 All right, so hello everyone, and welcome to this first seminar in our Trinity term schedule. Today our speaker is Sonia Contella. Sonia is a professor of biological physics at the physics department of the University of Oxford, and she also serves as the associate head for Equality, Diversity and Inclusion. Her work is at the interface of physics, biology and nanotechnology, and she is interested also in how biological shape relates to information processing. She is an honorary member of the Society for Natural Sciences and the author of a group called Nanocomps to Life, how nanotechnology is transforming medicine and the future of biology. Today she will talk about how a natural system can inspire information processing and computational techniques. The title of our talk is It's From Bit, the Future of Bio-Inspired Computing Machine Learning. Thank you very much for joining us, Sonia, and the floor is yours. And yes, as I have mentioned, the talk is recorded and will be available on the YouTube channel. Thank you. Good. I hope you can see my screen. Yes. Yes, we can see the screen. Yeah. Okay. So thank you for inviting me to talk. I was very hesitant to speak in this series because I'm no expert in machine learning, but I met Peter Bram, who is organizing the series with Maxense, and we were talking a lot about biological shape, which is why I'm interested about myself. And also I decided to speak in the end because I've been following some of the talks that have been given in this seminar series, and they're all very interesting, and many of them actually are touching on biological problems, but I also thought I felt when I was seeing them, I felt a bit a bit frustrated that most of this talk are missing the big picture, which is why are we using bio-inspired algorithms, such as machine learning based on artificial neural networks, to solve the physical or non-physical of complex problems of the 21st century, which is what we are doing. I've been thinking a lot about this. I've been thinking a lot over the last 10 years of why me as a nanotechnology started to do biology and became interested in information processing, and why not only me, but many people are converging, as we are seeing this series, into bio-inspired computing paradigms to solve their problems. So hence the title of my talk, It From Bit. It From Bit, for those of you that like quantum mechanics, you know, is a sentence by Johnnie Wheeler, who inspires my talk today and is one of my heroes, who was the first one, maybe one of the first ones to realize that there's a computational and you can interpret nature as a computational process. And he realized that from quantum mechanics, the idea that in order to have reality, you need a kind of computation, a kind of participation of the experiment in order to reality to become clear. So my argument in this talk is that it's not only the physical reality comes from a quantum computation, but perhaps also biological reality comes from a physical computation, which is what we are trying to understand when we develop these bio-inspired algorithms. So I will take you through my vision of how we are evolving into understanding comprehensive systems, trying to, at the same time, understand ourselves and our own intelligence and as physicists, why that the universe actually creates clever intelligence shapes. So as I said, I started out as a nanotechnologist in the 1990s. People like me that were interested in how matter was behaving at the nanometer scale, we started to become interested in biological systems because we are made of nanomachines. So up to then we had developed at least in the since the 1950s onward since the invention of crystallography for proteins and for DNA. A consistently and increasingly dominant picture of biology was emerging as a kind of digital computer in which you had DNA and proteins that there were bits and they will connect with each other in some kind of networks, complex ways that we didn't understand. And that was all that we needed to do to understand life. That yet people like me were coming from materials were coming from quantum and condensed matter physics. And we were interested in things like when you put many atoms together and they interact strongly with each other, the stuff emerges that is not like the atoms that is made from. So out of particles, out of bits, can come continuum, can come emergence, can come complex systems. So I put my atomic force microscope that was used at the time for use of nanoparticles to use up proteins and DNA. And I didn't find bits I found shapes. So, of course, the proteins like this one here, this is a potassium channel, one of the proteins you have in your cell membranes that allows potassium to go in and doesn't allow sodium to go in. And that's one of the reasons you can function. It is these proteins are the product of four billion years of evolution. One third of the age of the universe have been devoted to creating these shapes and of course have a digital part. They have a chain with a specific change with a specific components. They're made of Lego blocks, the amino acids are put together in the right way and they fold into the correct shape. So biological shapes at the nanoscale are not bits. There are clever combinations of bits and shape creating function. These things move a lot. They are run by the principles of non-equilibrium thermodynamics. When you are a string and you are in water and your nanoscale, you are able to create and reduce your entropy because you can dissipate energy into the environment. By example, a chemical reaction or a movement. By dissipating energy into the environment, you can create a shape. You can reduce your entropy. You can encode information. And if you are shaped and used in time, you're already encoding information about the environment. So there's something very profound that we were starting to learn as physicists working in soft matter at the nanoscale liquid, which is not only eat from bits in the quantum case, but is also an eat from bit in the soft matter case. When you're soft and you're in water and you can wriggle and you're a string, you can encode information in your shape and you can encode information in the pieces, in the polymer and polymer are made from from different bits. You're cutting code information into the chain that makes you, but also into the way you fold yourself. You create a shape. And that shape is entangled with the energy and whatever is happening in the environment and is changing and responding and changing the environment itself. It's extremely complex. It's linking thermodynamics, information process and time and mechanics through mechanics. So eventually these things move, but we do with our microscopes, and this is why I started to create these microscopes and work with people that make these microscopes, if I managed to make this movie, move there, is that these things move. They're mechanical shapes that are able to move. So for example, here at the top is the ATP synthase. It's a prism that you have in every single cell of your body and on every living algorithm. It's a rotary motor. It's the perfect nanomachine. Every atom is in place, is everywhere and catalyzes a chemical reaction with almost 100% efficiency, just using the temperature of the water around it and creates rotation in order to catalyze a chemical reaction. This is very far from just a digital vision of biology that most molecular cell biology was presenting us. This is shape link with digital information. And here at the bottom is another amazing linear motor, an acting molecule that sits on, sorry, a myosin molecule that sits on the track. You have these things all over your cells and they're responsible from hearing, from every movement you have, every time you have a contraction. And this relies again on the machine able to use temperature and shape. And of course the information encoded in this sequence by the 4 billion years of evolution to create amazing functionality. So that is interesting enough, but of course it's not the complete picture of biology and it still let me very dissatisfied when I was working on this, because the whole thing of biology is that out of those little bits, yet move on to another scale. So you're going across scales to create millimeter, meter size machines that are able to integrate all this and coordinate. They need to make computations that are able to control time. So all these things are coordinated because to be able to function and to understand the keys to control the time that all the bits are working at the same time and in the right way. So the key for me is to understand maybe a simple system, not so simple. Why a leaf sees the sun and moves towards it. That is not a digital computation. It's an analog computation that is underpinned by digital processes. So this is like me, and this is where it comes my frustration and why I move on to try to understand better information theory, go frustrated in that I can, from my biological physics world in which I live, I measure how these shapes emerge in a mechanical way. I measure the analog part. I measure the mechanical properties, the continuum properties in the whole story. So like me, not only biological physicists, many of us are trying, in all the sciences and technologies involved, are converging into biology. So I explore all this. Why are we right now converging into biology and everything we do in my book? I'm going to promote here. Nano comes to life, which is when I first started thinking about these problems. It's very interesting historically, the idea that most of the 19th and 20th century science is quite reductionist. And there are many reasons for it. One is that the mathematics, we knew the computers and the capacity to calculate complex system we had, but also historical development of things like colonialism. But I'm talking about that tomorrow in the talk in the Japanese study center here in Oxford about colonialism and physics, but not today. So basically we are at the end of reductionism in a way. Our computer technologies and also the kind of problems we need to solve cannot be reducted anymore to bits. We need to understand complex systems. So that means that people like me in nanotechnologies move to biology, but obviously people trying to make robots, robotics, drones for wars, creating new materials, materials that are smart, materials that are going to substitute a steel and concrete when we make new buildings are also looking into biology and how biology constructs, how biology creates wood or how biology creates other materials that are going to do the things we want to do. In the background of this is global warming. Of course, in itself a very complex system, a very complex problem that produces an enormous amount of data and that will require very clever algorithms to solve. And arguably the global warming itself has been caused by the reductionist way of doing science of the 19th and 20th century. So basically most of the problems we're trying to tackle involve big data, involve complex systems that evolve with time, and the computer capacities we have right now that are very limited, not only and also our capacity to understand the physics, the actual models that will allow us to have these algorithms that we need in order to understand and solve these problems. So what I think now, after all these years of thinking of why we cannot solve biological problems, why we cannot solve complex problems, and what is the way forward is this one. So biological systems such as my plant or the protein I showed before have been trying, people have been trying to be understood by the typical biology approach with these genes and proteins. So it's a digital representation of biology which has many limitations. One of the best arguments for this is that in the last 30 years a reductionist approach, basically a digital approach to cancer has produced very few drugs and the drugs that have been produced don't improve very much the outcomes of cancer treatment. And yet for example, now within immunotherapists which is using biology to tackle your cancer an analog system that can be primed like in the COVID vaccines. We can talk about that later if you want with digital algorithms that you encode in your RNAs of the area of your vaccine you can actually start to tackle cancer. So biology has its limitations looking at a biologist just from a digital point of view. However, this digital point of view is what has attracted most theoretical physicists. So theoretical physicists usually look at the interconnections between proteins and genes and the complex networks they make and use complex systems approaches to try to solve what is going on. Yet people like me and applied mathematicians and engineers that are trying to make applications out of biology such as creating tissues in regenerative medicine or COVID vaccines or try to understand how a plan grows. Gross, simply we use continuum models, differential equations. We use concepts from mechanics. We use models like finite elements or we use things like electricity. And these worlds don't talk. And you can see that from the lectures the series of lectures we're having already in your series of sevens of machine learning. So the people using machine learning are usually looking at the biological approach. They're looking at trying to find the complex networks between genes and proteins but they never talk about to the people they're actually looking at how these networks are connected by a shape. So the analog world is not talking to the digital world to understand complex systems. And I think this is where the opportunity lies right now for ambitious people to try to join these worlds. And what I'm trying to give you in my talk is I think there's a lot of possibilities and a lot of reasons of why would you like to do that. One is physics. We want to understand why the universe creates these shapes that are so clever and how they do it. And also in physics departments we have the tools to do this or at least that we link to engage with people that will help us to show these problems which is non-equilibrium thermodynamics as I told you at the beginning is time is how we control time and shape and how is that linked together with information and information series. So I think this is how all of this is coming together right now. You start to see examples in the literature in here and there of how many of us are trying to come to this new paradigm in which we link the analog and the digital to understand complex problems. Now I put here the paper of our Louis that came out recently and follows up one of the seminars he gives you in your series because he's also thinking about these things. He's actually looking in a more of the complex system way and I recommend you to read his recent paper which just algorithmic arguments called Mogorov complexity. He explains that in his talk that you can see in the videos. He explains how just simply from the algorithm nature in biological systems produces symmetry. So in the last month I've been talking lots to James Semple who is sitting here in one of the only squares there is no black and to Arjan Ardavan trying to develop our own vision of how from a physics department can we really make a difference in tackling complex problems in the 21st century if I have in this physics approach on using ideas from biology and nature to be advised by a scope region. So this is a little bit quickly the motivation of the ideas we have developed together. So of course the reason why we are using now digital computers to simulate physical phenomena and you include in machine learning to understand reality and to create new technologies is because our computers work really well after they invented the transistors in 1947 and all this early development of architectures. Well you know it's very it's been good to use your computer to solve your problems. We all learn how to program them and we have been trained on how to do it. The problem right now is that the type of problems we're trying to solve have too much amount of data and they're very complex. So the problems are becoming increasingly complicated to code and to handle the computers are never big enough. The computers in the current way they are designed semiconductors will also hit the limits. We are now in transistors below nine nanometers and there's not much further you can go and also they spend and this is a big problem an enormous amount of energy. So the idea is that when there always exists an alternative the analog computer an analog computer is trying to create an analog version of the physics you're trying to simulate in which you don't need to simulate it. You just have your device itself introducing the physics you want. So an example will be an analog computer where the electronic circuit elements are the differential instantiate the differential equations that you're trying to solve. Of course there's nothing new and for those of you that like history of devices probably you have seen recently in the press desk machine which is 2000 years old the Antikythera mechanism that was found in Greece which is an incredibly complicated analog machine full of clever wheels that is able to produce the positions of the stars and it's just using physics not digital. Of course these ideas are not new and the same guys that created the digital computer Alan Turing, Johnny von Neumann and their interactions between them they were thinking of by inspired computing from the get go. So they had ideas in the 1940s of indeed neural networks and representations using neurons of how the computations were going to work. But again the success of the digital computer means that we forgot all these other ways of computation until now. So as I said before we having in the 21st century to tackle a lot of problems that are very complex. So from tumors or pandemics to global warming to new materials that allow the global warming not to explode. So all these are actually interconnected and one thing we love doing and that we are seeing a lot which is autonomous vehicles or drones. I think that the pandemic that we had is going to bring a boom in by inspired computing not only in the design of things that I'll take you later but the way we design them including things like imitating the immune system to create new computers and the war in Ukraine is already seeing a search on the use of drones autonomous vehicles and all of these are using different versions of machine learning but also swarm computing that I will review very quickly now. Also increasingly on our computers so neuromorphic computers are already being designed unused and I don't know if they're being used in Ukraine my guess is that I should be okay here. My guess is that yes they are because most of these drones are actually have been tested in the wars in Syria and since 2014 in Ukraine. So this is the situation we have right now. This is a summary of what I told you before. The idea is that we take inspiration from biological systems to combine the digital and the analog as I told you in the previous slides biology at all its scales is using this dual code mixing digital with analog. So this is what I told you in words and now I'm going to go to the summary of what I'm going to tell you about what people are doing in this world of bio-inspired computing. So first I will very quickly review all the biology inspired computing models that I have found or I found a few more but these are the ones that are most used implemented in digital based computers which is how we're actually now doing things like machine learning based on artificial neural networks. So we get inspiration from a system that works in biological or natural world and then we encode that into a silicon based computer. The next wave which is starting to come out is for me more exciting is the Cyborg computing revolution which in which nature becomes the implementation hardware. So hardware and software merge into a material which can be biological artificial or in between and there are already many examples of this starting to happen. And finally we're going back to the concept of Johnny Wheeler on many others understanding and trying to look at nature as a computer. As I see it right now as soon as time moves forward there is some kind of computation something is going on is creating intelligence of some sort and this is something that Tristan Farrow which is in the audience knows much more than me and maybe we can discuss in the end. So I'm going to go very quickly about just telling you different types of inspired computing that are merging with artificial neural networks in machine learning. So we learn to download the machine learning algorithms from Python but the people working on this are working in many different strategies that combine with artificial neural networks to actually solve the problems of trying to solve. The oldest is the cellular automaton which was already thought by von Neumann and Stahl Ulam in the 1940s. The reason was it was the 1940s is because they were working on the Manhattan project and they were working on thinking computers together. So they're already thought of the cellular automaton which probably you know about this is perhaps the most famous ones which is The Game of Life created by John Horton Conway in the 1970s. By the way he died of COVID during the pandemic which is quite sad so basically you it's a final set of cells that you create and then each cell evolving parallel at these critical time steps follow some transition rules and by doing that you create a computer that is able to compute in parallel distributed fashion there's no tape no street of instructions but what you have the evolution of the cells is saying so basically tells you it's a computer by evolution of what happens in the behavior of the cells and how they interact with each other. Of course Stephen Wolfram has made this way of computing very popular with his with his books and he has this idea that the world is pixels and everything can be interpreted by cellular automaton of some sort. Many of us think that that might not be true but anyway particles and continues will always be the most perhaps most controversial topic of physics. So these things are still used now and they use a lot for all sorts of problems biological growth reproduction I put a few here pattern formation and diffusion. One beautiful thing that happens with every computing model that all these guys were developing in the 1940s and the 1950s is that if you look somewhere in nature you will find it implemented. These guys found that the leaves the stomatine leaves actually work by a kind of mechanism that is very similar to a cellular automaton. So in the leaves you have this is from my lab we look at the you have these little pores and the pores open and close to optimize and the amount of oxygen what do they take in. So they are in the leaves if you look in the leaf they look like that so they solve a formal optimization problem by adjusting their apertures and although they're separate they are actually work as connected networks of course because they are connected by a mechanical medium which is what I study I study how the time responses of this medium responds to the movement of the stomata so shape matters they're not only networks they're networks connected by shape by non-equilibrium several dynamics by mechanics so they synchronize as I said at the beginning one of the most important things in biological systems is that you need to synchronize all the bits in order to make the computation and they actually work like the cellular automaton which is amazing that time and again you find all these things in nature so you know artificial neural networks and I'm not going to go through them because this is the the topics of all the other talks have been going through they're very old they're also from the 1940s and they were proposed by these two guys which were no computer scientists and they were no physicists they were mixtures of logicians and and psychologists and they came up with the idea of artificial neural networks and they attracted the word for Neumann and others that that relate to what we have now another interesting way of I'm going to check the time another interesting way in in making computer algorithms that is not artificial neural networks is evolutionary computation so it's the idea of using Darwinian evolution to to create your algorithms so they're used in a wide range of things including finance, economics, complex systems because it's time-evolved and basically they consist of looking for possible solutions of a problem as individuals in the population and to include some kind of fitness to to decide the quality of the solution they're already used in real-world applications and I found you one nice one of an antenna made by NASA for a space technology five mission in which they use evolutionary algorithms to create the shapes of the antennas which are very cute they look like little sculptures that they were using in their in their system and it became the first evolutionary algorithm-based thing that has been put in space as a real device Swarming intelligence is becoming increasingly important in this area because it's the idea of how you connect things is important for companies that are trying to work on the internet of things in medical data set heating systems moving out just tracking of course traffic once we go to autonomous vehicles you will like to have some kind of swarm intelligence straight to these things move together more worryingly these things are going and are being developed very much at the moment because of the war in Ukraine so there is a lot going on right now in Ukraine of drones working together and swarming the first time a mission by drones and swarming could kill actual soldiers apparently happened in 2014 by the Russian army in the beginning of the Ukrainian war that now has exploded and there's already a lot of literature on the topic with a lot of ethical problems I think this will take over a lot of things deliveries from Amazon the way they track the way we move so yes welcome to the drone wars and as I said the immune system is an amazing thing that we are starting to be able to use for medical treatments one example that we can maybe all understand is the covid vaccines so the covid vaccine was an amazing example of how medicine of the future is already started it will work so the Chinese authorities released the the sequence of DNA of the virus very early on in the days of the pandemic and immediately the teams of Moderna BioNTech and indeed Baxitec in Oxford within days using machine learning algorithms and whatever they took they designed their vaccines they had vaccines ready for clinical trials within weeks these vaccines are combinations of nanotechnology in the case of Moderna BioNTech with RNA technology immunology and a great understanding or a better understanding of the immune system the idea that you can get the immune system to tackle not only pandemics or viruses but actually cancer which motivated these companies to exist in the first place both Moderna BioNTech are over 10 years old and they were trying to use the cleverness the analog cleverness of the immune system to tackle cancer and eventually they help us to survive so far the pandemic I expect people have been using artificial new immune systems in digital format for a while for things like antivirus in your computers but I expect that the search and research in immune system will also bring us a search and systems that use this the inspiration from immune systems to create new algorithms another area that is quite interesting and that is inspired a lot of new applications is the idea that if the more we know about the connectivity of neurons in living systems it can help us to design algorithms such as artificial neural networks that are first made for specific purposes so an interesting project that I bring here is open worm which belongs to the artificial life type of bio-inspired computing so in open worm people just look at how the neurons of C elegans C elegans is a little worm that biologists use for many many years as their model system to understand many things because it's very easy to grow in the lab and it's transparent and also has a very interesting a range of neurons there don't have so many neurons and interesting they have the same amount of genes as humans these little worms so people are trying to understand how the architecture of the neurons actually affects the way they the worm behaves so they are implementing algorithms that are mimic as accurately as possible the interconnections between these neurons and then they put them in little robots and these little robots have sensors that try to imitate in a very not very sophisticated way as you can see the sensing abilities of the worm and by doing biology computers and robotics we're trying to create new ways to understand how is the awareness of this worm of the world people are really using these for real world robotic applications including all sorts of things that are coming up with in cities even for navigating through pipes etc there are many other ways in which people are trying this approach is the idea of is quite philosophical how life arises from inner matter where are the living so basically in a bit is a good summary of how all of us are converging in these very fundamental questions why the universe creates matter that is intelligence and what does it mean another example of membrane computing which is based on ideas also from biology the way the proteins are embedded in the cell membrane the membranes of the cell communicate the inside well on the outside world and many problems people have thought that can be solved such as some problems of cryptography or or computer graphics with this approach I find this a very interesting way of making algorithms which is called amorphous computers so basically this you can see here a wave I got it from a lab in MIT that is working on this this wave is similar to the way equal Drosophila these little flies the biologists study there's a lot of work right now in understanding how the embryo develops into the shape of the fly their fizzes they're all sorts of people working on these problems each of us are working on it separate nobody talks to each other and these ones are making their computer approach to this problem so trying to understand the engineering principles that can be used to assert control organizers plot the behavior of programmable multitudes basically what they're trying to mimic is how cells communicate to each other in order to create shape over time so that is my summary of what people are trying to do to implement bio-inspired computing models into digital computers for me the most interesting part and is the way I'm starting to work on is in the next wave nature as implementation hardware the idea that you can get your software and your hardware within the same system so this story it starts with molecular or maybe not but I guess this was one of the early examples with Adelman which I read his work when I was doing my PhD in Japan and he was the first one that proposed the DNA could be used as a computer this idea has been taken by many people he proposed this idea in the 1990s and it's taken a long time to realize it's not easy to make a molecular computer I encourage you to look at this paper that was published by people of our department last week but John Bath Eric Benson Raphael in the group of Andrew Traberfield created they all created this beautiful DNA molecular printer capable of programmable positioning and patterning in two dimensions using DNA so again the idea that you can create a machine that encodes shape function and digital information within itself the most used and already making an impact in our world is neuromorphic computer a neuromorphic computer where Carver made thought about them many many years ago and is the idea that you can use the solid state devices that are in your computers not as digital but as analog devices so you can use them a little bit more like brain in sort of like neurons not digital but analog neurons a lot of people are investing a lot in this so NeuroGrid in stand for the human brain project in European Union has 1.3 billion been allocated to the project to try to understand a little bit what I was telling you before with the worm how the connectivity of the neurons in the human brain can lead to neuromorphic computing and the other way around can we use neuromorphic computer to understand better the working of the human brain so IBM is also working on this brain scales and perhaps the most advanced is the IMEC which is a research institute in Belgium has created a neuromorphic computer that is already being in use Heidelberg has brain scales and as far as I know I think these brain scales already implemented in a drone who knows where these drones are going to be deployed and a drone is the perfect place to test a neuromorphic computer because what a neuromorphic computer and an analog computer that is to be very fast it doesn't have to go through the algorithm the response like we respond we have memories in our brains but we also have emotions and feelings and we can respond very quickly to whatever is going on without having to do a computation just without or analog capacity and this is what this type of computers implemented in in drones are trying to do so out of my frustration of doing biology just thinking about mechanics and thermodynamics and not being able to do information process I created or a kind of network people interested in these problems and this is my first project that we are trying to get funding for with brains in a dish so it's a collaboration with Antoine Jerusalem and engineering Yoshikatsu Hayashi in the University of Reading he's very good at robotics and soft robotics and and neuroscientists mainly David Dupré here in Oxford so what we're trying to create is a neural network on a dish and we are going to try to understand what get neurons to oscillate and coordinate for example in the gamma frequency band the gamma frequency band is is linked to consciousness basically neuroscientists now think that in order to be conscious and put together all the different information we get from the outside world like hearing or seeing and that all comes into a coherent picture the neurons need to oscillate in this frequency range we're trying to make cells oscillate in a frequency range my student Hendrik had the idea of using piezoelectric lithium niobate substrates to achieve these and our I think what gives us an edge against other people trying to understand this is that we will try to understand and put together the physics the non-intermodal dynamics and the information theory the information processing of the neurons we are planning to connect our neurons to the outside world or to a computer to see in real time how the neurons evolve as they're getting information so this will be our brain in a dish and this is also led by a student in my lab Janmin who is there who is trying to make new materials so we can get better recordings for mice that we are able to feed into our chip eventually related to this way of thinking is the people working on synthetic biology so synthetic biology is the idea that we can hack into the kind of code that makes life possible so we can use, for example, cells as factories of the materials of the future to the manufacturing process that is able to hack into the digital and the analog capacities of cells in order to create materials I said analog although most of the idea behind digital synthetic biology is actually digital and they're mainly looking at genes and proteins perhaps the most interesting of this area is protein engineering I think you had a talk about how these guys deep learning a deep mind could be used alpha fold could be used to predict protein structures this work follows the work of alpha fold follows the work of decades of work of people like David Baker that they were developing the algorithms that they allowed them to actually predict the structures of proteins it's not just machine learning in order to create these algorithms the key is not the algorithm of machine learning but it's actually to take into account the evolutionary history of the protein that allows you to make a computation that correctly falls the protein to the correct shape the most interesting thing of this is that these guys not only stayed in predicting protein structures what they're doing now almost immediately after they're able to do that is to create proteins in the computer that they don't exist in nature go back to the cells like in artificial like in synthetic biology approach and get cells yeast or whatever to produce these proteins that don't exist in nature this is proper nanotechnology this is designed in a computer something with atomic precision atoms every atom is in place in this complex structure and you get cells to build them so these guys now are able to produce very complex assemblies that are designed in the computer and the computer program itself is just beautiful from the computational point of view because as I said the key here is not just a normal computation like physicists do it's not just a simulation taking into account the physics you need evolution the evolutionary history of life on earth in order to create these structures and they're so quick that one of the students of the lab that now has his own lab created a vaccine within weeks again against COVID that is already in clinical trials based on this this is proper clever nanotech the total design of a structure all the way down to single atoms and amazing and based on so many clever ideas of biology of physics of computer science I love printing and technology there's such a sample of how things will be in the future another and this is related is the idea I told you before there were evolutionary algorithms that will tell you things in silicon how evolution of things how evolution of a process will happen a bit like the game of life what is going to happen now in the 21st century is that the combination of what we know about evolution and the software we can create and things like 3D printing will allow us to create hardware that is able to evolve the evolution of things so this is again will take us to ways we're not now and we don't even know where we are we will be able to design materials for applications that can be replaced by itself as it needed and it's not so difficult right now even with the technologies we have right now to start creating this then we introduce you to the cenobots Michael Levin is a biologist that have been interested for many years in understanding how biological shape emerges not just from genes but from electricity and he had this idea together with his collaborators of creating the cenobots they create living these things these little blocks are made of tissue engineers so basically you get polymers and you put cells on them they're cells from frogs and these things they create them so it's a little bit like the game of life but proper these little things are there they're fit and for a strange in a very strange mechanism I very much encourage you to look at this paper they're able to self-replicate nobody knows what these things will be used for but right now we have already living toys that can replicate and I'm moving just finally to the next frontier I started my talk with it from Bit from Johnny Wheeler and I think another area where quantum computing sorry with the with the ideas of computing are being totally changes quantum computing as you know the idea again of using nature to creating new ways of computation but there are interfaces between quantum computing quantum devices quantum technologies and biology is the idea that proteins are in that interface between the quantum and the classical there are three nanometers across they usually operate in classical as far as we know but some of them may not be operating just in classical principles especially all these proteins that are interacting with light for example can be suspicious that they might be doing some kind of quantum computation so people are Oxford including Tristan Farrow are trying to look at biological systems for inspiration to trying to find what might be the design of the quantum devices of the future I also think that all these algorithms will help us to identify the proteins that are looking for that might have this quantum computing capacities and my interest here could be to understand lessons from nature lessons from biology to try to remove the noise that will allow us to create a computer at high temperatures biological systems need to control noise they work on thermal fluctuation and yet they perform tasks that go just over these thermal fluctuations I also have the suspicion that in some biological systems biology might be able to freeze enough the thermal fluctuation so that quantum systems quantitatively happen many of the systems that are in nature that are suspected to be quantum are actually crystalline in the biological structure such as this one bacteriodopsin in which I work for and also magnetoreception rods in the eyes of birds etc so to finish this of course is creating computing and creating materials and creating stuff out of inspired biology but what all these leaders is actually the underlying principle of all this is that there is a value in looking as nature as the computer the capacity of matter to the idea that we can look at nature from an information perspective is very promising for people like me as I showed you at the beginning my plant that turns into the sun I really think that the only way to integrate all these different visions of biological processes is by integrating digital and analog non-equilibrium thermodynamics and information theory yeah this is what I just said without the picture so of course this bit my plants and what I'm trying to do just to finish I'll tell you that there's nothing new here physicists have always try to understand why we understand nature and it's been a combination of social forces and technological forces that are completely pushing us into this idea of understanding the capacity of nature to create intelligence so the line over here is a bit of a history of physics in the 20th century and at the bottom part is a bit more the mathematics and the information theory so if you look at this line even though as physicists we don't we are not told very often to think that we are looking for complexity and intelligence in fact this is I argued what we've been looking for from the beginning so we started from elemental particles and quantum mechanics and just looking at particles but very early on we started to look at how individual particles interact with each other and emerge into things like superconductivity like left and down ideas they started in Harkiv as they're now being bombed it's very sad to think I've been shooting in 1944 his first idea of how non-eclimate thermodynamics will lead to life he made a prediction of the ex sustains of DNA that then it was proved by again by techniques involved by developing physics department X-ray diffraction he's Rosalind Franklin looking at the structure of DNA the first things we turn our microscopes to is life and we are obsessed in where are we made of of course Jean-Pierre de Genne with polymer physics the idea that you can create materials for soft matter physics and as I argue today in my talk now this is linking with information theory the idea that soft matter and nano-scale soft matter polymers actually link non-eclimate thermodynamics with information process as soon as the computer as the transistor was invented people use it for complex systems fractal chaos theory which linked to non-eclimate thermodynamics the work of Ilya Prygoshin Lars Onsager trying to understand how energy I use his principles actually to understand how plants grow and at the same time all these guys are starting from goodness and completeness theory the idea of what can you learn from a logic system and what is outside your logic system inspired Alan Turing to try to make a computer to understand what can you understand with the machine and what kind of machine knows and as I've been telling you today through my lectures in the Manhattan Project already from Neumann standalone in here in the middle we have Richard Feynman may look in somewhere else you know but engage with the computer discussion Claude Shannon linking a structure entropy and computing and we finish here with the lecture of Johnny Willer in 1889 in the Santa Fe Institute of Complex Systems trying to link information physics and quantum the search for links here's where we are now in biology we're just how many years 14-30 years behind Johnny Willer trying to find our it from bits in soft matter physics so I think I'll finish here I just tell you one beautiful example again of biology and how other people are trying to do it so Toshiyuki Nakagaki from Hokkaido University and started a few years ago to interrogate how molds compute so there's a search and research on this thing Fissarum polycephalum which is a mold that rods would and this mold is found to create amazing networks able to solve very complex geometrical computational problems looking for food so his work is inspired a lot of experiments and computer scientists are starting to try to understand how fungi are able to create actually have a language an electrical language they produce this oscillation similar to brains to communicate with each other so this is a beautiful paper that came recently demonstrated that the fungal world lengths are much that of human languages we're just starting to discover the world of computation in biological systems as I told you before and this is my final slide I think with the right framework of research a truly transformative way in which we do technology and we understand nature will come about not just with machine learning but with all this by inspired computing tools that I briefly or too long describe in my talk today and I finished with Johnny Wheeler my hero of my talk and one beautiful sentence where he's playing his insight into the idea that nature creates matter creates intelligence out of a computation or maybe I just read the last bit reality arises in the last analysis from the posting of a yes and no question and the registering of equipment evoke responses here's the analog part of Johnny Wheeler in short all things physical are information theoretic in origin and that we live in a participatory universe and with that I ask for your participation questions thank you very much for this very interesting talk so indeed visual clapping for our speaker and thanks again for bringing a subject that we do not often see in these seminars and very interesting this year this different perspective so if anyone has any question do raise your hand or don't hesitate to ask it already there's no one else they're already bored usually at first this is a bit more sorry so I actually I have a very nice question because this is really not my field but you mentioned the user neuromorphic chips I think for example and you mentioned that these different sub elements would be able to compute together but you then also mentioned the parallel with the human being where you had the appearance of large behaviors such as emotion which are probably I don't know if they directly intended in the machinery but they do appear and I was wondering if these biologically inspired machines do not suffer from problems of stability and control and do not have this image of uncontrolled behavior well of course well depending on how you're going to implement it the biological ones are very difficult to control and all these people doing synthetic biology in one way or another they're very difficult things to make so the difficult things to think about the difficult things to understand yeah they're not easy to implement I don't know myself much about neuromorphic computing I just implemented with semiconductor devices I know there's a lot of interest and the idea that they just mimic architectures in brains of some sort or imaginary brains in order to produce tasks a little bit like what I was showing you about the worm I get this is all still very preliminary in many cases not because they need to evolve the computer system together with the application that's why I think drones is one of the areas that they will use it yes I guess they do have problems of stability and I guess these are the things they're trying to tackle I think it's all at the moment very trial and error where things are going on but maybe someone in the audience knows more about this than me thank you that's also a very interesting this is probably a very naive question as well but is there well given that you're saying it's hard to control these do we have any concerns about producing things that are self-replicating yeah I think in the case of the center bots you wouldn't because as soon as you don't fit them they will die and they are difficult to keep alive the things I mean they're all in a pot they're tiny and then they need a lot of food to keep alive and they need temperature and this is not going to go anywhere I will I do have ethical issues and we will have ethical issues very quickly so especially in things like where I will try to do things like you know neurons in a pot that are starting to interact with the virtual world well are we making there in as long as it's just that it's fine if we put them in a robot I think we have a huge amount of ethical issues coming up there's no way to do this research without going away from that and luckily because of medical technology actually we have quite good research responsible research and innovation protocols in place they were both developed by the European Union and now implemented anywhere but increasingly so in this area of cyborgs we're going to need to have very strict understanding of what we want to make I don't think we should make whatever we want or what is possible but what we think is a good idea a good idea to make Tristan has a question I'm on mute and on video myself thanks for that very interesting talk I thought it was a great sort of vision overview of fields I didn't even know existed to be honest it's more a comment what do you think are the barriers to entry so at the beginning you said that we need to bridge this divide between the analog and the digital worlds and I couldn't help noticing in one of the penultimate slides that most of these people were theoretical physicists apart from Rosalind Franklin I think and from my own experience trying to bridge this divide I find that talking to say biologists or biologists talking to physicists we don't necessarily speak the same vocabulary so it might have been okay to design these basic ab initio principles for you know Godel and During and all these guys but nowadays I think you do need some knowledge for me of biology and vice versa physics for the biologists so I was just you know it's more as a sort of comment than a question but I was wondering if you had any thoughts on that yeah I do have because I'm a biological physicist and first thing is that there's a lot of physicists that actually know no biology like me and they work with living systems biology has moved to the physics department and that will facilitate that interface many of us are used to work with biological systems who can grow cells and we can work with biologists more or less I think is even more in a way I think what we do now is that in in physics labs we do analog digital so we do an experiment trying to understand a physical phenomena an experiment and the experimental equipment is an analog computer that is trying to extract the physics once we have that experiment we extract the physics in a way we can code and then we go into the simulation of that phenomena I think what we're trying to moving forward is integrate it all which is increasingly how we are all working so the experiment the analog experiment biological or quantum or quantum crystal or is linked to the to the digital model in itself I think we are already doing it but we do it in separate places or maybe in separate people you have the theoretical physics or the simulator physics and then your experimental physics so it's all put all together in a chip this is a bit what we are trying to do here which is a bit what I'm trying to say the future of computer is something like the hardware and the software is so the experiment and the simulation is in your chip and you can I don't know how we are going to implement that I gave you some ideas of how things might be looking but I think that's what we are already doing we are doing that so for example when I have when people are trying to understand tumors the idea was to look for the genes and all that that doesn't work and then you have engineers that make two more models in a dish or you can make organoids that make the two more model in a dish that's your analog computer so now the task is to extract from there the information which is what is missing so all these analyses are based on either the biological way of seeing things which is very basic network complexity network or the physics continuum way of doing things so I think it's the bridging of the two which I think is what you work in conceptually linking information theory with non-equilibrium thermodynamics the continuum and the digital is the big challenge so that's why I think this has to be done in a physics department at least partially anyway okay enough good I don't know if you still have a question just in your hand is still raised sorry I just forgot to put it down just maybe I'll make you can make another comment which is that on the bio-inspired computing side what interesting point is that some of these for instance if you take DNA base pairings the amount of energy per base pairing which you can in this paradigm look at as a logical step or logical gate is about 100 kT unit of energy and if you look at classical computers each computational step is roughly of the order of 1000 to 10 000 kT so I think that'd be if it's possible a great advantage in mimicking or building computers out of say DNA or organic systems that are much more efficient and that would solve some of the issues that we encounter with heating for example in in class of computers at the moment but anyway that's again more of a comment than a question there's a lot of there's a lot of things in which we can we can learn from biological systems that the problem will be to get the right people talking to the right people and having the right structures that allow people to create those computers I think their ideas are all there and as I showed today the ideas are there from the 1940s with nanotechnology means that now we have the means to manipulate and observe and create computers for example with molecules and now we even have physicists understand biology so I think all the pieces of the jigsaw are there is just the willingness or the capacity to to actually make some of these things happen thank you interesting I actually had a question myself related to this because I come from high energy physics where we obviously have the CERN which is a really large and Fermilab and many different very big national labs or COO or international labs where people collaborate on these things and I was wondering if this fields also has such labs or it's more shared between universities and private lab the main problem of these fields that nobody talks to anybody and this is the reason I made this talk nobody will tell you this story because nobody talks to anybody the biologists have their visions the physicists have their visions the engineers have their visions I think for me computing actually put all this research into the framework or computing from a physics perspective will give it will is the only way to really tackle big medical problems and also big complexity problems and so my contribution if you want is to put all the story together and at least put all the pictures together so people see they are related and that the biggest breakthroughs will come from understanding biological and complex systems from physics of non-equilibrium and information theory but we don't collaborate but we don't collaborate we need a CERN of biology that'd be interesting yeah yeah I think that's clear we need a CERN of biology good to know so that's the list of organizations from the particle physicists maybe that's what we need here yeah well that sounds interesting indeed I don't know if anyone else in the audience I want to ask her last question yes David David Hi Sonya, thank you for your talk I was just wondering I mean your the topics that you mentioned of information theory and physics it's a it reminded me of work by Carl Friston on free energy principle yeah I'm not sure if you've heard of this but I'm wondering if what relation your work has to it does it interface with it at all or I don't know what is the free energy principle um well essentially it's it's a framework that he uses to describe the organization of biological systems where a or essentially a biological biological system will have a predictive model of the world and it tries to take actions in order to minimize its sensory prediction error yeah I guess this is similar to the I didn't discuss this they're to the yeah so there's several ways in which people understand this so you you do it from energy um Kolmogorov Uspensky which are Russian mathematicians and actually Art Louis in his talk was talking about Kolmogorov complexity the idea that algorithms are always going to look for the a simple the smallest Kolmogorov complexity but also that biology might be implementing algorithms um that try to minimize things and and lead to symmetry not only information not only structure but also energy I think again many people have thought about these issues from different perspectives from energy perspective and from information perspective um and and it's our task now I think to put all this information together to see if we create a more profound picture of information processing in biological systems is is is difficult because it is difficult maths and it's difficult physics but I think this is where we are now and I you can see more and more people trying to unify these things for me the main problem is fragmentation different fields don't talk to other fields and it's going to be very difficult to make progress if we don't talk to each other okay thank you yes thank you very much and I think it's probably time to close so maybe we can give another virtual round of applause for our speaker thank you very much for this super interesting talk and have a good afternoon everyone thank you so much and yes to just to answer the question that I answered recently we will share the talk in the normally tomorrow during the day okay so it would be made available on Friday thank you very much have a good afternoon thank you very much for coming to see me thank you very much bye bye goodbye