Finding life we can't imagine - Christoph Adami

So I have a strange career. I know it because people come up to me, like colleagues, and say, "Chris, you have a strange career." I can see that point because, you know, I started my career as a theoretical nuclear physicist, and I was thinking about quarks and gluons and heavy ion collisions. I was only 14 years old. No, no, no, no, I wasn't 14 years old. But after that, I actually had my own lab in a computational neuroscience department, and I wasn't doing any neuroscience. Later, I would work on evolutionary genetics and work on systems biology.

But I'm gonna tell you about something else today. I'm gonna tell you about how I learned something about life. I was actually a rocket scientist. I was a rocket something. I wasn't really a rocket scientist, but I was working at the Jet Propulsion Laboratory in sunny California, where it's warm, whereas now I'm in the Midwest and it's cold. But it was an exciting experience. One day, a NASA manager comes into my office, sits down, and says, "Can you please tell us how do we look for life outside Earth?" That came as a surprise to me because I was actually hired to work on quantum computation. Yet, I had a very good answer. I said, "I have no idea."

He told me, "Bio signatures! We need to look for a bio signature!" I said, "What is that?" He said, "It's any measurable phenomenon that allows us to indicate the presence of life." I said, "Really?" Because it's not easy. I mean, way of life. Can't you apply like a definition, like for example a Supreme Court-like definition of life? And then I thought about it a little bit and I said, "Well, is it really that easy?"

Because yes, if you see something like this, then all right, fine, I'm gonna call it life, no doubt about it. But here's something: it goes like right, that's life! I know that extent. If you think that life is also defined by things that die, you're not in luck with this thing because that's actually a very strange organism. It grows up into its adult stage like that, and then goes through like a Benjamin Button phase, initially goes backwards and backwards until it's like a little embryo again, and then actually grows back up and back down and back up, sort of yo-yo, and it never dies. So it's actually life, but it's actually not as we thought life would be.

And then you see something like that, and it was like, "My God, what kind of a life-form is that?" Anyone know? It's actually not life. It's a crystal. Once you start looking and looking at smaller and smaller things, this particular person wrote a whole article and said, "Hey, these are bacteria!" Except if you look a little bit closer, you see in fact that this thing is way too small to be anything like that. So he was convinced, but in fact most people aren't. Then, of course, NASA also had a big announcement. In fact, President Clinton gave a press conference about this amazing discovery of life in a Martian meteorite. Except that nowadays, it's heavily disputed.

If you take the lessons of all these pictures, then you realize, well actually maybe it's not that easy. Maybe I do need a definition of life in order to make that kind of distinction. So can life be defined? Well, how would you do about it? Of course, you go to, you know, Encyclopedia Britannica and open that, L. No, of course you don't do that. You put it somewhere in Google and then you might get something.

What you might get, and you go there, anything that actually refers to, you know, things that we are used to, you throw away. Then you can might come up with something like this, and it says something complicated with lots and lots of concepts. Who on earth would write something as convoluted and complex and inane? It's actually, you know, a really, really important set of concepts. So I'm highlighting just a few words and saying definitions like that rely on things that are not based on amino acids or leaves or anything that we are used to, but in fact on processes only.

If you take a look at that, this was actually in a book that I wrote—that this was artificial life. And that explains why that NASA manager was actually in my office to begin with, because the idea was that with concepts like that, maybe we can actually manufacture a form of life. So if you go and ask yourself what on earth is artificial life, let me give you like a whirlwind tour of how all this stuff came about. It started out quite a while ago when someone wrote one of the first successful computer viruses. For those of you who aren't old enough, you have no idea how this infection was working, namely through these floppy disks.

But the interesting thing about these computer virus infections was that if you look at the rate at which these infections work, they show like the spiky behavior that you're used to like from flu viruses. It is in fact due to this arms race between hackers and operating system designers that things go back and forth. The result is kind of a tree of life of these viruses; a phylogeny that looks very much like the type of life that we're used to, at least on the viral level. So is that life? Not as far as I'm concerned. Why? Because these things don't evolve by themselves. In fact, they have hackers writing them.

But the idea was taken very quickly a little bit further when a scientist working at the Santa Fe Institute decided, "Why don't we try to package these little viruses in artificial worlds inside of the computer and let them evolve?" This was Steen Rasmussen, and he designed the system, but it didn't really work because these viruses were constantly destroying each other. But there was another scientist; they had been watching this, an ecologist, and he went home and said, "I know how to fix this," and he wrote the CHERA system. In my book, it is in fact one of the first truly artificial living systems.

Except for the fact that these programs didn't really grow in complexity. Having seen this work a little bit, this is where I came in, and I decided to create a system that has all the properties that are necessary to see in fact the evolution of complexity—more and more complex programs constantly evolving. And of course, since I really don't know how to write code, I helped, you know, two undergraduate students at the California Institute of Technology that worked with me: Charles Ofria on the left, Titus Brown on the right. They are now actually respectable professors at Michigan State University, but I can assure you, back in the days, you know, we were not a respectable team.

I'm really happy that no photo survives of the three of us anywhere close together. But what is the system like? Well, I can't really go into the details, but what you see here is some of the entrails. But what I want you to focus on is this type of population structure—about ten thousand programs sitting here, and all different strains are colored in different colors. As you see here, there are groups that are growing on top of each other because they're spreading.

Anytime there's a program that's better, that's surviving in this world due to whatever mutation that is acquired, it's going to spread over the others and drive the others to extinction. So I'm going to show you a movie where you're gonna see that kind of dynamics. This movie star—I mean, these kinds of experiments are started with programs that we wrote ourselves. We write our own self-replicator, are now very proud of ourselves, and we put them in. What you see immediately is that there are waves and waves of innovation. By the way, this is highly accelerated, so it's like a thousand generations a second.

But the system goes like, "What kind of a dumb piece of code was this? This can be improved upon in so many ways so quickly!" So you see waves of new types taking over the other types, and this type of activity goes on for quite a while until the main easy things have been acquired by these programs. Then you see sort of like a stasis coming on, where the system essentially waits for a new type of innovation like this one, which is going to spread over all the innovations that were before and is erasing the genes that it had before until a new type of higher level of complexity has been achieved. This process goes on and on and on.

What we see here is a system that lives in very much the way that we're used to. Life goes. But what the NASA people had asked me really was, "Do these guys have a bio signature? Can we measure this type of life?" Because if we can, may we have a chance of actually discovering life somewhere else without being biased by things like amino acids. So I said, "Well, perhaps we should construct a bio signature based on life as a universal process." In fact, it should perhaps make use of the concepts that I developed just in order to sort of capture what this simple living system might be.

The thing I came up with—I have to first give you sort of an introduction about the idea, and maybe that would be a meaning detector rather than a life detector. The way we would do that, it's like, "Okay, I would like to find out how I can distinguish text that was written by a million monkeys as opposed to texts that are in our books." I don't like to do it in such a way that I don't actually have to be able to read the language because, you know, I’m sure I won't be able to. As long as I know that there's some sort of alphabet, so here would be a frequency plot of how often you find each of the 26 letters of the alphabet in a text written by random monkeys.

Obviously, each of these letters comes off about roughly equally frequent. But if you now look at the same distribution in English text, it looks like that. I'm telling you, this is very robust across English text. If I look at French text, it looks a little bit different, or Italian, or German. They all have their own type of frequency distribution, but it's robust. It doesn't matter what it writes about—politics or about science; it doesn't matter whether it's a poem or whether it is in a mathematical text. It's a robust signature, and it's very stable as long as our books are written in English. Because people are rewriting them and recopying them, it's going to be there.

So that inspired me to think about, "Well, what if I try to use this idea in order not to detect random text from text with meaning, but rather detect the fact that there is meaning in the biomolecules that make up life?" But first, I have to ask, what are these building blocks? Like the alphabet elements that I showed you? Well, we have many different alternatives for such a set of building blocks. We could use amino acids; we could use nucleic acids; we could cover silica acid, fatty acids. Chemistry is extremely rich, and our body uses a lot of them.

So that we actually to test this idea first take a look at amino acids and some other carboxylic acids. Here's the result. Here is in fact what you get if you, for example, look at the distribution of amino acids on a comet or in interstellar space or in fact in a laboratory where you made very sure that in your primordial soup there is no living stuff in there. What you find is mostly lysine and then alanine, and that's a trace element of the other ones. Okay, that is also very robust.

What you find in systems like Earth where there are amino acids but there is no life. But suppose you take some dirt, dig through it, and then put it into these spectrometers. Because there's bacteria all over the place! Or you take water anywhere on Earth, because it's teeming with life, and you make the same analysis. This spectrum looks completely different. Of course, there's still glycine and alanine, but in fact, there are these heavy elements—these heavy amino acids that are being produced because they are valuable to the organism, and some other ones that are not used in the set of twenty will not appear at all at any type of concentrations.

So this also turns out to be extremely robust. It doesn't matter what kind of sediment you're using to grind up, whether it's bacteria or any other, you know, plants or animals. Anywhere there's life, you're going to have this distribution, as opposed to that distribution. And it is detectable not just in the amino acids. Now you could ask, "Well, what about these variants?" They're variants being the denizens of this computer world where they are perfectly happy replicating and growing in complexity.

So this is the distribution that you get if in fact there is no life. They have about twenty-eight of these instructions, and if you have a system where they're being replaced one by the other, it's like the monkeys riding on a typewriter. Each of these instructions appears with roughly equal frequency. But if you now take a set of replicating guys, like in the video that you saw, it looks like this. So are there some instructions that are extremely valuable to these organisms, and their frequency is going to be high? There's actually some instructions that you only want to use once, if ever. So they write either poisonous or really, you know, should be used at less of a level than random, and in this case, the frequency is lower.

Now we can see, is that really a robust signature? I can tell you indeed it is because this type of spectrum, just like what you've seen in boxes and just like what you've seen in many ways, it doesn't really matter how you change the environment. It's very robust; it's going to reflect the environment. So I'm going to show you now a little experiment that we did, and I have to explain to you the top of this graph shows you that frequency distribution that I talked about—okay? Here, in fact, that's the lifeless environment where each instruction occurs at an equal frequency. And below there, I show in fact the mutation rate in the environment.

I'm starting this at a mutation rate that is so high that even if you would drop a replicating program that would otherwise happily grow up to fill the entire world, if you drop it in, it gets mutated less immediately. Okay, so there is no life possible at that type of mutation rate. But then I slowly turn down the heat, so to speak, and then there is this viability threshold where now it would be possible for a replicator to actually live. Indeed, we're going to be dropping these guys into that soup all the time.

So let's see how that looks like. So first, nothing, nothing, nothing. Too hot, too hot. Now the viability threshold is reached, and the frequency distribution has dramatically changed and, in fact, stabilized. So now what I did there is I just—I was being nasty. I just turned up the heat again. And, of course, it reaches the viability threshold. I'm just showing this again because it's so nice. You hit the viability threshold; the distribution changes to alive. Then once you hit the threshold where the mutation rate is so high that you cannot self-reproduce, you cannot copy the information forward to your offspring without making so many mistakes that your ability to replicate vanishes, and then that signature is lost.

What do we learn from that? Well, I think we learn a number of things from that. One of them is if we are able to think about life in abstract terms, and we're not talking about things like plants, and we are talking about amino acids, and we're not talking about bacteria, but we think in terms of processes, then we can start to think about life not as something that is so special to Earth but that in fact could exist anywhere. Because it really only has to do with these concepts of information—of storing information within physical substrates. Anything—bits, nucleic acids—anything that's an alphabet, and make sure that there's some process so that this information can be stored for much longer than you would expect the time scales for the deterioration of information. If you couldn't do that, then you have life.

So the first thing that we learn is that it is possible to define life in terms of processes alone, without referring at all to the type of things that you know we hold dear as far as the type of life on Earth. And that, in a sense, removes us again, like all of our scientific discoveries, or many of them, it leads to a continuous dethroning of man, of like how I think we're special because we are alive. Well, we can make life. We can make life in the computer—granted it's limited—but we have learned what it takes in order to actually construct it.

And once we have that, then it is not such a difficult task anymore to say if we understand the fundamental processes that do not refer to any particular substrate, then we can go out and try other worlds, figure out what kind of chemical alphabets might there be, figure out about the normal chemistry, the geochemistry of the planet so that we know what this distribution would look like in the absence of life and then look like to large deviations from this, this thing sticking out which says this chemical really shouldn't be there.

Now we don't know that there is life then, but we could say, "Well, at least I'm gonna have to take a look very precisely at this chemical and see where it comes from." That might be our chance of actually discovering life when we cannot visibly see it. So that's really the only take-home message that I have for you: life can be less mysterious than we make it out to be when we try to think about how it would be on other planets. And if we remove the mystery of life, then I think it is a little bit easier for us to think about how we live and how perhaps we're not as special as we always think we are.

And I'm gonna leave you with that, and thank you very much.