Multi-disciplinary Revolution in Biology and Artificial Intelligence
Michael Levitt
2013 Nobel Laureate in Chemistry;
Honorary Dean of the Multiscale Research Institute for Complex Systems, Fudan University
Thank you so much for inviting me here on the occasion of the 20th anniversary of the Shanghai Forum and the 120th anniversary of Fudan. My talk is about a triad that really controls everything. Surprisingly, this is not well recognized.
On earth, we have three intelligences: biological intelligence, human intelligence, and machine intelligence. Biology is the smartest of all, because it made humans. And then humans make AI. So let's first say, what can we learn from each of these intelligences. So biological intelligence really teaches us a very important lesson, and it teaches us through evolution. Biology is really smart, as I just said, it can make amazing molecules better than any chemist. It can make tissue. It can make five fabrics. It can make materials. It can make hands, neurons, eyes, brains, everything, and it's all made from scratch. So basically, biology really started from very little about 3 billion years and slowly became more and more advanced until at the top of the tree, we have humans and higher mammals and so on. What is the secret of the complexity? If you look at the very bottom, life started out very simply. The key difference between simple life—bacteria that you can never see—and all the complexity you see around you is very simple. Two parents, a male cell, and a female cell that give rise to offspring that are 50 % different from them. Bacteria clone themselves. So a really fit bacterium makes all his kids just like him. That works fine as long as the future is the same as the present. The trouble is the future is never the same as the present. And the secret of life is basically not survivor of the fittest, but survivor of the most diverse. This also explains why we got more and more intelligent. A person is more diverse than a tree. If there's a forest fire, the person can run away. The tree has no alternative. So as life has become more advanced, we've become more diverse. So basically nature awards creativity, and in biology, mixing male and female cells leads to diverse life forms.
Let‘s now movee to human intelligence. And one of the really important lessons about human intelligence is that other humans are great role models for us. We all learn not so much from what we are taught, not so much what we read, but of the people we meet, of people who have been important in one way or another. I was extremely lucky in my life. Before the age of 20, I was really meeting some of the greatest scientists of the last century. First and foremost, was a man called Max Perutz. He's not that well known. A Jewish refugee from Austria who came to England and basically transformed biology. He was somebody who was working on the crystallography of a hemoglobin molecule shown there. He had an assistant called John Kendrew, who was the first person to solve the structure of a protein. A student of Max Perutz at that time was Francis Crick. Now, just think about this. Crick with James Watson wrote the most famous scientific paper, probably ever—the structure of DNA. He wrote this paper while he was the student of Max Perutz. But Max Perutz, his name is not on the paper. That is something that would be unthinkable today. But Perutz was operating in the scientific vacuum created by the Second World War in England in the 50s, and basically realized that the way you create great science is by giving young people independence and responsibility. Inspiration is contagious and meant to shape future scientists. So you want to let young people meet those who changed the world.
My heroes used physics and chemistry to reveal the secret of life. The secret of life was completely unknown until the 1950s. So we don't know the secret of life even for 100 years yet. And in the last 80 years, we had nuclear weapons invented, but we also discovered the secret of life. The information in the DNA sequence becomes information in a protein chain. The protein chain folds up. That gives the protein function. It can interact with the real world. It can perhaps digest food. If the cell can digest food, the cell multiplies, the DNA increases. That's how we have all life. DNA is a molecule that contains information in its sequence of nucleotide. In our body, everything is defined by a sequence of about 3 billion letters—that isn't very much. My hard disk could contain many thousands of human sequences. 1 billion is only a gigabyte, not very much. The important thing about nature is that the protein necklace isn't just a long chain. It actually folds up into a compact shape. It folds into a compact shape that has precise three dimensional structure. Our body has little machines in it that are smaller than the thinnest wire in the best nano technology we can make. This molecule is about two nano meters across. The best chips have wires that are 5 nano meters across. In our body, we have billions of machines all interacting with each other, making life possible. So basically, the secret of life is self-organization at the smallest level. We also have self-assembly with the laws of physics and chemistry are driving the process. And again, that is something quite amazing.
We defined that great institutions actually also win many Nobel Prizes. Max Perutz went on to form an institution in Cambridge that ended up getting 28 Nobel Prizes. Now, 28 is a big number. I think Japan has 28 Nobel Prizes for a population of 100 million people. The total number of people who went through this laboratory in Cambridge was maybe 3,000 who won a hundred honorable Prizes. I was one of the lucky people to be there. And this is all due to Max Perutz, who said, famously, in science, truth always wins. This is a very important thing. And the reason truth wins is that everything is checked. Everything is rechecked. In fact, in science, nothing is true until proven otherwise. So everything is assumed to be wrong until you prove it right.
Perutz had a recipe for Nobel Prizes. He never said the same thing. But as a student in the lab, I was able to understand what was his genius. So firstly, make young scientists independent and responsible. Have small groups. Don't give a famous scientist all the money. Finding big discoveries in science is like winning a lottery. And the only way you can win the lottery is by buying more tickets. So it's much better to have a hundred scientists, each with a group of three or four people than 300 scientists working for some great genius. The research groups there had ample support. No one had to apply for a grant. You wanted anything, you just took it. No visible bureaucracy. Perutz felt that a young scientist should never have to fill in a form, should never have to order equipment. If he wanted it, he just went to the workroom and said, give it to me. There was also intense peer pressure. You are only as good as your next paper. Everything you did in the past is irrelevant. What are you doing now? And there is also very little hierarchy. PhD students there felt as good as Nobel laureates. Great leaders inspire lots of great scientists. We really need to have great scientific leaders.
Let me move on to AI. I'm defining AI very broadly. AI is computers making us more intelligent. My very early work in basic fundamental computational biology with a form of AI started when I was 20, including working out the forces between atoms and molecules, working out how proteins could fold up, and also working out how enzymes could catalyze reactions. All of this work was done before I was 27 years old. Great science is done by the very young. It takes a long time for it to be recognized, but we need to basically focus on young people for science.
Science is also applied. While I was working in Cambridge and in Israel at the Weisman institute, I was doing basic science. But then in 1987, I moved to Stanford where it was almost compulsory to have a startup company. The first startup company I joined was a company called Protein Design Lab in California. And in some ways, working for this company, I probably did some of my most important work. Although I never really thought it was very important by working on antibody therapy for cancer. A paper published in 1989 is a humanized antibody that binds to the interleukin 2 receptor. Cancer cells have certain receptors on their surface. If you can have an antibody that attacks the cancer cell, you can suppress the cancer very much. Trouble is antibodies are made in mice or rabbits. And if you inject antibodies from mice and rabbits into people, they get sick. You have to do a makeover, a computer operation where you take the mouse antibody and make it look human like. There are different ways of doing this, but in this company, protein design lab had a patented technique. Based on this paper, this led to a big drug company. I think they spent $200 billion to prevent monoclonal antibody drugs. But today, some of these drugs like Avastin have been used for cancer, based on the papers that we were writing 30 years ago.
Another example is much more modern. There's a company called Insilico Medicine that in 2014 started to use generative AI for the whole process of drug design. Their idea was to realize, and it's not very difficult to realize this that the drug design is a long pipeline. You start at the beginning with an idea, a target, then you have to end up with something on the shelf. And if that pipeline gets blocked anywhere, you're stuck. There's no way around it. It's a very expensive pipeline, something like $2 billion to get a drug. They wanted to make this quicker and cheaper. And they did this by using AI everywhere. It wasn't that they used AI for target discovery or for lead optimization. They used AI everywhere. And they really have been successful. After 10 years, they now have something like eight drugs, either in IND phase one or phase two trials. This company is also a very global company. I don't even know where the headquarters are, maybe Boston, maybe Hong Kong. They have a big presence in shanghai and in Dubai. Dubai is very important because all the programmers from Ukraine and Russia work in Dubai now and in New York City. So this is what a modern company looks like. And they've done very well.
In science, there are two paths and one engine. Basic science explores the unknown, applied science builds on the unknown. It's very different. In basic science, everything you did in the past is meaningless. You're as good as your next paper. But in applied science, you have to build things on the past. So you have enough money to continue. AI today, and by “today” I mean 30 months ago, so I have been using ChatGPT 3.0 and 4.0, and whatever is next 4.5 every single day for the last 30 months. I think I've got something like 10,000 inquiries to the engine. And I use it for everything, medical advice, legal advice, psychological advice. I can write lectures like air for agriculture, or air for health. I write all my computer programs using AI, and I'm getting better and better at the AI because you have to learn how to use it. It's a very complicated object. It's much more complicated than the next complicated thing we know, which is our cell phone, even that is very hard to use. So AI actually has very high emotional intelligence. People don't often realize this, but it was shown by a paper from Tsinghua University, where they showed that AI has more high emotional intelligence than 90 % of humans. I joked and said they must have been male engineers. But ChatGPT is as smart in emotional intelligence as most women, which is pretty amazing.
More surprising, AI is the world champion in Diplomacy. Now, I don't mean real diplomacy. I mean the board game call Diplomacy. This is a game where you start it with pieces given out randomly on a map of Europe about 1901. The aim is just to talk to other players and say, I will give you half of Germany, if you give me all of Sweden. The idea is to eventually surround everybody and conquer the world or prevent World War I, depending on what your point of view is. It turns out AI is the world champion. A program written before ChatGPT by Cicero from Meta (Facebook) wrote this program. And the interesting thing is that human players love playing with this person, like it's gonna win all the time. It never cheats. It's honest, but it knows the map so well, that it knows exactly what the value of northern Germany is compared to Portugal. I hope that every single diplomat is asking AI for everything. Ai is not right over time. Ai in fact, is often wrong. But with AI, you always reach an answer better than you would have reached without it. So AI has the answers. You need the questions. Stay curious like an 8-year-old, be wise like an 80-year-old. Keep on asking questions.
My final section is about pairs of scientists. at Stanford, They started a program about 20 years ago, called the Interdisciplinary Initiative. Two people who were in totally different areas were able to get a seed grant if they wrote a 1-page proposal and the seed grant is, I think $50,000. Payers came together. This has been incredibly successful because they went on to apply for real grants based on the initial seed grant. And it brought in more than 10 times the funding on the seed grant. This is a kind of internal venture capital, but the key thing was different people. If you just took two random people in this audience and said, “if you guys can think of something, you'll get $50,000”. Just that conversation. And it wouldn't matter whom, it could be me with the person running audio. We would find some language in context. That would be a new idea. Here's an example of something that I did in 2022. I was working on how molecules moved. Somebody was working on our people moved, and we put them together to understand molecule walking and people walking. And this led to a big institute funded by NIH later on.
I think AI and humans, exciting things happen at the interface. I got AI to draw this from me yesterday. It's not that great. But everything exciting happens when you put together different things. It started out with biology mixing male and female for a diversity of life forms. And in science, interdisciplinary collaboration leads to diversity of thought. And you may well ask, you've talked about everything you did. What are you actually doing now? And what I'm doing now is working incredibly hard to understand what went wrong during Covid. People were convinced it was going to destroy humanity. In the end, it was like a bad flu pandemic, essentially zero excess deaths. In all the countries, we can see it, except the United States and Eastern Europe. In most countries, it looks like what happened in 2010 in Hong Kong, in Taiwan, South Korea, Japan all have essentially zero excess deaths. Now, I am writing these papers. I was actually very criticized for working AI on Covid. I tell people I'm writing these papers now, not for people, but for AI, because the next pandemic AI will be the one that tells us whether we should panic. We need to lead. We need to believe AI, the same way that we believe that when we're planning to drive across Shanghai, we don't say we know the right way, we use the GPS. Thank you all for your attention.
