The extent to which academics are likely to follow the Nobel Prize Committee's media attention, money and critical acclaim is a question that Julian Togelius, an associate professor of computer science at New York University's Tandon School of Engineering, who works on AI, irritates. “Scientists generally follow a combination of the path of least resistance and the most bang for their buck,” he says. And given the competitive nature of academia, where funding is increasingly scarce and directly linked to researchers' job prospects, it seems likely that the combination of a trendy subject that – as of this week – has the potential to land top performers a Nobel Prize . It may be too tempting to resist.
The risk is that this can get in the way of innovative new thinking. “It's difficult to take more fundamental data from nature and come up with new theories that people can understand,” says Togelius. But that requires deep thinking. It is much more productive for researchers to instead run AI-enabled simulations that support existing theories and engage existing data – producing small leaps in understanding rather than giant leaps. Togelius predicts that a new generation of scientists will do exactly that, because it is easier.
There is also the risk that overconfident computer scientists, who have contributed to advances in AI, will see AI work rewarded with Nobel Prizes in unrelated scientific fields – in this case physics and chemistry – and decide in to follow in their footsteps and encroach on the borders of the world. someone else's territory. “Computer scientists have a well-deserved reputation for sticking their noses into fields they know nothing about, injecting algorithms and calling it progress, for better and/or for worse,” says Togelius, who admits to having been tempted to do something before adding deep learning to another field of science and “advancing” it before thinking more about it because he doesn't know much about physics, biology, or geology.
Hassabis is an example of the use of AI Good to advance science. A neuroscientist by training, he received his PhD on the subject in 2009 and credits that background with helping to advance AI through Google DeepMind. But even he acknowledged a change in the way the industry is achieving efficiency gains. “Today, [AI] has become more technically demanding,” he said at his Nobel Prize press conference. “We now have many techniques that we only improve algorithmically, without reference to the brain.”
That could also influence the type of research that is done – and who does it, their level of knowledge in the field and the incentives that keep them going into it. Instead of researchers who have dedicated their lives to a specialty, we could see more research by computer scientists, disconnected from the reality of what they are looking at.
But that likely takes a back seat to the celebrations for Hassabis, Jumper and the colleagues they both thanked for helping them win the Nobel Prize this week. 'We're almost done cleaning up the [AlphaFold3] to release code so that the academic community can use it freely,” he said earlier today. “We'll go from there.”
