Oh, that is cozy. Nice." Germans have it, but it doesn't quite translate, right? It is a different kind of word. So there are cultural representations. The Inuit have a lot because the Inuit have to live in the cold a lot, right? But let's go back again: what is intelligence? We don't know. Can we measure it? We don't have a measurement. We know what temperature is. These AI tools are able to do very important, very interesting things; they can automate a lot of tasks.

AI Predicts, Humans CREATE

Lee Cronin: But I am going to make one statement and I will keep coming back to it: the AI only knows what it is trained on. And if something is embedded in that data and you query it correctly, you can get that out, which is great, right? And the AIs are showing there is a huge amount of knowledge that we haven't explicitly found on the internet in human knowledge that we can pull out.

Host: But it can't use the way that humans query and reason over time to put together patterns to be able to predict how they are going to query in the future and potentially outrun humans before we can get there?

Lee Cronin: I don't understand that statement until outrun humans.

Host: Because if they can patternize everything—let me give an example because it is a little bit complex. If I am talking to an AI and 100 other people in the population are talking to an AI, and we are all trying to figure out something about paint, and we are asking all these questions about different colors and how we paint a room and whatever, the way I ask questions is a series of queries. Even if it is very similar to the next person, the wording might be different or a specific thing I may be interested in might be different. But the AI is collecting at volume and at scale all these different ways of querying things, so the AI can use all of its computational technology to patternize what everyone is saying and be able to predict what the optimal questions might actually be. So in that way, the way I see it is it could be outrunning humans—front-running them is maybe a better way to put it.

Lee Cronin: No, that is not what is happening.

Host: Thanks for clearing that up.

Lee Cronin: There is a fallacy here which is super interesting, and again, I am very happy to be wrong on this, but I want to come to it with data. AIs are prediction machines; humans are creation machines. When the creations are kind of shallow, the AI is great at finding them. If you take all these Erdős problems or ways in mathematics that the AIs are solving, disproving certain conjectures, AIs are very good at disproving things by finding a counterexample. But an AI has not come up with a single creative mathematical act, right? It hasn't come up with a single creative scientific act. Now, people will argue about this. I want to actually stop this argument because it is obscuring what the AIs are good at. In the last year and a half, the AIs have supercharged my ability to be creative, because I don't have to spend my time writing code and collecting data and processing things. The AIs are splendid at it. They are really great, but what they can't do is they can't create. They can appear to create, and those creations are relatively shallow.

Intelligence I would measure by the ability to solve problems that are unseen. AIs are able to solve problems that are seen or embedded in their data. AIs can do new things, right, but they can't do novel things. And so the question is, what is the difference between novel and new? New is something that you can get from your dataset; it is just a combination of things, it is fairly shallow. You can search for something new in your dataset and find it in the end. So when the AI does something and you go, "Oh, that is exciting," you can see that it is just new and it is entirely embedded in that data. Humans are able to do something which I call novel, and novelty is defined by that thing that is in principle not predictable and not embedded in the data. Fashion and music are just human. You are not going to know what humans are going to do. You might have me back on your podcast one day, and I might come back wearing a color you weren't anticipating. Is that new or novel?

Host: Well, it is novel because in prior data you are just wearing pink.

Lee Cronin: Right, the AI is just going to make me in this suit. So when I come in wearing turquoise or whatever else, you would be like, that is novel. But there is a certain amount of misselling here because we are trying to sell AI as intelligent, going to replace PhDs, replace this.

Host: Mhm.

Lee Cronin: Human beings are going to use the AIs to massively flex their cognitive capabilities, not replace them. This is where everyone is going wrong. AIs are going to replace science? No, they are not, because science is about problem solving.

Host: Well, you actually just made a point a couple minutes ago, and I can't believe I have never looked at it this simply. It is like the most obvious 30,000-foot-in-the-air point. We have had this AI crisis being talked about extensively over the last decade, but particularly the last three years where everyone is using it on LLMs. But they haven't created their own E=mc² yet or anything like that, which is kind of making your point for you. They have not figured out the laws of quantum—I am just making up things right now—but these things that scientists are constantly trying to test every day. Therefore, not only have they not replaced the highest-level guys like you, they haven't replaced science in any way at all technically to this point.

Lee Cronin: People have used AI to come up with very good new ideas for making new drugs, folding proteins, solving some physics problems, but these problems are already well-defined by the human for data processing. So the AIs are a great tool. They are a tool. There is something really interesting sociologically and culturally happening. We have this problem right now where people are saying the AI labs need a lot of money, they need a lot of infrastructure, they need a lot of energy. And in academia there is this political push against academia: academia is full of left-wing people who don't critically think and are all woke and so on, therefore let's just replace all academia with AIs to do science. That is just such a fail on all levels. Sure, academia has problems, ideologies are everywhere. But given that the AI builders don't know what science is—they simply don't, they are populated by computer scientists and mathematicians now. And that is not me being elitist; it is just solving scientific problems in a wet laboratory, whether it is a biology lab, chemistry lab, physics lab, or mechanical engineering, is a very visceral thing, and it requires creation. Like the governor in a steam engine: someone is like, "Oh yeah, I need to make sure the boiler doesn't explode, so I need to build stuff." We interact with the world.

Now, I am building a world model for chemistry. What does that mean? Well, I am just learning all the chemical reactions by doing them, and then once I have learned them I will be able to make any molecule. That is pretty cool, but there is no magic. I have to collect the data, I have to have a verifiable process, and then I have to be able to run it. Science has never been in a more exciting place. It is just a shame that we have this culture war going on at the same time, because we have to make a decision: do we want to spend all this money on GPUs, CPUs, water, energy, whatever, to crunch the data and give back products to people, or do you want to spend the money on humans doing that as well? How much does it cost to produce a human that can solve a problem that you can use ChatGPT for? There is no magic. If I want to replace a human with ChatGPT, I can, but what is the cost? And I have got the human out there who now has not got a job to do. Maybe the human will do something more interesting. We don't yet know what intelligence is; we haven't measured it. We are using AIs to do very important, very interesting things at scale and at speed, but it is not diminishing a human.

Host: You said at the beginning of when you were explaining this with the intelligence loop, there is three layers: we often talk about intelligence, general intelligence, and then superintelligence. So if I am understanding you correctly, because we can't even measure the bottom layer, you are saying it is like 10 leaps and a skip to get to the next layers, obviously.

Lee Cronin: Yeah, I think the providers of AI—and I actually like the products, they are great—

Host: I don't know about your glasses though. They were doing some math before. You might want to check those things.

Lee Cronin: I am not doing any product promotion on your podcast. They are not Meta glasses.

Host: Yeah, but Meta might have actually gotten that right, not that I am trying to shill for them.

Lee Cronin: No, they worked, it was just—anyway. Let's go back. AGI: lots of people will say that an AGI is the term we would give to something that can automate most tasks. Now, is that an AGI, an artificial general intelligence, or is it just an artificial general automator?

Why AGI Isn't Actually Intelligent

Lee Cronin: It would be great if I have an AI that can clean up my emails. Actually, I do use an AI to clean up my emails; I use one that is about 8 years old. What I do is, I am an obsessive email cleaner. I answer all my emails that I want to answer and I basically file everything. When new stuff comes in, it doesn't get lost in the noise. I get so many emails every day, I find it incredibly depressing because I lose stuff in the noise I need to answer. I have only got so much cognitive bandwidth. So an artificial general intelligence that the frontier labs are discussing is literally an entity that could do a load of tasks. Does it solve general problems? No, humans are good at this. And maybe there are some jobs that humans have that are filing, but our failure in human society right now as people go from school to university or into a job is maybe we need to generate jobs that have more value and meaning. We can do that, and AI provides a great chance for that.

Now let's go to superintelligence. Nick Bostrom made up this term.

Host: Yeah, we got off this last time, we didn't go deep on this. I have read that whole book.

Lee Cronin: It means just making stuff up, right? If you can't define a thing and you say it is super, I mean I could be a super chemist, what does that mean? A superintelligence for me would be somehow beyond our current understanding of physics and mathematics, like magic. Again, superintelligence plays on this kind of, "Oh, there could be this entity that is way more intelligent than me, knows more." What does that mean? Human beings are able to learn how the universe works; we are able to understand far more than maybe evolution has equipped us with, it would seem. Now, what the doomers say is, well, this superintelligence could understand how to do things that could basically end humanity or build a super-duper weapon or something. Well, I don't understand the evidence for that. And so what I try to do is deflate the term superintelligence, because I don't know what that means.

Now it could mean, let me take a stab at three different things: it could be just a faster way of thinking. Sure, chess computers can do chess really fast, so I could buy that, but why call it super, just fast? It could be that it is able to play through some kind of chain of reasoning on decision theory, and if I plant these seeds in these decisions I can basically create disruption, like I could disrupt a market by planting a meme over here, over here, everyone buys their stocks in a certain way. Could do that, but the market is so noisy and humans are so good at doing weird stuff, that is going to get washed out. And the other thing is maybe suddenly it can come up with a brand-new thing that doesn't have precedent. Well, we haven't ever seen that. You just said there is no E=mc².

The problem is within AI, let's just imagine we go back to the time of Einstein. With special relativity, Einstein basically asked the question: what happens when I sit on a light beam? He started to realize the speed of light was the fastest speed limit in the universe, you can't go faster than light. What happens then in general relativity? Einstein said mass curves spacetime, and things don't take a straight line, they go through curved spacetime. Cool. So suddenly when you use general relativity, you are able to think: if I fire a satellite into space, the further the satellite gets away from the Earth, the greater the drift in time on the atomic clock on the satellite versus the atomic clock on Earth. And in fact, if the satellite is putting out a pulse with its clock, and I measure it and compare it with my clock on Earth, I can use that to position using GPS, right? We understood general relativity before we put satellites into space, and when we put satellites into space we understood frame dragging and we could build GPS. Hold that thought. We have built these systems which can basically keep distilling data and finding relationships, and we are assigning a property to these relationships that we don't understand because we haven't understood the fundamental theory of intelligence. In the same way, imagine if we put satellites in space without general relativity; we would be like, "What the hell, they are losing time," and we would have made all sorts of stuff up just to understand that. So the frontier labs have got to some incredibly capable systems that appear to reason and do things that we didn't anticipate before, and that is because we don't have a good theory for intelligence.

Host: Now, I like the baseline there of you actually already have something defined so that then you can measure it once you test it, like Einstein and testing satellites with GPS. If I go back to what Bostrom did pretty early on here, all things considered with how early he wrote that book in our recent timeline of AI, the way I always looked at it was he created a bunch of decision trees. He created possibilities. And I think I buy your argument that there wasn't an underlying definition on that, but his decision trees pointed out on a scale from things could be fine all the way to we are doomed. It pointed out if this, then that; if that, then this; and it created all these different worlds where you could see code getting out of control, where some things he described—I wouldn't even make the leap and say this AI became sentient, I am just saying it was coded to do something, with the mistake of a human being. And one example was it could overcreate paperclips and cover the entire surface of Earth in paperclips and drown us. I see things like that, it sounds crazy, but at the same time I could see that kind of situation happening if—and this would actually go to your argument—human beings are not responsible with how they create the code. So do you think it is useless if he is pointing out things where we make mistakes and then the AI does things that end us?

Lee Cronin: The AI won't do anything. The AI is programmed by humans. All the decisions come from the human. This is a thing that people don't understand. And this is one of the reasons why I think it is kind of fascinating. My research, why I have built a company to make a world model for chemistry and discover drugs and new stuff, why I am basically building engines for chemistry, is I want to understand the transition that we go from physics to chemistry to biology to cognition or culture. What is happening there? We don't really understand it. And the issue I have with superintelligence is it requires a mechanism that doesn't exist. You say you can see the leap: if this thing, then that thing. Okay, let me say I am afraid of AG, anti-gravity. What is AG? Well, anti-gravity is a thing that happens when I build a new technology and suddenly it turns off gravity and we all float away and we all die, because we are not stuck on the planet anymore, we have no air, nothing. Shall I write a book on super anti-gravity that could appear one day? Sure, I could tell you a decision tree: if anti-gravity, then float. But what is the mechanism for the anti-gravity? I just made something up. Unless you can define something and understand the mechanism—paperclips, what nonsense! If you fly on an airplane above the world, most of the world is not covered with machines; it is empty space, fields, trees, oceans. What do you mean paperclips?

Host: He was using a ridiculous example of something.

Lee Cronin: Yeah, but it is entirely ridiculous. There is no scientific merit in it, it is divorced from reality. It is fantastical nonsense, and it is science fiction in the worst way because he plays it off plausibly and calmly. All thinkers, Nick, all of us, have a duty to actually criticize our own ideas. We do not understand what human consciousness, sentience, decision-making is. The fact that people are going to start saying in a few years that AI should own its own copyright is just an excuse for people who basically want to take your copyright, feed it to an AI model, and then—if I am being really rude about it—I am saying, "Oh yeah, thanks whatever company for reading the internet and stealing my IP and training a model and then selling my IP back to other people." And then you have the AI pros going, "You just can't accept that the AI is smart." And I am like, "No, it just stole my IP, right? It literally just stole my creation. Shall I steal your house? Is that okay because the AI said so?" No. There are three things: we don't understand what intelligence is, we don't understand what creativity is, and we don't understand what novelty is. The fact we don't understand those three things and we are using them all interchangeably—I think I can tell you how it works.

The Drake/Weeknd AI Song and Plagiarism

Lee Cronin: The reason why everything appears computable is that once something is created by the world, I can then label it and put it in a database. I can decompose it, but that doesn't mean I can predict the future faithfully. I can predict the next token. And sometimes, if the search space is shallow enough, it is going to appear to be acceptable for some evaluation, right? But what we are going to see is that what humans are doing is quantitatively different. They are in a different universe to what we are doing with silicon compute, which is great. Again, I don't think there is something beyond the material; I am a materialist. But similar to Roger Penrose, and I think my collaborator Sara Walker, many other people, we don't know what that material is doing.

Host: Can you expand upon that? Like when you say we don't know what it is doing.

Lee Cronin: Yeah, so I don't understand how matter is able to process information in space and time. We have some models, but there is so much we don't understand about how biology is created by chemistry, how evolution produces biology, and how biology then produces consciousness and intelligence. We just don't know. We are at a really wonderful time where we have all these tools and data, but AI is really good at predicting the past. I will say that again: AI is the best at predicting the past because it has the past. Let me qualify: AI is very good at predicting the past when it can train on a faithful representation of the past. What it cannot do is predict the future, because the future is something different. The past is a probability space; what has happened, you can look at probabilities there. The future is a possibility space. Until people understand the difference between those two things, we will keep confusing it.

Host: How do you define the difference?

Lee Cronin: A possibility space is something that hasn't happened yet. A probability space is something that has happened, because I can't define a probability on something that hasn't happened yet. What do I mean? If I take a die and I roll it, I know that I am going to hit 1 through 6. I know it has happened before, I can train my own model on it. But I could create a die on paper that you can't anticipate, and therefore it is a possibility, not a probability. Until we understand the subtle difference between those two things, we are going to continually misunderstand what AIs are actually doing. There is a lot of vested interest and money in trying to shore up these models. But this is why AI can never do science. AI is a tool that I can use to do science.

Host: Yeah, it is not a rocket; it is extra nice rocket fuel.

Lee Cronin: 99% of scientists will argue with me and say that AI does do science, but they just don't understand what science is. Science is the ability to identify a problem and then to create a hypothesis about what is going on, and then to create an experiment to test it. What AIs are very good at is running experiments. But all that PhDs do is assemble problem statements. That is why doing a PhD is so annoying. Some of my PhD students are like, "Why are you making me think in this crazy way?" But it is really hard to identify a problem or an anomaly. Lots of people think the AIs are able to do this. They are not. The AIs are doing something really shallow. They can identify anomalies looking at large datasets and finding a hole, but that is not the same thing.

Host: How is it not the same thing?

Lee Cronin: One is the inverse of the other. You can use the AI to say, take this data where I have already defined the experiment and find an anomaly within that experiment. That is great, but I have already defined the experiment. What the AI can't do is actually define an entirely new set of experiments. They will always define experiments based upon what they know. The AIs doing science are a big delusion. The other day I used an AI to write a simulation of the origin of life using assembly theory. I said, "Hey, here is what I want you to do: here is my assembly theory calculator for molecules, let's generate a load of random molecules and join them together." And basically it does a pretty good job. It uses some toolkits from GitHub, puts them together with the code, and generates a simulator. But it was generating random molecules that were batshit crazy and not stable, and I said, "No, don't do that, remove that, remove that," adding constraints. I curated it to my taste.

Host: Right, you had to fix it up.

Lee Cronin: It was my taste, right? I feel like a music producer: "That didn't sound too good, I will change that." A lot of the technical stuff in science the AIs will replace, and we need to train people to use those. But we mustn't use the AI just to do boring science and pretend science. When the AI starts to produce new theories, explain quantum gravity, or come up with something that we could never conceive of, I will be looking out for that. As of today, August 2026, silicon-based computers as we know them are not capable of creating novelty. That doesn't mean to say we will not create a machine that is able to go beyond computation as we know it.

Host: So are you saying we could create something that right now is not something we are labeling AI effectively?

Lee Cronin: Sure. If you want to create a brain, there is the old-fashioned way of doing it: I have got two brains I created, one is 20 and one is 18 this week, my two boys. And they were also created by Darwinian evolution going back to LUCA. I need to create life in a lab before I can understand consciousness. Probably there is no other way to get there. We are making a massive leap, and maybe we are entering in a new dark age of science, like the Middle Ages, where creative science will stall because the weaponization of science to produce outputs and papers will go through the roof. Everyone will use LLMs to write papers, use LLMs to assess the papers, use LLMs to write grants, use LLMs to assess the grants. And who is going to have the taste? Then Bostrom's prediction of paperclips will come true, but in the scientific world, and the paperclips will just be papers and grants.

Host: Well, it is a stretch from what he was saying, but—

Lee Cronin: Same thing, right? What happened with the printing press? Everyone said it was terrible, writing things down would take something away from us. But actually, if it wasn't for our ability to write, we wouldn't be able to create what we created. So what I am excited about is what we will be able to use the AIs to help us create.

Host: Yeah, and this is where I think of the common example when you look at music that AI creates. There was a whole stir a year and a half ago where AI made a song of Drake and The Weeknd that actually sounded pretty good, and people were like, "Oh, is this going to be a problem?" But it didn't take off. First, because people learned it was AI, and there was something about that where people were like, "Human creativity didn't make this, I can't feel the same way about it." But the other reason is going back to your point: it didn't make anything novel. It took what the types of music those guys had already created and just took that style and said make it again.

Lee Cronin: I will give you a scoop from something I am writing just now. If I come up with something novel right now and say, "Hey, here is this novel thing," we know full well that we can then label that, put it into a database, and understand the pathway of getting there and therefore the probability we get there, because now we have a precedent. AIs can combine things together, but the size of the space of combinations for the AI is big, but not that big. Roughly speaking, if a space is big but searchable, the AI can do quite well in a shallow space. What humans are able to do is pluck something out of the air that is so deep you cannot ever search it with the compute you have available. Humans are surfing on the wave of almost infinite possibilities. That is where creativity is. And as soon as they find something, they chuck it back, and suddenly they go from that infinite edge to the finite encodable present. AIs predict the past, they are encoding the present, and the interface between the present and the future is where the infinite is a continuous substrate. When something is continuous, it is infinite, and you can't in principle know what is going to come. That is where creativity comes from. A lot of people are pretending to be creative, but they are just plagiarizing one another. If something exists in the database and you literally put it together and pass it off—I have done it. I invent so many ideas all the time and then I am like, "Oh, someone else had that idea, it wasn't my idea." Does that make me a plagiarist? No, because I didn't knowingly take their stuff and pretend it was mine. Most of my life I have shallow ideas, just a shallow thinker. But occasionally you have a thought that is quite novel. And humans are quite good at doing that: great artists, musicians, fashion people surfing that edge between the present and the future.

Host: Where do you think ideas come from?

Lee Cronin: No idea.

Host: You have no clue?

Molecular Computing and the DNA Computer

Lee Cronin: I have something that I am going to propose. If I wanted to give you a recipe for having a novel idea, something creative, what you want to be able to do is make a leap into a space that is so big you could not in principle search it. And what humans are really good at is imagining in that space. Imagination is uncomputable in principle, because it is too big to compute. Everyone says, "But that is stupid because your brain is imagining." Yeah, what is my brain doing that is not computable in principle? Imagination seems to be proof that the human brain is doing stuff that is not computable by a Turing machine, a counting machine, a labeling machine.

Host: How are you going to build a brain then, if it performs tasks that aren't even computable? How can you compute in the real world to build that?

Lee Cronin: I take a physical object and interface with it digitally. I plug it into a digital computer. And then the brain is able to come up with imaginations that the digital computer could never predict or conjure because the size of the space is too big. There is something about the polymer brain that is able to collapse the interface. I have an intuition for how it might work, but I am struggling to put it into proper mathematics.

Host: How long have you been working on this?

Lee Cronin: Quite a long time. It has been in the back of my head for maybe more than 20 or 30 years.

Host: Oh, wow.

Lee Cronin: Yeah.

Host: We got on this tangent a while ago on AI that originally came from the brain. And the way you were explaining it was after you created 49 wells where you were getting 0s and 1s, on and off with all the lights, you were then taking that and putting that into gels. Is that what you said?

Lee Cronin: That is the next step: to build a brain gel right now.

Host: Can we go back to this and explain that some more?

Lee Cronin: What I am trying to do is make a material where processes can be encoded digitally but then become analog, a bit like an oscillation: on-off, on-off, tick-tock. You encode it digitally because you nudge it with a pulse of electricity, and then you read out what is happening later. I want to understand how that digital encoding and that readout are related, and is there a complex or a linear relationship between the two? Right now, the phenomena that we are looking at I don't understand, and I infuriate a lot of computer scientists doing what we call molecular computing.

Host: What is that?

Lee Cronin: They are using DNA. DNA has four base pairs, and they can connect those base pairs to a computation. They can make a Turing machine, basically a digital computer, and make a DNA computer.

Host: A DNA computer?

Lee Cronin: Yeah, they have done it. DARPA funded it.

Host: DARPA, man. Andy Jacobson said they were talking to dolphins telepathically in 1992, I am just saying.

Lee Cronin: DARPA is a funding agency. They fund people, they give people money, and the people do the crazy things.

Host: Have you been to one of their secret labs?

Lee Cronin: They don't have any secret labs.

Host: I don't buy that. That is what I would say if I were funded by DARPA.

Lee Cronin: DARPA is based in the Washington area. They have program managers, and it is all out in the public domain. They identify very interesting areas and they fund them. They come up with an idea that, if this could work, it would change the world. The internet changed the world. Self-driving cars: DARPA had a self-driving car challenge where they had cars going through a desert, and they were falling over, but it arguably accelerated a lot of the visual compute and robotics. So I don't want to burst the balloon of DARPA.

Host: No, I think it is a really fascinating place for sure. It is not like when they come to you, Lee, they are like, "Here is all the 1 billion things we are working on right now." They are coming to you for a very specific thing because you have an area of expertise. When they approached you with the first grant, were they saying, "We are working on this kind of thing, that is what you work in, do you have an idea that you could add?"

Lee Cronin: It is very collaborative. They have a number of different performers that are funded, and we all get together and share and pool stuff. They are trying to push together the boundaries of science. The UK has got a DARPA version as well called ARIA, the Advanced Research and Invention Agency.

Host: And they are cool with you working with DARPA even though you are a UK guy?

Lee Cronin: Sure. ARIA will also fund stuff in the US, the same way DARPA has funded people.

Host: Oh, that is interesting.

Lee Cronin: They fund where the best ideas are. Obviously there is taxpayer accountability, but if someone is building something of great use to you, it makes sense. Going back to the DNA computer thing: there are computer scientists thinking about DNA as a computer, and they reduce it to this conventional computing paradigm. I am suggesting the brain is not a conventional computer. The brain is capable of doing things we don't understand, i.e., imagination. Isn't it wild? You can imagine a thing that doesn't exist yet, and that imagined thing has causal power because you can imagine that thing and you can make it work. This is what entrepreneurs do all the time: "I am going to make this widget, a mobile phone or an iPod Nano." Human beings are uniquely able to imagine a thing that doesn't exist—it just exists in their head, molecules and synapses—and you can literally grab that from the future and drag it into the present.

Host: Specifically from the future.

Lee Cronin: That is what I like to call it, because for me the future is unpredictable and is a possibility space, and imagination works in possibility space, not probability space. Whoever at SpaceX came up and said, "You know what, we will make the rocket out of steel and we will catch it. We are going to make something the size of a skyscraper that will take off, go supersonic, land down subsonic, and by the time we catch it we hear the sonic booms." That is nuts, but it is now science fact. This is why humans are really good at science fiction, like Jules Verne. Digital computing is not everything. And the fact is, we think that the entire world is a simulatable entity. Nick Bostrom said we are all in a simulation, but that is basically unfalsifiable.

Host: How is it unfalsifiable that we are in a simulation?

Simulation Theory Called "Rooted in Faith"

Lee Cronin: Where does a simulation exist? You have an infinite regress.

Host: What do you mean by that infinite regress?

Lee Cronin: If I am in a simulation, great. The simulation has to exist somewhere. Let's say I exist in a computer somewhere. Where is that computer? Is it another simulation, and a simulation in a simulation? If I can't falsify simulations all the way down, therefore it is the same as having a religious commitment, a belief. I am not saying I can't falsify God, therefore God doesn't exist; I am saying it is not amenable to the scientific process. If something is not amenable to the scientific method, then it is a commitment of faith. David Deutsch writes about this really nicely in The Beginning of Infinity: the simulation argument is a garbage argument because it stems from faith.

Host: The part that I wonder if it is actually beyond faith is twofold: number one, we don't even know how far the universe could go and we can only know what we have been able to physically observe. And number two, there are unexplainable things that happen in human patterns in a way that would suggest that perhaps we are within layers of simulation.

Lee Cronin: We are in the universe. Who created the universe?

Host: Who do I believe created it?

Lee Cronin: We can say there is a creator, or it just is. But the fact we can't falsify that makes it a nice conversation. Was there a designer, was there a God? Sure. But what can we use the scientific method to explore?

Host: But you said you are not a Lawrence Krauss guy that something came from nothing. You believe it started from something.

Lee Cronin: The "something from nothing" argument is, for me, kind of similar to the simulation argument. They both have fundamental scientific flaws. I prefer as a materialist to work in the material world and do experiments on the stuff around me. That doesn't mean we do not exist in a simulation or something didn't come from nothing; it is just I am not able to build an experiment to falsify that.

Host: How do we find a way to build an experiment to falsify that?

Lee Cronin: I don't think it is possible. With my small intellect, I am not able to falsify that right now.

Host: They couldn't falsify gravity in the 1400s, and then an apple fell.

Lee Cronin: Yes they could. They were throwing stuff at each other. They just didn't know how to formalize it. This is where philosophy becomes really important. If we are going to assume there are certain things that exist—an ontology or a metaphysics—and then look at relationships within that metaphysics, those layers have to be self-consistent. What you will find with both the simulation hypothesis and something from nothing is an infinite regress. With an infinite regress, you can't make progress. So you put that to the side and say that is something that is outside the scientific method. That is why the simulation argument is silly to keep discussing as if it is a serious argument, because we can spend our life discussing so many other interesting things: how do we cure cancer, can we go to Mars quickly using a fusion engine? These are far more interesting questions because we can affect them.

For me, simulation is super interesting because I am building chemistry robots. If I build a robot to do something and I build a simulation, I need a thing called verification. In the AI world, where the AIs are working really well is they have got very good verification on certain problems. One of the things I am inspired by from AI is building the correct verification loop for physical chemistry. I invented a programming language for chemistry a few years ago and everyone said, "This is nonsense, you just made it up." I did make it up, but it is not nonsense. The idea came to me because I wanted to program my 3D printer to do chemistry, and I wanted to make sure the 3D printer didn't catch fire. So then I built this programming language, and that allowed me to make a digital twin of it so I can verify and test it in a sandbox. Simulation that allows me to verify something in the real world is great. If I can take the real world, make a model of it, and check that when I do it in the real world it doesn't catch fire, I will do that.

Now, that doesn't mean to say that one day we can't create a simulation where we can put digients inside. Ted Chiang wrote a really great novella called The Lifecycle of Software Objects. Ted Chiang is a fantastic thinker, writer, external faculty at Santa Fe Institute. I have met with him several times, debated assembly theory with him. These digients are entities that live in a world and can manifest in physical robots. I could imagine a world where I could create a sufficiently rich computational universe where I could create things that might appear sentient, which I could do tests on, where I would actually have to think about moral philosophy and ethics. I can imagine that, but I don't know what physics I would have to build for that because I don't understand what life is, how chemistry became biology, how biology generates consciousness, and how consciousness makes intelligence and free will. Until I know those things, I can't build a simulation of them.

Host: Right, I got you.

Lee Cronin: Bostrom, he is a really nice guy.

Host: Have you ever talked with him?

Lee Cronin: I haven't. I have been at several meetings when he was there, but I think I would probably just be too annoying. I would make the Bostrom confabulator: a philosopher that just puts a load of things together and pretends they are real. What would happen if we simulated the wrong coffee cup in this alternative universe and it caused a singularity and the entire world disappeared?

Host: I don't know if he went to that. I think he went to where things are coded to create or destroy in general, not like floating objects that defy physics.

Lee Cronin: It has allowed people to think there is this thing possible called superintelligence. It is a bit like saying we are going to go faster than the speed of light. You can't go faster than the speed of light, ever. It is the law of the universe. As you put energy into a mass to accelerate it, it radiates radiation to stop you getting there. We do that in proton beams all the time at CERN. The speed of light seems to be 99.99999% absolute.

Host: So you are saying there is a chance.

Lee Cronin: No, there is no chance. But as a scientist, it is impossible for me to say anything as absolute because I have to have an open mind. Putting limits on things allowed us to build a technological society. When you go to a hospital and they put a stent in your artery, there is a chance you will die, but look at all the statistics: the probability that you can survive with a stent has prevented many deaths. You never say never in science, you just go to the edge of the scale and say, "I am 99.999999% sure we are not going to go faster than the speed of light."

Host: When you referred to CERN, are you involved with CERN at all?

Lee Cronin: No, I am not. I have been to CERN, there was a TEDx talk at CERN. CERN is a great machine in terms of the media associated with it, the way they drum up interest. It is a massive experiment and it costs a lot of money. We used CERN to find the Higgs boson. Great.

Was the Human Genome Project Worth It?

Lee Cronin: Was that worth 5 billion euros or dollars or whatever it was? I don't know.

Host: That seems like such a huge part of academia: how much can you drum up buzz and hype and creative sci-fi interest in the general public versus actually funding the ideas that have the merits of being the best to create something groundbreaking?

Lee Cronin: When you fund any scientific enterprise, you get rewards coming out, and then when the rewards tail off, what do you do? You probably kill it and start again—creative destruction. Was the Human Genome Project worth it? Well, it wasn't worth it at the time, but it is going to be worth it now because we now know how to edit the genome.

Host: Can you explain that more?

Lee Cronin: When the Human Genome Project was done, it was a massive collaboration between the UK Medical Research Council and the US National Institutes of Health. They spent quite a lot of money building the technology to sequence the human genome. And then Craig Venter, a great pioneer and entrepreneur, started a company to beat them to it because he wanted to get the IP. To justify it, we said, "Well, we are going to cure all disease." And the human genome was solved, and we didn't cure all disease. What is beginning to happen now, which is really exciting, is there are a lot of gene therapies where we understand how to correct a diseased gene. There are young people that have been born blind where they have been able to correct protein by gene therapy and they can see. That is amazing. There are also gene therapies where we are understanding how to reprogram the body's immune system to cure cancer. There is a very strong chance that most cancers we know about, within our lifetimes, will have a combination of molecular and genomic therapies where we will use small molecules.

Host: Do you think they already have them and they are just not giving them to people?

Lee Cronin: No. Why would you—because that is a stupid thing to do.

Host: Why is that stupid? We have utilized the same type of brutal treatment for decades now.

Lee Cronin: There is no conspiracy; that can't exist. There are lots of pharmaceutical companies doing medicinal chemistry. Developments in biology are moving apace, but the most critical problem is we have to translate those developments from the lab into the clinic, and people are hard, and people die. There is an interesting contrast between China and the US and UK right now about how much regulation we have. There are some clinical trials going faster in China because of different regulation. In the West, we are much more risk-averse. Could we go faster if we accepted a few more deaths? Probably. Any limitation that we have in the West is a function of our regulatory environment, which is a function of our voters.

Host: China has moved in some ways that have taken risk with that because they don't have the same guardrails on it. There is something to be said for having guardrails, for sure.

Lee Cronin: There was one Chinese scientist who used gene therapy to edit some embryos that became humans and caused a problem. He got put in prison because he basically misedited the genome.

Host: What did he do that they are not going to have as good a life?

Lee Cronin: I think it was something to do with the HIV virus, but there was something he did that was not particularly smart.

Host: Chinese scientist who produced genetically altered babies sentenced to 3 years in jail. He Jiankui and his two collaborators were found guilty of illegal medical practices. Let's see what they did: the Chinese researcher who stunned the world by announcing he had helped produce genetically edited babies was sentenced to 3 years. The court in Shenzhen found that he and two collaborators forged ethical review documents and misled doctors into unknowingly implanting gene-edited embryos into two women, according to Xinhua, China's state-run press agency. One mother gave birth to twin girls in November 2018; it has not been made clear when the third baby was born. The court ruled that the three defendants had deliberately violated national regulations on biomedical research and medical ethics, and rashly applied gene-editing technology to human reproductive medicine. All three pleaded guilty. In November 2018, he announced that he had modified a key gene in human embryos in a way thought to confer resistance to HIV. He wanted to spare the babies the possibility of becoming infected with HIV later in life. But it caused other serious drawbacks.

Lee Cronin: That was completely the stupidest thing to do, because HIV is pretty much now—it devastated populations, mainly homosexual males. And then we got therapies to put it under control, and now you can get it under such control that there is no viral load, which means it is almost impossible to pass it on. You can chemically control it and suppress the virus. What an amazing accomplishment for human medical technology. He just went one step too far. We have to debate how we do gene editing in the future. The Human Genome Project was an incredible achievement, in the same way that the use of CERN to find the Higgs boson was an incredible achievement. We are in this illusion right now that AI labs are going to automate all science so it will be cheap and we will just cure all disease. That is just not going to happen. The AI tools will help us cure some disease and produce new molecules, but I don't know the exact extent.

Host: If it is just a huge accelerant that allows us to solve things way quicker where human beings are still leading the way, this is the opposite of doomsday.

Lee Cronin: I have sympathy for the AI doomers because they are trying to ask politicians to understand what is going on, and I have sympathy for the AI abundantists who want to use these tools for human flourishing. What I don't like is the extreme: that we are all going to die, or we are all going to live forever and have infinite stuff. Both those things are clearly stupid.

Elon's Claim: No Money in Ten Years

Lee Cronin: There is a worry that we will use AI systems to hack into computer systems and all software is going to be perpetually unsafe: bank accounts, encryption, and so on. You don't want your Tesla to go nuts on the motorway. On the flip side, you want to use AI to accelerate technology and cure disease. But some AI people getting quite rich are saying, "Oh, money won't exist in 10 years."

Host: Yeah, I don't understand that argument at all.

Lee Cronin: It is just Elon making stuff up again. Elon is a genius, but he is not a genius at communicating.

Host: How does money not exist to where people are all just going to be the same, especially that quickly?

Lee Cronin: Money is a representation of the allocation of resources. Where there are humans deciding on what to do, money will be required. The resources required to build certain things will drop and asymptote towards zero. There are some microprocessors you can buy that are just 15 cents when they used to be many dollars. But there is always going to be a bottleneck: Planet Earth has a finite size, there is a finite amount of accessible energy today, and a finite amount of physical resources. But there is an infinite amount of possibility. We are going to recycle things, find new energy sources, and create entirely new economies. This idea that we have infinite abundance in less than 10 years might be a reaction because the people creating these AI tools have been astonished by their capability. You can use Claude or ChatGPT to make a PowerPoint and it is pretty good.

Host: Yeah, you can make apps with these things too.

Lee Cronin: But if you tell Claude just make a PowerPoint to do X and you don't give it enough information, it is garbage. Garbage in, garbage out. If I say, "Here are the points I want to make and here is the fundamental data for them, please make it and iterate with me," I have no problem using AI tools to unleash my creativity. When it comes to writing, do I use the AI to write for me? No, they write in such a semantically uniform way. It is always, "It is not A, it is B." Please do not write like that! My writing is flawed, but it is my voice. I do use the AI now to correct typos, but it tries to clean up my grammar and I say, "No, I want it to sound like this." What I found quite fascinating is I use the AIs now to look at all my paper proofs. I had this paper that came out in PNAS on alien detection systems: "Molecular Assembly Is a Universal Biosignature Measurable by Mass Spectrometry." I put the proofs into ChatGPT and it went through and found 10 typos I did not find.

Host: Let's read the abstract: "Detecting life beyond Earth requires biosignatures that do not depend on the chemistry of known organisms. Molecular assembly (MA), derived from assembly theory, quantifies how difficult it is to build a molecule from basic building blocks, linking complexity directly to selection and evolution. Here we show that MA can serve as a universal biosignature that is both interpretable and experimentally measurable. Unlike information-theoretic measures, MA can be inferred directly from mass spectrometry data without structural elucidation. We demonstrate that using a machine learning model trained on standardized single-stage spectra predicts MA with three-fold lower error than baseline methods. Simulated multi-stage data reveal that small instrumental variations can double prediction error, highlighting the importance of calibration. These findings establish molecular assembly as a physically grounded, quantifiable biosignature measured by mass spectrometry whose interpretation depends on a careful control of instrumental effects, offering a scalable route to life detection on future planetary missions." Let's go down to the graph.

Lee Cronin: You can use machine learning because there is a theory there. People have been trying to use machine learning to look for life and they were just making stuff up; assembly theory helps you understand the basis for it. Temperature: we know things are hot or cold, but how do we quantify it? With life: we know things are alive or dead, but how do we measure it? Living systems uniquely make molecules with many different parts in high copy number, so we can measure that. Our equivalent to the thermometer is a mass spectrometer. What a mass spectrometer does is weigh a molecule, fires it into a vacuum electric field, and measures how heavy it is. Then you hit it with energy and it fragments, like taking a plate and hitting it on the ground. You count the number of parts, and you can use that to measure the assembly index. There are three mass spectrometers on Mars right now, and NASA is sending a nuclear-powered mass spectrometer to Titan called Dragonfly. It is a quadcopter powered by a plutonium slug that is going to fly around Titan, sniff the air, and use mass spec.

Host: Can you do any of this to try to find life on exoplanets?

Lee Cronin: You can. There is a paper coming out soon where we use assembly theory to measure the probability of life on an exoplanet. The exoplanet world is very used to talking about one marker for life, like methane (CH4), oxygen (O2), or water (H2O). These are all very small molecules, so they don't carry enough information to know the difference between life and non-life. Methane on Mars could be produced by geological processes. Oxygen can be produced by breaking water down with UV light. The answer to inferring biology from gases is going to be assembly theory. You can measure the assembly index in the lab using three techniques: mass spectrometry, NMR (nuclear magnetic resonance), and infrared spectroscopy.

Assembly Theory and Life in the Universe

Host: Were aliens something you thought about a lot as a kid?

Lee Cronin: Not obsessively. I was probably thinking about why am I here, why does life exist, how can I take this apart? But understanding what life is as a chemical phenomenon is critical for understanding the phenomena of life on Earth, which will tell us about aliens. If we only saw our Sun in a black sky, we would obsess about how the Sun got created. We know the Sun is created by the gravitational collapse of hydrogen, getting hot enough to overcome the strong nuclear force and undergo fusion. We can see stars everywhere in the universe. But we only have an N of 1 for life on Earth. If we can understand how likely it is chemically that life emerged on Earth, we can start to bound the probabilities.

Host: How can we do that if we don't understand how big the universe is?

Lee Cronin: We can estimate the size of the Milky Way, and we have bounds on the light cone of the universe. Every star looks like it has planets around it. Those planets are either: dead and never will have life; abiotic but have the possibility for life; alive; living and technological; or post-technological. Life is probably only possible within a certain zone of temperature: too hot and the chemical bonds fall apart; too cold and nothing happens. Listening to David Kipping's Cool Worlds podcast, they were talking about the habitable zone, and I realized planets probably can respond. Planet Earth, after the Late Heavy Bombardment, started to cool down, and feedback processes regulated its temperature, like James Lovelock's Gaia hypothesis. If life starts to form on a planet and the planet drifts outside that zone, evolution will respond to counter that, like a thermostat. The planet is able to self-regulate for the emergence of life.

Host: If a planet was traveling around its star such that it was getting closer and heating up, are you suggesting the planet could adjust?

Lee Cronin: Pre-intelligence, no; post-intelligence, maybe. That brings up one of my favorite Chinese sci-fi movies, The Wandering Earth. If a planet is moving around a star such that it gets closer and hotter, it might regulate the atmospheric pressure: when it gets too hot, make the atmosphere reflective—shields up; when moving further away, trap light in—shields down. Climate change is annoying for humans because we urbanized around the coasts and equator where the zone was right for agriculture, but Siberia is going to become really habitable. On a timescale of a few hundred years, buy real estate in Siberia. We will probably come up with a technological solution for climate change. Maybe Earth will survive longer than it otherwise would have because it produced life.

Chemify Genesis and the Job Replacement Fear

Lee Cronin: Over the next 5 billion years, the Sun is going to get larger as its hydrogen gets depleted and it starts fusing helium, and it will get hotter. We have got a couple billion years' notice; can't we just push the Earth gently that way? Just fire a few nukes: boom, shift it a few millimeters a year. Given 2 billion years, we can do a lot.

Host: I have got to get you and David Kipping in here for a podcast together. He is coming back in November.

Lee Cronin: That would be really cool to exchange ideas. I listened to his podcast and I was like, "Nah, no."

Host: You are a "no" guy. The minute you hear something questionable, you are completely unafraid to challenge it.

Lee Cronin: Real-time critical thinking requires you to be uninhibited by people pushing back, willing to be wrong, and open to education. We are mystifying AI because we have been told it is magic. OpenAI hit upon a cognitive treasure trove when they trained their model, and suddenly these chatbots started to be really good at predicting the next token. As you increased the context window, it got even better. When you built guardrails and chain-of-thought abilities, these systems were able to daisy-chain together. But we have got to demystify it and define what intelligence, creativity, and novelty actually are.

My company, Chemify, has got a few hundred people in it now. I was at a meeting where people were saying, "I have been able to shed all this number of people from my workforce." I was like, "Guys, what are you doing?" Human beings are fantastically flexible, infinite problem solvers, and infinite creativity machines. You want to employ more of them, not less of them. I find the whole job displacement fear rather distasteful. Chemify's aim is to build a world model for chemistry. If you go to chemify.io/genesis, it explains the world model.

A Virtual Library of Molecules

Host: "The molecule on the screen is only a hypothesis. Ask, Make, Test."

Lee Cronin: People think that drug discovery is discovering a molecule. In the old days, you would look at plants, find molecules that have biological activity, and discover the drug. Chemical space for possible drug molecules is typically said to be about 10 to the power of 60. The number of atoms in the universe is 10 to the power of 80. I redid that calculation using assembly theory: it is not 10 to the 60, it is 10 to the power of 117. We published this on arXiv: "Elucidating the Size of Chemical Space with Assembly Theory." If you allow your molecule to have an assembly index of 25 steps with the chemistry that is available, there are 10 to the power of 117 molecules accessible. When your space is so big you cannot search it, you have to generate something novel. That is where Genesis comes from.

Genesis is not a catalog, not a reaction search; it is a computational executable loop wired to real chemical capacity, turning ideas into verified matter. All these AI companies doing drug discovery are just generating random graphs on a screen; they can't make them. What Chemify has been doing is making molecules in the physical world with chemical robots, and Genesis connects the AI models directly to our robotic synthesis engine. People buy Genesis, keep all their intellectual property, and we federate the operational data. Chemify gets faster at doing chemistry, but customer designs remain completely encrypted and segregated, similar to how the UK Biobank operates. We built a chemputer farm in Glasgow. Every time we do a reaction, our world model gets better. Chemify's virtual library of digitally verifiable molecules is 10 to the power of 40, drawn from the total space of 10 to the 117.

Yes, I Want to Play God

Host: When you look at the scaling of tools like AI and the abilities humans can leverage at an exponential rate, do you ever worry about our incoming ability to start playing God?

Lee Cronin: No, I want to play God. In fact, I love playing God. We have mobile phones, we have drugs, we have cars. I want flying cars. As Arthur C. Clarke said, any sufficiently advanced technology is indistinguishable from magic. Imagine going back in time with a working smartphone and FaceTiming someone; people would think you are a god. We have to make sure humans are flourishing and thriving. By any measure, human beings have never had a better time on Planet Earth. We burn fossil fuels to make ammonia to feed the world. If we attempt geoengineering to solve climate change and 50 people die of something unforeseen while saving billions of lives, do the calculation.

Host: What about genetic engineering, like Colossal Biosciences recreating extinct species like the dire wolf or woolly mammoth?

Lee Cronin: George Church is presumably behind all this stuff. George is very mischievous. The same way I like to experiment with chemistry, he likes to experiment with biology. He is testing the limits of what is ethically and commercially acceptable. Is low IQ a curable disorder? If you could genetically engineer higher IQ in humans, would you do it? Society will have to decide what is acceptable, just as cesarean sections became accepted. In the UK, we approved mitochondrial donation therapy—so-called three-parent babies—to prevent severe mitochondrial disease, and healthy children have been born who would have otherwise died of muscular dystrophy.

Twelve-Foot Giants and Genetic Destiny

Host: That is hard to argue with.

Lee Cronin: If you supercharge mitochondria, could that give rise to super-athletes? In Western Europe over the last few hundred years, due to malnourishment, people did not achieve their genetic destiny; now in the Netherlands, people are six foot whatever tall. Am I worried people could use Chemify's technology to mass-manufacture bad stuff? We build encryption and GPS-geolocated licensing into the systems. There is extensive firewalling, real-time telemetry, and fail-safe designs built into our chemputers.

A hundred years ago, "computers" were people in skyscrapers using slide rules, performing several hundred thousand operations per second globally. Today, the number of chemists doing reactions every day is between 100,000 and 1,000,000. Chemify's physical AI will drastically increase chemical operations. With falling birth rates, we need to keep people healthier for longer, driving the economy through subscription healthcare and advanced therapeutics.

Lee vs. James Tour on Piers Morgan

Host: As long as Dr. James Tour isn't right and you're wrong. I saw you guys going at it on Piers Morgan.

Lee Cronin: James Tour is a chemist first and a scientist second, whereas I am a scientist first and a chemist second. James does chemistry and uses that authority to exert authority over his religious belief system. He claims everyone is clueless about the origin of life. Why would I do experiments if I knew the answer already? I am the only person willing to debate him publicly. Piers Morgan has a huge following, and it was a fun debate to have. One of my sons watched it and told me, "This is the first time I have heard you talk about something and I actually understood it." If it inspired him, it was worth doing.

Lee's Work

Host: It is always great to have an open dialogue in science. Thank you as always, Lee. We are going to have to do this again. Let's go get some steak now.

Lee Cronin: Thanks, great to be with you, and yeah, till next time.

Host: All right everybody, give it a thought. Peace. Thanks so much for watching. Please hit that subscribe button, leave a like, join the Patreon, and join the Discord clipping community below. See you for the next episode.