Interview with Professor Morten Meldal on academic freedom and his idea about creating artificial intelligence especially for research.
In 2022, Prof. Morten Meldal was awarded the Nobel Prize in Chemistry “for the development of click chemistry and bioorthogonal chemistry”. Click chemistry means that molecular building blocks snap together quickly and efficiently.
Meldal, a professor at the University of Copenhagen (Denmark) and head of the Center of Evolutionary Chemical Biology, shared the prize with Carolyn Bertozzi and K. Barry Sharpless; in 2002, Morten Meldal and Barry Sharpless, independently of each other, developed an elegant and efficient chemical reaction: the copper catalysed azide-alkyne cycloaddition. This is now in widespread use and is utilised in the development of pharmaceuticals, for mapping DNA and creating new materials.
“We as researchers are responsible individuals who make results, and if we didn’t, let’s say within the last three years, publish anything or didn’t do any activities proper, we have to find something else to do. Because this is a very competitive occupation”, said Prof. Meldal.
Science should be done without constraint – that’s what you said in your lecture at the 72nd Lindau Nobel laureates meeting. What did you mean?
Morten Meldal: I think that we should just specify that a little bit. Science is categorised into applied sciences, into fundamental sciences and strategic sciences. Strategic sciences there could be, for example, something within the environment that we really need to do strategically in the society in order to be avoid some big problem in ten years ahead, let’s say.
That’s very goal oriented. So is the applied part where you actually take something from one of the other boxes and make it an industrial application. That is also targeted. In the fundamental box, that should probably be the largest one, which is not, but that’s where we learn everything. That’s where we get all the new ideas and all the new possibilities for companies and so on.
That box needs a certain amount of funding, but it doesn’t need control over what the funding is used for.
The reason for that is because all the big discoveries more or less serendipitous. Because if you could actually plan behind your office desk, you could plan fantastic research. It would have been done before by somebody else in the world.
All new discoveries are surprises by nature. And therefore we should not constrain research. It cannot be politicians who tell us what the next discovery is. It must be the researchers themselves.
How do you see this freedom of academic research preserved? How can it be done?
Morten Meldal: I think that we have a perfect system actually in place to do that: the peer reviewing system. And with the accounting systems for our teaching, the dissemination of different kinds and so on – all of that you can probably make a good arithmetic formula with that tells you how much money you need for the next couple of years.
The most important thing is that we as researchers are responsible individuals who make results, and if we didn’t, let’s say within the last three years, publish anything or didn’t do any proper activities, we have to find something else to do. Because this is a very competitive occupation.
The GDP of the country and how it is distributed is a very important thing, too. In my country, in Denmark, we are extremely lucky and we have a very high taxation rate I think. I guess this is what I like about Denmark. We pay a lot of tax because we have a very good health system and research funding and all. Denmark is actually a rich country in spite of the fact that we pay that much tax. And the taxation system has also allowed us to build some very large foundations which are philanthropic today and really support basic research. I think that’s a very important thing in Denmark.
Artificial intelligence was one of the main topics at the Lindau Nobel laureates meeting. It is something that will dominate the future? Some people are afraid and suspicious of it.
Morten Meldal: It’s clearly very useful. But it is also very dangerous. It is a tool with an inherent danger because this tool can be used for any kind of very rabid handling of a lot of data. And it can be muscled into really giving you a particular result which you want. But it can also be allowed to do things on its own. And that’s a bit of a problem when you have something like large language models that can actually create a code by themselves.
I’m sure there are many, many places in the world where they are working on enabling these programs to make sort of logical decisions. These logical decisions can be logic based on our understanding of logic, but if they’re able to program themselves, they might be able to take over that logic and decide what is logical from their point of view.
It’s very useful for defense purposes to be able to overview all the different critical issues and take a decision. And it is very useful in particular in science, because science is generally about data, it’s generally about filtering data. And a good scientist is the one that is good at filtering data from the literature. We would filter through all the information that we know is there in the literature. The more the more knowledge we have, the better we are filtering.
But there’s no way any scientist in this world can cover it all anymore. When I was young I knew all the people in my field and what they have done.
ChatGPT is not the right kind of thing for science though, I think. We will have to develop special artificial intelligence that handles all the information in a different way, because in science there’s a demand for truth, ChatGPT doesn’t fulfill this.
Therefor I think that we have to develop a particular artificial intelligence that deals with science, similar like you have with Google and Google Scholar. It could be one tool with different options to select or ask questions. You could ask questions in different ways within the same tool.
Should there be a regulation about what AI could and should not do?
Morten Meldal: You cannot decide what AI should not and should be allowed to. It is too late, it cannot be done. How would you regulate something which is happening on everybody’s computer around the world?
In general I think it’s too late to actually establish regulations for AI. So it is a real danger in my opinion.
What about some ethic codes for the science community regarding the use of AI?
Morten Meldal: Yes, that can be done, and it has been done before in other areas. And I think we should do it definitely – to find good codes for operation in all types of computational approaches.