3 ms·
It seems more like they used a classifier to evaluate the situation and decide what state the system should be in. Not sure that the LLM contributed that much.
by Animats 3y ago
It seems more like they used a classifier to evaluate the situation and decide what state the system should be in. Not sure that the LLM contributed that much. LLMs are mediocre planners at this point.
- hlfshell 3y agoYes and no. LLMs are definitely struggling with long horizon planning in complex situations, and can hallucinate/misunderstand environmental states such that they can recommend nonsensical or impossible steps. But we are starting to see more research that "grounds" LLM answers in a set state. Where they shine in the planning stage is they provide a sense of contextual heuristics we didn't have available before without a lot of expert domain programming. LLMS in their current state are certainly not the answer, but point to some exciting possibilities that we couldn't do before because of sheer complexity.
- YeGoblynQueenne 3y agoThey're trying to get it to do hierarchical planning. People (like Yann LeCun) have complained that nobody knows how to do that in AI, only when they say "AI" they mean "neural nets" and ignore planning -as in Automated Planning and Scheduling, the discipline that has its own conference and journal, and international competition and thousands of people working at it, and dozens of systems, some commercial, some free, and all; but I guess that actually works so it's not AI. Less sarcastically, automated planners need domain knowledge to break up a task into sub-tasks, and the neural net folks don't like that because they are taught at ML 101 that everything shall be a) end-to-end differentiable and b) have no human in the loop. LLMs seem to encode some kind of domain knowledge, the "common sense" that the article's title mentions, and so there's various teams trying to get robots to follow instructions with LLM-in-the-loop. A year ago I think it was LLMs generating PDDL, now it's natural language labels that obviously need a translation. This won't work though a) because it's trained by imitation learning and nobody cares about a robot that can scoop marbles from bowl to bowl but would need re-training to do the same thing with e.g. strawberries and b) the LLM itself isn't grounded anywhere and has no way to self-correct its self-correction, once it falls into error. This, too, shall pass. Btw, are those people mad? Marbles? Couldn't they find something less painful? What happens when your robot overturns the bowl and there's marbles all over the floor of your lab?