Viewpoint: FP10 should fund AI for peace, ageing – and European leadership

09 Apr 2024 | Viewpoint

For too long the EU has been chasing – too late – after US Big Tech. With the next Framework Programme, Europe should get ahead of the game and harness its own creativity

Afonso Ferreira is research director at the CNRS, working at the Toulouse Institute of Computer Science Research.

What kind of AI research should get funded in the EU’s next R&D programme, Framework Programme 10? When pondering that question, I thought it only fitting to reach out to a Generative AI system for help. Unfortunately, it had never heard of FP10.  So, here is a text I produced myself.

Though I’m not made of silicon, I do have some relevant experience to offer an opinion. I was coordinating my first FP project in the early 1990’s, in the beginning of a career that experimented with both sides of the European funding counter - as a researcher at France’s CNRS and as a programme and policy official at the European Commission. Based on this and later jobs, I believe that my main recommendation for FP10 should be to avoid running after what the US technology firms just did. Instead, the EU should try to fund what they’ll be doing next. Or after that.

Let me give an example of how the EU should not do it. At the Millenium, when Google was rolling out its world-changing search engine, the Commission was busy funding Framework projects in circuit-switching networks, thanks to the lobbying of the telecommunication incumbents from some large EU member states. At some later point, the EC woke up to the fact that the digital future would be based instead on packet-switching and something called the Internet Protocol. That’s when EU leaders decided that Google had to have a European competitor and started calling for projects on – wait for it - search engines.

But by then Big Tech had moved on. They were building very large digital infrastructures in the form of huge data-centres to collect all the data they could and then some more. IBM, Google, Facebook, Amazon, Baidu, Tik-Tok and others began using such data to advance the scientific state-of-the-art in machine learning and other fields. And the EU by that time? It was proudly funding projects on digital infrastructures and data-spaces.

Now the buzz is about generative AI. And my conclusion: No. FP10 should not put all its AI funds in the belated development of a European counterpart. Instead, let’s think ahead of the curve. It’s time for peace, drugs, and rock&roll.

AI for peace

For instance, we could start with a focus on peace. The EU currently has two wars raging on its doorsteps. What about funding projects that aim at using AI for research and innovation that foster peace amongst peoples? After all, there is a great deal of data about warring and peaceful periods in history and their dynamics. There are also plenty of theories of war and of peace, as well as studies about the origins of conflict.

Hence, FP10 should call for clever ways to use AI to solve optimisation problems that try to minimise the probability of war or maximise the probability of peace, based on parameters like dialogues and other conflict-avoiding practices and new ones yet to be discovered. Certainly, this won’t be easy, as they are multi/inter/trans-disciplinary by default; and finding good performance indicators is surely a challenge. But I’m sure that it’ll be easy to find many experts willing to contribute to such a distinguished goal. With a nod to John Lennon: All we are saying is give peace a chance.

Drug development is another area that can benefit a great deal from machine learning tools. Humanity stands to benefit even more from new drugs that can be developed more cheaply and quickly. As Google likes to advertise, modelling how proteins fold is a very difficult problem. Out of the roughly 200 million proteins, only 170,000 had been modelled, mainly by very clever cohorts of PhD students, by the time Google’s Alpha Fold was developed. Then, Alpha Fold displayed a level of accuracy so high in proposing proteins’ structures that the community considered the protein–folding problem as solved. Now, again according to Google, there are 1.7 million scientists working with Alpha Fold and more than 18,000 citations to it in the scientific literature.

Of course, large drugs-related databases that can be used for training new models already exist, but they are mainly the property of Big Pharma. Incentives would be needed for those companies to cooperate with each other, and with researchers from academia. And there are some high-profile, high-reward fields in drug development in which AI could change our lives, such as ageing research. FP 10 should have a programme on AI for drug development, with a strong line of funding for projects related to combating ageing and increasing human lifespan in good health conditions. This is less about who wants to live forever than it is about being forever young.

Let a thousand projects bloom

Finally, the EU should ride the waves created by the democratisation of GAI models that happened since OpenAI unveiled its GAI in November 2022 and since Facebook’s GAI model Llama was leaked. Today everyone and their neighbours can develop their own GAI models on laptops. This is the power that’s begging to be harnessed at the level of a Framework Programme.

With these tools, FP 10 should call for myriads of small, target-based AI-powered research projects. It should have both topic-agnostic calls, leaving the ideation of the application areas open to the applicants, and calls with very ambitious pre-determined grand topics. Besides the peace and pharma challenges just mentioned, these big themes could be in biology. For instance, what makes a collection of liver cells behave like a liver? What is consciousness? There could be big projects in societal challenges: How to save humanity from collapse brought on by climate change, while still growing the economy? An approach similar to that of venture capitalists would be ideal: Many projects running for some 42 months and funded at a maximum of €2 million, well synchronised to coincide with three-year PhDs.

With FP10, let’s start an AI revolution in science. For those about to rock the European research landscape, we salute you.

Afonso Ferreira is research director at the CNRS, working at the Toulouse Institute of Computer Science Research. The opinions expressed here are his own, and do not necessarily reflect those of his employer.

Never miss an update from Science|Business:   Newsletter sign-up