TechTakes

Speaking at an event in London on Tuesday, Meta’s chief AI scientist Yann LeCun said that current AI systems “produce one word after the other really without thinking and planning”.

Because they struggle to deal with complex questions or retain information for a long period, they still “make stupid mistakes”, he said.

Adding reasoning would mean that an AI model “searches over possible answers”, “plans the sequence of actions” and builds a “mental model of what the effect of [its] actions are going to be”, he said.

wait, you mean the same models that supposed AI researchers were swearing had “glimmerings of intelligent reasoning” and “a complex world model” really were just outputting the most likely next word for a prompt? the current models are just fancy autocomplete but now that there’s a new product to sell, that one will be the real thing? and of course, the new models are getting pre-announced as revolutionary as interest in this horseshit in general takes a nosedive.

LeCun said it was working on AI “agents” that could, for instance, plan and book each step of a journey, from someone’s office in Paris to another in New York, including getting to the airport.

these must be the multi-agent models that AI fans won’t shut the fuck up about now that multi-modal LLMs are here and disappointing. is it just me or does the use case for this sound fucking stupid? like, there’s apps that do this already. this shit was solved already by application of the least-terrible surviving algorithms from the first AI boom. what the fuck is the point of re-solving travel planning, but now incredibly expensive and you can’t trust the results?