The paradox at the heart of AI and science

Un long monologue de Terence Tao sur la productivité des modèles de langage appliquée aux mathématiques. Je l'apprécie beaucoup, car il a une certaine faculté à clarifier des sujets compliqués, que cela soit des mathématiques pointues ou ici son usage raisonné et éclairé de modèles de langage dans son travail de mathématicien.

⊕We are seeing now in programming that many expert programmers  are reporting that their ability to write code has increased by a factor of five or 10 or 100 with these tools but they are also they can feel themselves learn losing the ability to code by hand and sometimes they cannot review the code that that that comes out of these  agents. There's a trade-off you know speed and is not everything. I very much like the collaborative aspect of mathematics. I didn't realize it was so important until relatively late in my career that, like a lot of the mathematics I've learned, I learned after grad school by working with mathematicians and scientists in different fields. I teach them what I know, they teach me what they know and I become much broader as a result. When you're working with a collaborator that you've been working with for a long time, there's a point where you become almost mentally attuned.

You know that we'll be able to optimize everything that we can do with our current technology, but we may not actually develop really original new ideas anymore. We will need to really have a much more open discussion about what basic science is and what is useful for, and why it's important to still have curiosity-driven research. Why we still need a community of humans to explore things sometimes slowly, sometimes, you know, in ways that are not as efficient as the latest model AIs and also the insights that we gain.

Étant moi-même réfractaire, voire hostile, à ces dispositifs simplement pour toutes les externalités négatives qu'ils engendrent. J'essaie de rester ouvert aux personnes qui m'en expliquent leurs usages tout en me retrouvant souvent à froncer les sourcils à cause d'une forme d'aveuglement, de déni ou de dissonance cognitive.

Je me pose donc la question de pourquoi son discours me semble acceptable. Est-ce que je ne suis pas en train de me laisser séduire pas son prestige et sa position d'autorité charismatique dans le champ scientifique ?

À ce titre, cette vidéo me semble importante, car il sait à la fois de quoi il parle, il ne rechigne pas à expliquer les détails techniques en passant, tout en soulignant les impasses vers lesquelles tout cela semble mener. Il n'est pas dans une sorte d'optimiste forcené ou d'un usage désespéré. Il énonce très clairement les risques, notamment la nécessité vitale de former les prochaines générations d'humains.