←back to thread

Ask HN: Any insider takes on Yann LeCun's push against current architectures?

385 points vessenes | 4 comments | 10 Mar 25 19:41 UTC | HN request time: 0s | source

So, Lecun has been quite public saying that he believes LLMs will never fix hallucinations because, essentially, the token choice method at each step leads to runaway errors -- these can't be damped mathematically.

In exchange, he offers the idea that we should have something that is an 'energy minimization' architecture; as I understand it, this would have a concept of the 'energy' of an entire response, and training would try and minimize that.

Which is to say, I don't fully understand this. That said, I'm curious to hear what ML researchers think about Lecun's take, and if there's any engineering done around it. I can't find much after the release of ijepa from his group.

Show context
probably_wrong ◴[14 Mar 25 18:38 UTC] No.43365762[source]▶
I haven't read Yann Lecun's take. Based on your description alone my first impression would be: there's a paper [1] arguing that "beam search enforces uniform information density in text, a property motivated by cognitive science". UID claims, in short, that a speaker only delivers as much content as they think the listener can take (no more, no less) and the paper claims that beam search enforced this property at generation time.

The paper would be a strong argument against your point: if neural architectures are already constraining the amount of information that a text generation system delivers the same way a human (allegedly) does, then I don't see which "energy" measure one could take that could perform any better.

Then again, perhaps they have one in mind and I just haven't read it.

[1] https://aclanthology.org/2020.emnlp-main.170/

replies(1): >>43365789 #
1. vessenes ◴[14 Mar 25 18:40 UTC] No.43365789[source]▶
I believe he’s talking about some sort of ‘energy as measured by distance from the models understanding of the world’ as in quite literally a world model. But again I’m ignorant, hence the post!
replies(3): >>43365847 #>>43366229 #>>43373547 #
2. tyronehed ◴[14 Mar 25 18:45 UTC] No.43365847[source]▶
When an architecture is based around world model building, then it is a casual outcome that similar concepts and things end up being stored in similar places. They overlap. As soon as your solution starts to get mathematically complex, you are departing from what the human brain does. Not saying that in some universe it might be possible to make a statistical intelligence, but when you go that direction you are straying away from the only existing intelligences that we know about. The human brain. So the best solutions will closely echo neuroscience.
3. deepsquirrelnet ◴[14 Mar 25 19:22 UTC] No.43366229[source]▶
In some respects that sounds similar to what we already do with reward models. I think with GRPO, the “bag of rewards” approach doesn’t strike me as terribly different. The challenge is in building out a sufficient “world” of rewards to adequately represent more meaningful feedback-based learning.

While it sounds nice to reframe it like a physics problem, it seems like a fundamentally flawed idea, akin to saying “there is a closed form solution to the question of how should I live.” The problem isn’t hallucinations, the problem is that language and relativism are inextricably linked.

4. anonymoushn ◴[15 Mar 25 16:35 UTC] No.43373547[source]▶
This sort of measure is a decent match for BPB though. BPB=-log(document_probability)/document_length_bytes and perplexity=e^(BPB*document_length_bytes/document_length_tokens). We already train models by minimizing perplexity, and model outputs are already those that are high probability. Though like with EBMs, figuring out outputs with even higher probability would require an expensive search step.