“The Evolution of Biological Information” with Professor Christoph Adami

Hello there, welcome to this episode of Bridging the Gaps.

I'm your host, Dr. Vaseem Akhtar. Today I'm joined by Christophe Adami, Professor of Microbiology and Molecular Genetics, as well as Physics and Astronomy at Michigan State University. He is a leading expert in applying information theory methods to the study of evolution. He has received several awards including the NASA Exceptional Achievement Medal and the Lifetime Achievement Award.

from the International Society for Artificial Life. Today we are going to discuss his new book, The Evolution of Biological Information, How Evolution Creates Complexity. from viruses to brains. Chris, thank you very much for joining me and welcome to Bridging the Gaps. Good to be here with you, Haseem. Chris before we discuss evolution and biological information, let us set the scene for our listeners.

Information can be described as an abstract concept, something which has the power to inform. Now, Shannon's information theory describes and explains information using mathematics. And you say that information is a precise concept that can be defined mathematically. Help us to unpack and understand this statement. Sure, I'd be happy to. So, you know, in our daily lives, we use the word information all the time.

Right. We make informed decisions or we need some information to make decisions. And if you think about it, information is always somehow involved with decision making. Right. Instead of making an informed decision, I could just simply flip a coin. But you can ask yourself, well, what has a better choice, a chance of being successful, flipping a coin or making an informed decision?

So this idea of information is that which allows you to make predictions with accuracy better than chance. That is the central concept of Shannon's theory. So it turns out that the concept of Shannon... will allow you to mathematically quantify how much better your predictions are being with x amount of bits that you are in possession with let me give you an example suppose you go to a city that you don't know about

right um and you're sitting in your hotel room and goes like you know i would like to have a coffee right so what you could do is you could go downstairs and just randomly walk around you know in in the city and hoping to find a starbucks because let's say you know that there's going to be a starbucks somewhere right and you're going to possibly find one by chance but suppose somebody actually had handed you a map

with all the Starbucks locations. What that means is that now instead of randomly moving around, you basically pick a Starbucks that's closest to your hotel and you can go right to it.

which means you've made a prediction and that prediction is very valuable to you right because you don't waste time going around you could go immediately to some place so you may you you had information in the form of this map you used it to make a prediction and then you exercise this make a decision that allows you to in fact reap a reward which means getting the coffee without having randomly walked around and the difference is in a sense in this case the time spent looking for the coffee

right and you can apply this to all kinds of other things but information is something that is valuable because it allows you to make predictions with accuracy rather than chance what the value is depends on the situation but always there's a value associated to having information as opposed to not having the information. And there should be a context as well, because I don't think that we can say that there is absolute information.

information would always depend on some context well that's in a sense a very central aspect of information that is contextual so for example if you are in chicago in your hotel room and somebody handed you a map of the Starbucks in Los Angeles, is that good for you to make predictions? The answer is no. In other words, that map is only information.

about starbucks when you are in fact in los angeles but if you're in chicago this is not information for you in fact it is disinformation to some extent it's your you moves our decisions are not going to be better than random chance so in other words the same map

It's going to be information when you're in one city, but it's not going to be information when you're in another city. And that is a central aspect of information, that information is always about something. Now, in this case, it's about the Starbucks in Chicago. It's not about the Starbucks in Los Angeles and vice versa, right? So anytime somebody tells you, I have information, you should in your head ask yourself, what is the information about?

Is this person able to tell me what the information is about? Because if they can't tell me, then in fact it is not going to be information. It is not going to be useful. So people sometimes talk about useless information. There's no such thing. Because if you can't use it, it's actually not information. This description of information makes sense.

I twice read your paper what information is. So the description of information makes sense. But the idea of entropy seems a bit challenging. I'm just keen to check with you that your description of entropy in the context of information is exactly the same as Shannon outlines entropy when he presented the theory of information. I was just...

Going back and forth your paper and Shannon's description. So if you could help us to understand that and unpack that. I'll be happy to. So indeed, Shannon's description of entropy is exactly the same as my description of entropy. Entropy, you can think of it as potential information. So for example,

All the maps about the Starbucks anywhere in any city of the United States, for example, they have potential information, right? But depending on whether you are in fact in any of these cities, you can't use it. right so any data could potentially be information but if you don't know how to use it for you it's essentially just variations okay so entropy is uncertainty it's stuff that could have an application

But in fact, you are not sure what it could be used for. So it is also sometimes called uncertainty. Uncertainty is basically just you have a quantity that you're measuring even in bits of. how much variation is there, right? That could potentially be applied for making predictions. But entropy is, in a sense, non-contextual. It is just variation without any knowledge about what it could possibly be used for. But once you have a use for it, it turns entropy into information.

think an interesting point that we should acknowledge here is that Shannon's theory not only quantifies the concept of entropy and gives us a mathematical description of information it also characterizes information transmission channels it discusses transmission capacity and how information can be protected from noise and presents the idea of error correction also. I just want to make this point because when we will get to our next topics and evolution so all these points

you are using when you are describing information in the context of evolution down the line. Am I right in saying that? No, absolutely. So in fact, the vast... majority of Shannon's contributions in his seminal papers was about the transmission of information. In fact, it really was about the transmission of entropy.

Because he was not concerned about the meaning, about what... you know like his words or his symbols that he was transmitting over a challenge what they could actually represent or what you could predict with them to some extent he didn't care he said like look i just want to know how well you can transmit symbols over a noisy

channel and he was able to prove mathematically something that people had not suspected, namely that in a channel with any amount of noise, except in a sense like a completely randomizing noise, you can actually send information about this channel with accuracy that is essentially perfect.

People had thought that that's impossible. They had figured, for example, out that if there is a large amount of noise, then in fact, at some point, the rate of perfect information transmission is going to go to zero. And he showed no. actually not the case and he showed that error correction was in fact the key to being able to do perfectly accurate information transmission over noisy channels. That was the main contribution of Shannon.

he constantly not just in the paper but also in interviews later said well like it doesn't matter what the meaning of information is and that's true but you know when you when you look back it turns out that Shannon didn't really fully understand what meaningful information is. In his papers, he seemed to sometimes call entropy and information by the same name.

Once you go back to some of the interviews, he really was not clear about this to himself. He didn't fully understand his own theory, believe it or not, because in a sense, he didn't care. What am I going to do with the information? spoke about you know how in the brain we have information transmission but he was sort of vague about it because for him it was much more important to understand how can i send any

symbol across a channel without noise, never mind what meaning is attached to these symbols. And that is the reason that some people find it difficult to understand entropy as described by Shannon in some of his research papers and talks and books and then entropy that is described by the second law of thermodynamics and now that you are bringing this idea and you are promoting this idea of using

information theory to understand biological information and its evolution. So I think that maybe you have come across some other places. while engaging with your students or PhD researchers that this is the challenge that sometimes you try to look at entropy. through the lens of shannon and you are trying to make sense of that so so there is a bit of lack of clarity there or something that is confusing right so so in a sense like you pointed out shannon is partly to blame

Because he did not make that difference clear. He did not make the contextuality, for example, of information clear because he was really only interested in sending entropy, meaning potential. information over a channel never mind what you're going to do with that information right the article that i wrote called what is information was

was written precisely to address the fact that entropy and information are two very, very different things. The French physicist Louis Briouin, for example, had talked about negative entropy or negentropy all the time. This is a nonsensical concept because what he realized is that, well, there was something like minus entropy, but...

Information is actually maximum entropy minus actual entropy, which is then a positive quantity. And so Briouin had to simply... forgotten about the first term in that equation that makes it information and so then he came up with this idea of negative entropy which is

complete nonsense. What he really means is information without the maximum entropy term that makes this positive. So it was important for me to clarify this difference between entropy information because I have seen so many papers and even seasoned researchers that are just not clear about the distinction. i figured like rather than me having to correct you know the the you know people all the time i could just point them to that papers like just read that because that would make the

the distinction between entropy and information more clear. And for the listeners, the distinction is very simple. Entropy is not something they can use for making predictions. It just tells you how much you don't know. Information is... The maximum amount that you don't know minus the actual lack of knowledge and the difference between two is how much you know. Chris, let us gradually...

Proceed to the main idea that you present in this book. You use the concepts developed by Shannon in this information theory and then you start applying them on the evolution. of biological information, biological organisms. So let us start with this thing that information content of genes and proteins, you discuss this in the book.

And you say that living organisms know some very important facts about the world. They live and survive and thrive. And you say that when you use the term know, you strictly say that This is in the sense that these organisms have information. Talk to us about the emergence of the concepts of genetic information.

Right. So the important thing there to keep in mind is that when biological organisms need to succeed in a particular environment, it would be very good if their decisions wouldn't be random. And so all organisms make decisions all the time. A cell, a single cell is constantly making decisions about, you know, whether to grow or not, whether to activate certain genes or not activate them. Whereas, you know, microbes make decisions.

about whether they should be swimming in one direction or another. Obviously, higher order organisms constantly have to make decisions. Let's say, squirrels have to make decisions on whether to actually... save certain nuts for the winter or things like that. So constantly every form of life needs to make decisions that further in a sense their well-being. Make sure that in a sense life is going to continue so that you're not going to be eaten.

How do you make those decisions? And of course, mathematically speaking, the only way you can make decisions with accuracy rather than chance is with information. So where does this information come from? Well, of course, evolution has transformed transferred this information in a sense from the environment into the genomes, into the proteins that they're using, the transcription factors and so on. So that means that the decisions that cells are making or

microbes are making, they're not random. They are informed decisions. So there's information inside of the protein that tells them about the world in which they live. For example, it tells them about the fact that there is water in the world, that there are certain chemicals in the world. For example, making one decision to activate a particular gene would make no sense if, in fact, the sugar that you're now trying to harvest isn't even there. So there are sensors.

that the cell has that sense for example hey look there's maltose and if there's maltose that's a good sugar for me so i should activate the genes that actually break down maltose and then shut down all the genes that activate glucose, for example. This information, that's in the genes and in the proteins that are involved in the pathways that make this decision. And...

One of the things that people have always talked about is like, yeah, somehow there's information in the genes. People were intuitively aware of that, but they didn't know how to calculate it. And 25 years ago, I wrote this paper, which shows you how you can actually do that.

that basically the information inside of a particular molecular sequence can be calculated as the length of that sequence, which is essentially the concept of the maximum amount of information you could possibly... store in that sequence minus the remaining randomness that is in that sequence like for example there could be three or four nucleotides you could

change them it makes no difference that's the remaining randomness right and the difference between the two is the amount of information that's stored in that sequence and we can actually measure this without having to in a sense do very complicated things. We can do this for example with an alignment of existing sequences.

And then we can, using something called the information decomposition theorem, we can sort of make an approximation about the true information content, which is relatively accurate.

it's more accurate if we have more example sequences but basically it tells us not only how much information is in the sequence but even where it is and the general you know idea behind it is if there is a nucleotide or an amino acid or you know any type of monomer that if you are changing it and it kills the organism carrying it, well then obviously this combination of bits was really important and it carried a lot of information.

But those areas where you make changes, mutations, and they do not affect the fitness of the organism at all, then they are not information. What that means is that in evolutionary biology, there is a direct link. between information and fitness. Why? Well, the information is being used for you to survive better. So someone who has more information about the world in which it lives, that it will be fitter than

some organism that does not have the information. It makes better predictions and the better predictions give you an evolutionary in this case actually fitness advantage you are your lineage is going to have more offspring per unit time why because you are better to predict not only the environment that you're in right now but also the future environment

Because in our genes, we even have information about how an environment might change. Actually, the majority of the information about how an environment might change is actually in our brains. Our brains are storing information. about the world in which we live, but it's actually able to store information that we can update during our lifetime.

Whereas, of course, our genes, we cannot update them over our lifetime. This is a very, very slow process in which the information goes from the environment. via the evolutionary process into our genes. But in terms of brains, we can learn something very quickly. We're touching a hot stove and it goes like, you know what?

I'd better not do that again. And that means in that instance, some information was stored in your brain that will help you survive in the future. For a few more moments, I just want to continue.

and explore the concept of genetic information a bit further. So genetic information is stored in DNA molecules, in our genomes. What is the nature of this information? How much do we understand? the information that is contained in a genome, are there still areas that are untranslated, undescribed, untranscribed? How much do we understand a genome?

So, this is a good question, and I don't think I can answer that with 100% certainty, but I can tell you that we know how much potential information is.

within our genome that's essentially the three billion base pairs per chromosome obviously because we have two sets there is a little bit more than these three billion uh but of course by the way three billion base pairs would translate to six billion bits because each nucleotide represents two bits because it's base four not base two so the potential information

disregarding the fact that we have two sets for chromosomes for a moment, is about six billion bits. Now, how much information is actually in there? This will depend on how do we understand what is the fraction of the genome that actually contains information. So there are projects that have attempted to do that. The amount of coding...

In other words, the number of base pairs that are actually involved in coding genes is just a few percent. However, there is of course a vast amount of genetic material that is used for regulation. of genes and that is not being translated. But it used for regulation. So the ENCODE project, for example, had estimated that about 8% of the genome is actually functional information. And if you take that seriously, then that means that there's about a half a billion bits.

That is information that we as a human have about how to live in our environment. Now, that is not a number. That would put a lot of accuracy in because that 8%, of course, is also fairly uncertain. For example, there's a bunch of different. So first of all, there are alternative, you know, splice forms that are clearly information. So you should look at the transcript.

in order to understand how much information is. And of course, the ENCODE project has done that. Now, there's other things that could be informative. There could be post-translational modifications. However, you have to keep in mind that the post-translational modifications are executed by actual kinases which are coded for

in open reading frames. So it is not clear to me whether any post-translational modification should be thought of as extra information. Having said this, there is more information inside of most organisms because there is still the microbiome. We have a gut microbiome, we have an oral microbiome, and all of these work in conjunction.

with our other genes in order to affect our fitness. So there is possibly, and in many cases, demonstrably information that is important to our survival, that is not stored within our genes or even our brain. but in fact in the microbiome and the microbiome interacts with all those genes. So it's actually very difficult to calculate how much extra information is in these, in a sense, dark areas that we usually don't look at.

But having said all this, I have not yet found a case where I would say, hey, there is information in, let's say, a patient or a microorganism, and I have no idea where it resides. I'm very confident saying that if there is information somewhere, we're going to be able to find it. Because keep in mind that information has to have a physical substrate. It's not an idea. It's something that is encoded in a physical state.

Because information always has to have a physical basis. It's not metaphysical. It's physical. And so it would be a fantastic surprise if we would find… degrees of freedom within a biological organism that are carrying information and that we have not thought of as carriers of information before. I'm not ruling it out because we are very complex bodies.

But right now, I should say that we are not aware of any such places. Earlier in the discussion, we looked at the information theory. Then we discussed genetic information, the approach that you outline. in this book, which is to apply information theory methods to study genetic information and the transmission of genetic information. How did this all start and which direction this research is going?

Well, I mean, it started out 25 years ago when I first learned about information theory. And, you know, the sort of quirky nature of it is how did I learn? classical information theory? Well, I was actually in the lab at Caltech, in the lab of Steve Kuhnen, which is a nuclear physics lab, but he was getting interested in quantum computation.

And so myself and a colleague of mine, Nicolas Cerf, who shared an office with me, we were looking at some... interesting facts of quantum information which involves in fact quantum teleportation. And we try to understand how this is possible. And we realize that, hey, that's because, in fact, quantum entropy can be negative. And in order to really understand quantum entropy, I had to actually learn.

classical theory of information, Shannon's theory, because that is the foundation of quantum information theory. And as I was learning those, I somehow realized that I can actually use that in order to understand biological information. So I read a paper by Charles Bennett who had...

sort of attempted to do something like that. He talked about something called fuel value in one of his papers. And I looked at this and looked at this and then suddenly realized how that was related to Shannon information. So it was that. from going from quantum information, having to learn the foundations of classical information, at the same time having actually studied evolution in a digital life system that we were developing at the time.

where we would see in fact self-replicating organisms inside of the computer become more and more complex and i realized what they are doing is they are accumulating information about the environment so i figured well you have to be able to quantify that and sat down with now armed with the classical theory of information and i was able to do that and then wrote that paper and that was really the very beginning was like the moment i had learned how to use classical information theory

independently of all my other work in quantum information, I realized how powerful this tool was. I realized that nothing in biology makes sense except in the light of information because everything that you see in biological organisms has to do with how information is stored how information is transmitted and how information is being used in order to make predictions so that that furthers the survival not just of the individual but the lineage that this individual is part of.

So everything in biology is involving information. In fact, one of the things that is so astonishing about biological systems is that they are far from equilibrium. In physics, in statistical physics, we talk about this concept of statistical equilibrium. And statistical equilibrium just means that

Nothing changes anymore. So all the things that could potentially change as we approach the equilibrium state have already happened. So it's maximum entropy state. And the maximum entropy state is really a zero information state. So the physical, non-biological world is a zero information state. Whereas any form of life, of course, is...

packed with information in the genome which they're using in order to further their lineage. So it occurred to me that the difference between being Alive and not being alive is literally the amount of information that is reducing this maximum entropy to the actual entropy that you have which is much much smaller than the maximum entropy and in fact how much smaller by the amount of information that you have about the environment

This nicely brings us to my next question which is about the key topic, the core topic that you discuss in the book and that is evolution. In the book you discuss principal mechanisms or should I say principal laws of evolution, inheritance, variations and selection as measurement.

processing and transmission of information not in this sequence though so talk to us about that these three principles of evolution how they are supported by this new idea of transmission and measurement and processing of information sure so um When Darwin wrote his seminal book, which he, by the way, thought was just an abstract of his theory, and he wrote another 10 books later, sort of...

you know applying it to different kinds of organs but the origin of species he makes clear uh what are the central pillars of of an evolutionary process and there's three of them So one of them is simply the concept of inheritance, the idea that if you have a particular phenotype, and then you give rise to an offspring, that this offspring will be very similar.

He was not, of course, aware of the idea of genes, but he knew that there was something that is controlling what an organism looks like, how an organism behaves. He knew that there was a substance behind that. basically realized, okay, what's going on when parents create offspring is that they are inheriting these characteristics. Not the characteristics itself, but...

In a sense, we would now today say the genes that give rise to the characteristics. But, you know, Darwin was well aware that you're not inheriting just the characteristics. But this idea of inheritance. That was one of the three pillars. The second pillar was the idea of variance, namely that in the process of creating offspring, the offspring don't look exactly like their mothers. There is changes in them, which he called the variation. Again, he was not, you know, you couldn't.

conceive of the concept of a mutation like we are today talking about a genetic mutation but he was well aware that in that process this substance that gave rise to this you know, to the phenotype, to the appearance in a sense, that that was being changed a little bit, not dramatically, so that, you know, you could not tell the difference between an offspring and a parent, but slightly.

And that it was this difference that is being used in order to actually determine whether an offspring has capacities that their parent didn't have. Now, what are these capacities? Again, the first concept was inheritance. The second concept was variation. And the third concept is selection. Namely, he said, look, these phenotypes, these variations that you have they're being interpreted by the environment in other words they are either useful or they're not useful

So variation creates all kinds of things, some of them useful, some of them not useful, but the environment is going to select which ones are useful. So if you have evolved a feature that helps you. or you have a variation that helps you, that variation will be transmitted to your offspring. Whereas the variations that are not useful or that actually could be deleterious, they are not going to be transmitted.

most of the time to your offspring. Certainly those that are lethal are not going to be transmitted. So this idea of selection he realized was an important concept. Now, in hindsight, we realize that these three laws are really informational laws. The idea of inheritance is really just the replication of information, the copying of information.

a string, let's say the DNA of an organism, and when you're making an offering, you're copying the information. That's what inheritance really means. It's the copying of information. The variation is just the mutation of information or the changing of information with random changes. Now one of the things that people of course later realize is that in physical systems it is essentially impossible to do perfect.

Because of the fact that we have a noisy world, there will always be at least a tiny amount of changes that are going to be made when you are attempting to do a copy. But generally speaking, you can think of what Darwin said was the concept of variation or mutation turns out to be just changing information. And now the third element, which Darwin calls selection.

That, of course, is the meaning of information. It means like whatever variation you have, is this useful to make predictions, right? So the meaning of information is just simply the value. that you're getting from having the capacity to make a prediction better than, let's say, some other organism, right? So now we realize, yes, all of evolution.

really can be thought of completely in terms of how you're storing information, how it is being replicated, how it is being changed and how it is in fact used in order to make predictions in the world. that will actually be translated in your lineage having more offspring than a competing one. And if we look at these three major principles of evolution, inheritance, variation and selection.

If I just for a moment focus on variation, variation can only happen if the organism is interacting with the environment and information is being changed that what variations may make the organism a better survivor in that environment. So it's copying the information and then replicating the information and making sure that the offspring

Gets the information is one thing, but the variation can only happen when the organism is engaging with the environment. So how does that aspect of information exchange work in your view? Well, so it is true that of course the variation that is being introduced must be in the information that is being replicated.

So for example, we have, of course, information in somatic cells. These are the cells that are not contributing when we are creating offspring, right? So if I have a change in a somatic cell... which by the way happens in cancer all the time that is not information that is being inherited it might be something that is important for the organism but it is not something that is being transmitted in time you know in the lineage so we have to

understand that the only information that is inheritable is that information that is in fact inside of the germ cells right but in the germ cells these germ cells are being replicated and these changes occur precisely during the replication process which is imperfect on top of that of course not only is it an imperfect process but of course there's recombination going on in

you know at least in sexual organisms and that recombination can bring mutations together which actually can create information just when there wasn't information before. And that's a very important concept. I don't know if we really have the time to get into that, but think of it this way. Suppose one parent has a particular variation, which by itself is not meaningful.

and another parent has variation that by itself is not meaningful it is possible and it happens actually fairly commonly that if you bring those two unmeaningful variations together in one genome that these two together suddenly not suddenly but at the moment they are being together acquire information about the environment even though each of the components was not it's a very important part of understanding the concept of information that information is in a sense not

just contained in the value of the monomers, but also in the relative state of monomers, so that information can be stored in pairs, it can be stored in triples or quadruples, quintuplets and so on, and that it might not be apparent. within any of the single parts. So people sometimes talk about the whole is more than the parts. In this case, each part could have zero information, but you bring them together.

And then suddenly you realize that there's an enormous amount of predictive aspects of it. In other words, parts that have zero information by themselves can, when put together, have an enormous amount of information. Which is why I'm saying that mutations during the copy process, which are essentially a faulty copy process, are not the only ways in which new information can be generated.

in a process of creating an offspring. It is also possible to create non-informative things that are being joined in one chromosome that creates information that way.

But generally speaking, we understand how information can be created. And it is, of course, it has to be during the process of making copies, which in the case of sexual organisms involves both mutations crossovers insertions deletions all of the things that actually change the genome and there are many many many of these processes i just mentioned you know the most important ones but there are others all of these potentially can uh create

information within the genome of course selection will decide what is information what is not because those that actually increase your fitness that make you more powerful those are information those changes that are either neutral or even deleterious to to you they are not information in other words what is information is determined by how well you are doing in the world and again it's of course contextual because you might be doing well in one world but the same genome

The same combination of nucleotides or amino acids could be terrible in another world. Now, to describe evolution using this information theory,

You borrow the concept of Maxwell's demon, and you discuss that in chapter three. So this idea that you borrow Maxwell's demon, that how evolution works when we are using information theory. Is this a general description or does this only apply to when we talk about the possible origin of life using this information theory? So perhaps We should explain to our listeners, Maxwell Demon, and the way you are describing that how this natural...

Maxwell daemon type process might be responsible for kick-starting life or kick-starting Darwin's evolution? Right, so... I don't think that the concept of this natural Maxwell demon, or sometimes call him Darwin's demon, is... is helpful for us to understand the origin of life, but it certainly is helpful in understanding the evolution of life. So let me give you a little bit of background. So Maxwell,

Of course, it's well known to physicists as the person who understood gas theory and thermodynamics and things like that. And he was interested in trying to understand the second law of thermodynamics. It basically tells us that a system that is not in an equilibrium state will move towards an equilibrium state almost all the time. And when I say almost all the time, it basically is like there are

very low probability cases where this cannot happen, but we can essentially disregard that. The second law of thermodynamic sense, if you are at a low entropy state, you're going to move towards a high entropy state until you are in equilibrium and then you are in the maximum entropy state. And then Maxwell came up with a counterexample. And he said, but hold on. What if, in fact, I can make measurements?

Now, what are measurements? Measurements are things where I have an object and I would like to ascertain its state. And the way we normally do this when we're doing a measurement is like we have a measurement device and then we have the object. In a sense, we would like to copy the state of the thing I want to measure onto my measurement device, which is well understood. And once you realize that.

you realize that hey you're taking something that you don't know anything about and you're making it copying its state so that you now have two of the same but if you're starting with two that are not the same and you're coming up with two that are the same well mathematically speaking you have lowered the entropy oh yes lowering of entropy means of course that you have to

extract information then of course that's what you're doing in a measurement right you're extracting information that's what a measurement means so a measurement is going to reduce entropy and then Maxwell thought well isn't that a violation of the second law of thermodynamics. The answer to that is no, because the act of a measurement is in fact not an equilibrium process.

it is in a sense the opposite a measurement reduces entropy and as a consequence it is working against uh against equilibration so maxwell created this hypothetical situation where you have two chambers that are filled with gas molecules. And between those chambers, there is a door that could be operated by this demon. And basically he said, well, the second law of thermodynamics says that I won't be able to create a differential between two of these halves.

But he realized, hey, but if in fact if I can measure the speed of the molecules and when I see a high speed molecule approach the door and I'm going to open this door and if I see a low speed molecule approach the door and I'm going to keep it closed.

And if I do this long enough, then I will have high-speed molecules on one side and low-speed molecules on the other side. So I've created this differential. It was like, oh, am I not violating the second law of thermodynamics? The answer… later was given that no you're not doing that because you can actually calculate the mathematics and this was done by the German-American physicist Rudolf Landauer.

Interesting development, a very, very readable paper of his. But the upshot is, yes, a measurement process can actually create this differential, and it creates information, which, of course, we now understand is what...

is the difference between a living system and a non-living system. And now once you understood this, you're going like, well, then in which sense is evolution doing the same thing? Namely, creating an accumulation of of uh information um that puts us further and further away from equilibrium which clearly that's what we want and the answer is yes every mutation is

Like one of these attempted measurements, right? And now, if the mutation happens, the question is, shall we keep it or shall we reject it? Just like, you know, Maxwell's demon is making a measurement of the speed of the molecule and asking whether I should... open or close the door between the two. And it is, of course, the survival of the organism within this environment that is giving you the result of the measurement, namely those mutations that are useful for us are being maintained.

We keep reopening the door and those that are not useful for us, they are being rejected in the sense that the organism dies and therefore that mutation will never show up in any offspring. So as a consequence, we can now map this process of information that allows you to decrease entropy and increase information on your device that records all of these speeds, that's mathematically equivalent to what happens.

in the Darwinian process of evolution that leads to an accumulation of the information inside of the genome in such a way that we actually now have fitness and can therefore better survive. that this analogy gives us new insights. In fact, it's just a re-description. of the actual Darwinian process. But it is actually very useful to understand that information has to accumulate and the information is accumulated by measurements.

So you can select, well, nobody's doing any measurements. No, it's not. But the mutations are happening and we're testing.

the organism is testing whether they're useful or not and those that are useful are being retained and those that are not useful are not being retained and that is mathematically equivalent to a measurement and so what really happens is that the information that's in the outside in the environment via these measurements is seeping into the genomes so that we can use it to further the lineage that this genome is part of.

Very good description. Thank you very much for these details. Now, millions and perhaps billions of years of evolution have shaped this biological ecosystem that we see around us. And this is very complex.

So as the evolution continues and the information is collated and is copied and is further enhanced then the complexity increases. This is a very general statement that I'm making here but you talk about this in detail in the book that this complexity that we see around us so how does this approach of applying information theory help us to describe measure, understand this complexity.

Well, so that's a good question, and it is in fact one of the questions that I tried to answer 25 years ago in this paper that I wrote, which is actually called Physical Complexity of Symbolic Sequences. So people have worried about this concept of complexity. They've used the word all over the place and people have worried, what do I mean by that word?

So people have tried to come up with mathematical ways of defining complexity or ways in which complexity could be measured. But there was always going to be somebody who could say, This concept of complexity that you defined, for example, structural complexity or sequence complexity or developmental complexity, it works well in this area, but it might not work well in another area.

And so I was wondering, is there a way to define complexity in such a way that is really a universal measure of complexity, which would allow us to compare complexities? not just between organisms, but maybe in engineering, right? So that we have a better way of actually understanding complexity in a quantitative way. And it occurred to me.

that as I was studying concepts like Kolmogorov complexity, which is a very famous construction by a Russian mathematician, then i realized that one can in fact define a complexity of a sequence from an automata point of view but the key insight was that if i average

this complexity of a single sequence over a group of sequences, I will get Shannon information out of it. And once I realized that, I realized, hey, that means that Shannon information is in fact how I should be measuring complexity. what that means is that if you have more information then you are going to be also more complex and you could say well how is that possible and the answer to that is that information that is useful to make predictions

is nothing without, in a sense, a body to actually implement these decisions, right? So in other words, in order to actually... exploit information, you have to have structure that allows you to do that because you have to interact physically with that world. So in other words, just information alone is not going to be helpful because I can have

like a big dictionary. But if I can't actually use this information to interact with the world, then it's completely useless. So it turns out, of course, that the majority of the information in our genome is used to build the structure. us, right? And then, of course, the stuff in the brain allows us to also act on the environment and so on. But it turns out that all of this genomic complexity is being used to create functional complexity.

And the functional complexity is again, reflected in structural complexity because how can you be functionally complex without actually having any structural complexity again because you have to interact in complicated ways so it turns out that the informational

measure of Shannon information basically gives rise to functional complexity and structural complexity and then in order to actually measure how well you're doing, how complex you are, all you have to do is look at the Shannon information. And now we actually have a mathematical measure that allows us to actually tell the difference in complexity between the organisms.

And now people have actually tried to say, well, can I now measure the complexity of bacteria and compare them to other organisms and see how they differ in complexity? The answer is yes, actually, you can do that. This was done by Eugene Koonin, for example, in a paper where he used my measure to actually estimate the amount of information that each of the different classes of organism has and you find out that they clearly are aligned.

In fact, it turns out that according to that measure, humans are in fact the most complex organism within the ecosystem. And there was a long discussion for decades where people said, oh, you might say that humans are more complex, but what do you know? For example, some would say, but bacteria are far more successful. They have far more of them. It's like, yeah, but in fact, that doesn't mean that they're more complex.

Having more of them is not actually a good measure of how successful you are, for example, in predicting your environment. Bacteria do make predictions about their environment, but they were very short timescales, like the next minute. not next year or

you know, saying that, you know, within 50 years, the climate will be so hot that we're going to have very, very, you know, big problems. That is a prediction that we are able to use using mathematics and our brain that a bacterium simply could not do. Right. So the complexity of what we are is reflected in how well we are going to be able to predict not just the environment, how it is today, but also how the environment is in the future.

Being able to predict a future environment is going to be of a huge fitness value. And the further in time that you can make this prediction with accuracy is better than chance, the better. So this is why, of course, people are worrying about global warming, because they can see that that might be a big problem in the future, even if right now we're fine. But we're talking about survival of lineages here. It's what information is important for.

And so what we find is that, therefore, the amount of information that is stored in our genomes really is a good... proxy for what we might call complexity. And when we evaluate that across all biological life forms, it turns out that indeed humans are the most complex of all.

And this nicely brings me to my next question. So humans are complex, perhaps the most complex organisms. But does it tell us? how the intelligence have evolved. Now in our conversation, you kept referring to that information in our genome and the information that we put in our brain or we get in our brain through learning.

So this intelligence that we use where we can store information, we can understand our environment, we can react to our environment, we can make better decisions, we can even predict the future in the sense that climate change might... destroy us down the line. So what do we know about the evolution of intelligence? Right. So that's a good question. So think of it this way.

the amount of information that we have stored in our genome cannot change over our lifetime. And so before there were brains, literally, the only thing that...

life forms would be able to do is to exploit environments that do not change. And in fact, the earliest ones before the advent of what we call nervous systems or brains were just simple organisms that would live at the very bottom of the sea and that would be foraging for detritus or so you know and all they did was go into circles without really any sensory and just gather stuff

So if the environment does not change, then your genes are perfectly fine for survival. But if the environment might change, well then in fact your predictions were about the past environment. And there might have been good predictions and useful predictions in the past, but in the future they're not. And remember that information is always, you know, conditional on the environment. So an environmental change.

might be that something that was information yesterday cannot be information today even though it's the same sequence because remember information is about the relative state between two things the thing that you're predicting and the thing that you're making predictions with so we now realize that once the environment starts changing we better have a way in which we can update information

which in fact is what brains in the end are doing right so give you an example there is a form of life it's basically a barnacle that has two different life stages a juvenile stage where these things are swimming around um finding mates uh uh and and and you know finding food and things like that uh and then creating offspring so on but once they have

done with mating and making offspring what they're doing is that they are attaching to the side of a rock or a boat right and at that point they now are going to be completely stationary they were swimming around and using a brain to do that

You know what the first thing that they're doing after they become stationary? They eject their brain. Because the brain takes a lot of energy. But now that you're stationary, essentially that means that your surroundings are essentially also perfectly predictable. And then you don't need a brain anymore. So the brain is an organ that obviously evolved, but it was an organ that was evolved to make predictions on the changing environment.

And in particular, there are different forms of brains. And you can see that early forms of... brains would have ways to make predictions about slowly changing environments. But the faster the environments are changing, the faster I have to be able to react to them. And the concept of a neuron. from a cell point of view, evolved precisely at the time where the environment was changing so rapidly that the standard chemical ways of doing brain-like things was not sufficient anymore.

So our brain is a world prediction machine whose states are constantly being updated when we realize, hey, what I thought was information has... you know is not information anymore because i i need to acquire more information in order to change what my model of the world is and of course our sensory organs they are all connected to the brain they are acquiring information about

the world so that we can update what we think the world is because the way we are making predictions about the world is we're using this internal model of how we understand the world and so it's like well you know i think the following is going to happen

right and then you're going to act on it and when you realize that that actually didn't work out so well then you try to find out why and then you are trying to incorporate this new information that you gathered in order to make better predictions So there are all kinds of mathematical formalisms that are being or have been constructed to understand this process of changing your internal model. It goes via Bayesian inference and concepts like that. But it's always the same.

You're trying out things based on what you currently know. And if that works, great. And if it doesn't work, update your model until you're not making the same mistake again. In other words, you learn. And on the learning, you're changing the likelihoods for certain decisions. And as you're doing it, essentially you're increasing your fitness because you're making your measurements or your predictions more precise.

And in the end, that's all that is going to decide our future. The future of our lineage is... how good our predictions are. So in an unchanging words, we can rely on our genes to make the predictions, but this has to be augmented by the information that we have acquired over a lifetime in our brains. And the ability to do that.

evolution gave us that evolution gave us the brain right i mean the brain is obviously encoded in you know in sequences and in genes that give us a an organ the brain that will allow you to make predictions that can change as the environment changes. So we have discussed that evolution. is all about information information is being replicated information is being mutated information is being improved enhanced and that's evolution but

What does your research tell us about the origin of life? That before... This very effective process of evolution started creating better and better organisms. There must be a place where these molecules came together. Very simple. Viruses are RNA strands. So what do we know about the origin of life? Right. So obviously the origin of life has been, you know, on the forefront of research for many, many decades.

It is clearly a concept that is important and interesting because we would like to know where we came from to some extent. Now, the standard ways in which people are trying to study the origin of life. They're trying to understand the chemistry that in a sense gave rise to the very, very first replicating molecules. So we believe that this probably happened fairly quickly after the Earth was formed and cooled down sufficiently so that not everything is constantly burning.

And so that for example, so that water could actually be stable on the surface. And that might be something like 4.2 billion years ago. Trying to understand the chemistry of an early planet like that is, of course, very, very difficult. And so you might… think that well let's try to make all possible different environments and see if something happens and you'll be there forever and so

The question of trying to understand the origin of life from an experimental point of view is fraught with almost insurmountable obstacles. And one of the things that I've tried to do is try to look at the origin of life from a different point of view. Namely, once you understand that the essence of life is information, then you could ask, well, what's the origin of information? And once you contemplate that question, you realize very quickly that, ha.

There is a problem with trying to understand the origin of information, namely that information is inherently fragile. It is not easily maintained. In other words, if you have a piece of information, and you just leave it be, it will deteriorate.

due to the interaction with a physical environment, the information will disappear. In fact, that is the essence of the second law of thermodynamics, that you're moving from a low entropy state, which is the one that has information, to a higher information state. In other words, The standard in a physical system is loss of information. In other words, every information that you're starting out with is inherently fragile unless...

you can refresh it. You can keep it information. And what is a good way to refresh and keep information? The answer is to replicate it. If you keep making copies fast enough. Then all these forces that are trying to change it will, of course, be, you know, they're there, but we can protect against it. And that is what Darwinian evolution actually does for us because it is replicating information.

Darwin evolution is a way to maintain information in the onslaught of the physical forces that constantly keep trying to destroy information. But in the absence of replication, how do you... And the answer is you cannot. You cannot. So if you can't protect it, well then how could information that is large enough to actually encode replication occur by chance? And the answer is it can't.

it simply can't and you might think well then the origin of life is impossible it's like ah no fortunately no you might think well maybe it comes from a different planet and rain down or like like it had to also emerge on a different planet So how can a sufficient amount of information be generated that is large enough that in fact it can have information about how to copy the information, which is the essence of life?

And while this is difficult, because you can estimate how much information you need for that, and you can sort of calculate, well, this cannot just simply... emerge by chance but in fact there are mechanisms by which we can actually understand that information can in a system where sequences are being replicated and

Molecular sequences, of course, occur in non-living systems, but they don't carry information. But if they're being replicated, for example, on a scaffold like clay systems, for example, as long as they're being replicated, even though they are replicating... entropy, not information. It is possible to think of a process where in a non-equilibrium manner, information can actually seep from the environment into those sequences. And I talk about such possible mechanisms in my book.

They are very speculative because in fact they have not been proven experimentally yet. However, what it does, it tells you that there are in fact avenues. which are non-equilibrium processes. Of course, on an early Earth, you can imagine many different environments which are strongly non-equilibrium because there's flow of energy.

like in these dark smokers underwater, for example. Plenty of energy and plenty of non-equilibrium where you can imagine ways in which information can accumulate, but only because it's a non-equilibrium system. And then all you need is that at some point you have accumulated a sufficient amount so that, in fact, it has information about how to replicate, at which point Darwinian evolution can start to take hold.

the the problem of the origin of life is really how you bridge the gap between zero information to go to a critical minimum information that is sufficient so that you can actually have information about how to replicate the information Because that's what life is. Life is information that has information about replicating the information. Very profound statement, Chris, here that life is information that has the information to replicate itself.

My next question it goes into a different but relevant direction because I am speaking with a researcher who is a professor of molecular biology. but is also a professor of physics and astronomy. So, Chris, my next question is that you just said that life is information. A very interesting statement.

In the discipline of physics, we are trying to make sense of the reality around us. And we are trying to understand what is this matter, this universe, what it is made up of. And there is a concept emerging in the discipline of physics that maybe everything... is information and this famous saying that it from bit do you see any parallels that information is there at the core of life and maybe information is there at the core

of matter are there any parallels so so this is not an easy question to answer because there are many aspects that one has to talk about when we talking about this idea of, is information more fundamental than matter? So first of all, I don't think that's true. But having said this, There is research that is being conducted where people try to understand the concept of space and time and try to understand whether these are emergent concepts.

right and it turns out that you can try to understand concepts like time in terms of measurement so this is a complex part that I'm sure we cannot get into. But when you're thinking about the foundations of theoretical physics, you realize at some point that time doesn't make any sense it's sort of a weird thing like for example in quantum physics we're talking about observables like everything that's physical has to be observable but time is not an observable

right it's not something like in a sense that we can really measure like it's not i mean we can measure it but it's not a fundamental thing And so many people have argued that perhaps time really is just a convenient thing, but it is not fundamental. believe that is correct carlo rovelli for example has talked about this so in my work in quantum physics for example i found out that the only thing that's really matters is the order of measurements

One measurement occurs before or after another, and that is actually fundamental. And in a way, how we are interacting with our world is just a series of measurements. And the only things...

That's another thing that we learn in theoretical physics, is that the only thing that matters are actually relative states. Absolute states do not have any meaning. The meanings are always relative. We know energies are relative, but we also know that only... difference between entropies namely information actually has any meaning by itself and that this in a sense um analogy

It's not accidental. Really, physics is about the relative state of measurement devices. That's what it is. And once you understand that and you think about measurement devices. or any other interactions really are just interactions between measurement devices then you're realizing that time is really just the relative

influence between measurement devices. In other words, which measurement was occurring before or after measurement. So only the causal relationship between those matters which we actually perceive as time. And once you understand the time emerges in a sense from a network of measurement devices, you realize that the same is possibly the case about the things that are being measured. The qubits, for example, will in fact.

become the origin of space and time this is all very very speculative but i don't think that these speculations you know are meaningless i think there is probably something very profound in you know understanding what is now known as tensor networks and things like that as the origin of space and time. But having said this, the idea of information is more foundational than matter.

I think that Wheeler, for example, who was trying to coin that was maybe really only thinking about the fact that what is really important when it comes to the relative state. of things is their relative state and he thought well the relative state must be information which is not wrong right but then again matter is matter whether or not it is something that we're using to make predictions, whether it has a relative state to another piece of matter. So I wouldn't go too far.

with this idea that everything is information because after all, the laws of physics, how we have constructed them in the standard model, are not talking about information, they're talking about the state. of elementary particles and how they interact so when we understand the origin of the universe and we understand you know how you know, as the universe cools down, that we're giving rise to actually stars and the stars give rise to the elements that we are.

you know, seeing both the elementary ones like hydrogen and helium and so on, but later also the heavier elements, you know, created in nucleosynthesis and so on. understanding of the universe does not talk about the relative state it talks literally about well in a sense the entropy We're talking about the thermodynamics of it. So there is value in thermodynamics, even though meaning, in a sense, is created by the relative state. Now, I don't want to sound as if I know everything about this.

knowledge but my point of view is yes while information is also important to the physical world. In fact, the majority of the physical world is in equilibrium, at least on the time scales that we are looking at. And there, you know, these are zero information.

All of equilibrium statistical physics is a zero information state. Now, if you're looking at the evolution of the universe and you're solving equations about that, Clearly, we're going from a lower... low entropy state to a somewhat higher entropy state but that's also because of the fact that the universe is expanding which means of course that it creates more and more entropy now having said this from a quantum point of view in fact

the entropy of the universe is not changing it is zero it has been zero at the very beginning of the universe and it will remain zero because in fact the quantum entropy of of a wave function is in fact zero so It's very possible that we can come to the conclusion that all of what we're seeing, all of the entropy and variation that we're seeing, is really just the way we are perceiving a quantum reality, which in... true reality nothing ever goes on because

If there is a wave function of the universe, then that wave function evolves in a unitary manner where everything is zero entropy all the time. And then indeed, the only thing that we are talking about are relative states. In other words, You can make predictions about other things, but these are not degrees of freedom that allow you to make predictions about anything that is in fact fundamental.

Because this fundamental level, we are not able to access it because the quantum level of the joint state of the universe is not accessible to beings like us. we are discussing your book the evolution of biological information how evolution creates complexity from viruses to brains we have touched upon a number of topics that you discuss in the book obviously there is much more in the book this is a

big book with many very interesting concepts however is there anything else that you think we should touch upon before we close this discussion well i mean one of the things that you um mentioned uh you know off uh camera on off mic uh is is the difficulty of doing interdisciplinary research so i'm in a sense a clearly

interdisciplinary research you know just this year of course you know saw the publication of my book about the biological information but it also saw the publication of a review paper about the the quantum information theory of black holes At the same time, I have also a paper about quantum measurement and collapse of the wave function. I have other papers about neuroscience and about artificial intelligence.

in a sense, famously interested in pretty much everything. And you could ask yourself, first of all, where does this come from and how does this help? And where it comes from, I think, is that I was... trained in the lab of Steve Kuhnen at the Kellogg Radiation Lab at Caltech. And famously, his group was very interdisciplinary and his mantra was always, look, if there's a problem, that you think is interesting and that you have the tools to solve it, go ahead and do it and don't be afraid.

uh to learn the details we'll figure out the details if you have the chops to do it so in other words he trained you know his group and me you know in particular to not be afraid of uh of attempting to solve problems that you think is are interesting and so i've done this now i also often point out that if i would coming at the on the job market today i could

probably not be hired and it's because people would look at the list of papers in my cv and goes like what on earth is this i don't understand three quarters of the stuff right and so you know our job market and our

infrastructure to fund people is really about funding and understanding specialists. And they are not set up to understand and fund people who are thinking outside the little box uh that that that they're in uh and so that means also that i have gotten news that almost all of my grants are being rejected i mean literally i mean i've written grants in in in quantum physics or black holes where the reviewers go like this is

the craziest thing i've ever seen this guy you know has written five papers on black holes but he's actually in a microbiology lab so i should give him an award for time management how is he able to do that but they also said oh by the way i'm rejecting the proposal because who the hell you know does this guy think he is right and so interdisciplinary thinking which we all understand allows you to solve maybe problems that people who are in the field are not able to solve because they are not

thinking about the possible ways to understand this particular problem from a different angle, that is not being rewarded, not in publications and not in the funding infrastructure. And while I'm lucky, I've had tenure now for over 20 years, that I can simply say, look, if it takes me 10 years to publish a paper, and by the way, my first black hole papers, it took 10 years to get them published.

It would be impossible for somebody like me to go on the job market today. And that is really deplorable. Many times people just simply are very risk averse when it comes to hiring. And when they go like, look, I can't. I can't try to understand all the papers on this person. There's 300 applicants. I don't have the time to do that. I'm going to just pick a safe person, right? There's plenty of applicants.

So it selects against people who actually have potential in the future. So you can talk all night long about how that has a Darwinian evolution analog and so on. But the truth remains. that people that really have broad interests and that are able to see beyond the boundary of the disciplinary education and maybe solve problems that seemed unsolvable.

They are being selected against and I don't know what to do about that. One of the things I'm doing right now is to form a company that will make enough money that I will be able to just simply fund a research lab without… having to convince external reviewers that my research is worth funding because invariably 20 years later what I've been proposing is now like the thing that everybody does but I needed it funded 20 years earlier not you know when it suddenly becomes fashionable.

So it's deplorable, but on the other hand, I can complain about it or I can do something about it. And so in this case, I'm trying to essentially create a company that will generate enough income to fund an inter... disciplinary research laboratory fascinating insights chris in this challenging world of working in a multidisciplinary environment these final comments have led to a thought in my mind that

We have been discussing your book, fantastic book, but I would like to have you again on Bridging the Gaps if you have time. When I was doing research for this discussion, I came across this paper, Information and Black Holes. Perhaps we can talk about something there also and some of your work in quantum physics is also very interesting. Quantum information, you have done some work, whatever you said in past three minutes.

that leads to this request that we would like to have you again on bridging the gaps and we can talk about some of the other fantastic work that you have done i'll be happy to so basically you know i wrote this book of biological information because i wanted to write down everything i've learned about biology and information in a sense so i could start focusing on other things so there's another book i want to

right in this it's about the physics of information and is looking at obviously you know classical information from the point of view not of biology now but you know about applied to physical systems the quantum theory of information the relativistic theory of information the gravitational theory of information and how all of these things are allowing us to actually understand the physical world better like for example

There's this famous black hole information paradox that Stephen Hawking has been talking about since 1975. It turns out that the quantum theory of information solves that paradox.

in that paper from 2024 which talks about stimulated emission and the black hole information paradox is basically a recapitulation of how we can actually understand the information paradox once we actually are trying to calculate information as it is transmitted, basically thinking about the black hole as an information transmission channel.

So I'm perfectly happy. There's many things we can talk about, the quantum measurement problem, black hole physics, relativistic information, and so on. But yes, it would have to be a second interview. Do you have a timeline on this book? Well, no, because I'm focusing a lot of time now on this company.

And in a way, I'm thinking once I have established that company in such a way that we can have a steady stream of income and I can hand it off to other people, then I will probably start writing that book. I mean, I have an outline of the book already, but it's basically going to be the other book I'm going to write, and that's probably going to take a decade or so.

Professor Kristof Adami, thank you very much for being with me. Thank you very much for this fantastic discussion. It has been an absolute pleasure having you on Bridging the Gaps. Well, I enjoyed myself and that was a fun discussion for me as well. Thank you and goodbye. Goodbye.