Hi ha un article provocatiu al blog Marginal Revolution, titulat The future of the scientist in a world with advanced AI, on s’hi planteja que…
AI will know almost all of the academic literature, and will be better at modeling and solving most of the quantitative problems. It will be better at specifying the model and running through the actual statistical exercises. Humans likely will oversee these functions, but most of that will consist of nodding, or trying out some different prompts.
The humans will gather the data. They will do the lab work, or approach the companies (or the IRS?) to access new data. They will be the ones who pledge confidentiality and negotiate terms of data access. (Though an LLM might write the contract.) They will know someone, somewhere, using a telescope to track a particular quasar. They may (or may not) know that the AI’s suggestion to sample the blood of a different kind of gila monster is worth pursuing. They will decide whether we should be filming dolphins or whales, so that we may talk to them using LLMs, though they will ask the LLMs for cost estimates in each case.
At least in economics, this will be continuing trends that were present before current high-quality AI. The scarce input behind a quality paper is, more and more, access to some new and interesting data source. More and more people can do the requisite follow-up technical work, though quality variations have by no means been eliminated.
“Science as an employment program for scientists” will fall all the more out of favor. It remains to be seen how much that will disfavor serendipitous human discovery.
On any given day, on the quest for more data, a scientist will have to think quite creatively about what he or she should be doing.
Les respostes i comentaris són interessants, algunes molt apassionades. No queda pas gaire clar quin serà d’ara endavant el paper dels científics. Però sí que sembla les grans idees, la intuïció… seguirant essent característiques rellevants de les persones. La IA, però, proporcionarà idees novedoses donat el seu gran abast de coneixement… si és que és capaç de tenir prou coneixement.
Ja ho veurem, potser sí, o potser no, el paper de la gent de ciència serà més important.
M’ha fet gràcia el que proposa un comentari: situar la IA l’any 1905, per veure si descobreix les mateixes coses que l’Einstein:
Let’s do an experiment. Let’s train an LLM on all possible knowledge up until 1905 and see if the LLM can come up with Einstein’s theories.
Una mica seria com els models meteorològics que fan servir dades anteriors per millorar la seva capacitat predictiva (i que, al menys per aquí Girona… encara fallen força).