Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Position: LLMs Can't Jump

openreview.net

Post reply on HN
Print view View original

101–110 of 233 posts

Re: Position: LLMs Can't Jump

#101

Post by dguest »

jimbo808 wrote: post #72
quantum_mcts wrote: post #30

The popular retelling of how Einstein created Special Relativity to "Resolve the contradictions of Michelson-Morly experiments" is very reductive to the history of the question. The epitome is the quote from the paper: > From the two postulates, Einstein derived the Lorentz trans- formation ... If Einstein derived them, who is "Lorentz"? The groundwork for Special Relativity was the study of electrodynamics and symme…

I get really tired of Einstein being portrayed as if he’s the greatest scientist who ever lived, when he doesn’t even belong in that conversation. If Einstein never existed, the fields he was in were already moving strongly in the directions of his conclusions. If Maxwell or Newton had never existed, the world today would be quite different.

Yeah, once you dig into economics, technology and science of the era it often feels like the big discoveries were inevitable. That doesn't make these people any less great, it's just to say there are many more great thinkers that people are forgetting about.

But I'd like more explanation as to why Maxwell or Newton are any less replaceable: all 4 of "Maxwell's equations" have other names attached to them (the exception is Ampere-Maxwell law [1] which Maxwell contributed an important term to), and Newton had a number of contemporaries who were making similar discoveries but are often forgotten.

[1]: https://en.wikipedia.org/wiki/Amp%C3%A8re's_circuital_law

Re: Position: LLMs Can't Jump

#102

Post by yk »

The paper is from the 30th of April this year, openAi announced the counter example to the unit distance problem on the 20th of May. That is to say this paper seems to have aged not much but quite poorly.

Re: Position: LLMs Can't Jump

#103

Post by nearbuy »

sobiolite wrote: post #10

The theory is that creative leaps in theoretical physics require a grounding in sensory experience, but the obvious counter-argument is that humans can make creative leaps in abstract fields without such sensory grounding. They do address this at the end, saying "In abstract domains such as Mathematics or Computer Science, the Sense Experience (E) may be grounded in high-dimensional topology or have other goals such…

The paper also fails to show that their central example, Einstein, relied on sensory experience for his intuition leaps rather than general reasoning. They just kind of claim that thought experiments require sensory experience. But you can easily ask an LLM to perform a thought experiment and simulate an outcome, and the SOTA LLMs generally seem to do about as well as a human. An LLM would certainly know that freefall feels the same as zero gravity, even if they haven't felt the sensation, which was the key intuition the paper talks about for Einstein's General Relativity. The paper's author would probably say any examples of this don't count, but without clear criteria for what would count, their claim is unfalsifiable.

Re: Position: LLMs Can't Jump

#104

Post by bee_rider »

jvanderbot wrote: post #2

Came for: "A computer once beat me at chess, but it was no match for me at kick boxing." TFA was actually about leaps of intuition, sadly. One of the experiments I've heard proposed around here is to somehow create an LLM from all text up to 1980 or 1990 and see if it can get back to making itself.

> One of the experiments I've heard proposed around here is to somehow create an LLM from all text up to 1980 or 1990 and see if it can get back to making itself.

This is an interesting experiment but I wonder if it would be possible to prevent some sort of retrospective bias. For example, I’d expect the experiments that lead to relativity to be over-represented in our catalogue of scientific literature prior to 1900, just because in retrospect they were important, so the records about them were preserved.It would have to be a very intentionally constructed corpus, I think.

Re: Position: LLMs Can't Jump

#105

Post by tpolm »

reliablereason wrote: post #93
buzzin__ wrote: post #68

Earlier quoted context omitted.

Or some randomness is aomehowntroduced in the output, which happens after every word, unless you set the temperature to zero.

That would not be intuition that is just randomness. Intuition is not randomness. A jump in intuition comes from automatic processes reorganising the relational structure of conceptual models. There is no reorganisation of the model durring inference.

> There is no reorganisation of the model durring inference.

one could argue that model can reorganize / interact with prior knowledge captured in text form (edit files) hence there can be reorganisation

Re: Position: LLMs Can't Jump

#106

Post by wnmurphy »

I had a related insight, but in the domain of humor [1].

LLMs are inherently probabilistic, and there's currently no mechanism for producing an orthogonal directional change in the path traced through a latent space which is also contextually relevant (landing on a punch line).

In other words, LLMs are fundamentally incapable of making intuitive/orthogonal leaps in context.

It might be possible to add this capability with a new architectural component like transformers, but specifically for making "left turns"/intuitive leaps.

[1] https://wnmurphy.com/llms-cant-do-humor/

Re: Position: LLMs Can't Jump

#107

Post by unfitted2545 »

jvanderbot wrote: post #2

Came for: "A computer once beat me at chess, but it was no match for me at kick boxing." TFA was actually about leaps of intuition, sadly. One of the experiments I've heard proposed around here is to somehow create an LLM from all text up to 1980 or 1990 and see if it can get back to making itself.

If would be interesting to see 5 billion LLM's working together, each with random mutations (temperature ig). Would we essentially be looking at a society through a petri dish? Ofc 5 billion is quite a lot of compute.

Re: Position: LLMs Can't Jump

#108

Post by goatlover »

yomismoaqui wrote: post #61

Why everybody is obsessed with replacing humans with LLMs when it seems like the most profitable use cases (like coding agents) rely on enhancing human capabilities? Until LLMs have some 0% error humans will have to be in the loop (even if they only serve to take responsibility of the process).

Would have been better if AI was known as Augmented Intelligence as it's a tool to to boost our own intelligence. Instead the term that promises science fiction futures predominated, and now it's being used to raise a ton of money.

Saying you want to make workers more productive and provide better tools for people just isn't that sexy.

Re: Position: LLMs Can't Jump

#109

Post by StevenWaterman »

daun_gee wrote: post #78
yomismoaqui wrote: post #61

Why everybody is obsessed with replacing humans with LLMs when it seems like the most profitable use cases (like coding agents) rely on enhancing human capabilities? Until LLMs have some 0% error humans will have to be in the loop (even if they only serve to take responsibility of the process).

[dead]

> The only way to justify trillion dollar valuations

Also possible if you make god

Re: Position: LLMs Can't Jump

#110

Post by ACCount37 »

vatsachak wrote: post #95
ACCount37 wrote: post #33

It's a very shaky position, and the empirical track record of "LLMs can't..." is in itself a reason to call it into doubt. Every "can't" of this nature was followed by a discovery of "they can, just poorly", and then by that "poorly" improving steadily generation to generation. The paper doesn't provide a way to measure or quantify this elusive "jumping" capability, not even as an approximation. It just throws "can't…

Okay here's something LLMs can't. They can't solve problems that are longer than ~10 pages of math. They also can't maintain codebases without supervision. It's because they are have no memory and use various tricks to supplant that fact.

Now, how long before someone rolls out some sort of 10M context hybrid attention active context management monstrosity and ruins this guy's "can't"? Start the clock.

My opinion of claims like "LLMs need memory to manage codebases" has also hit the dumpster bin a while ago.

Why would knowing how to make a maintainable change to a codebase require any more "memory" than knowing how to play an optimal chess move? The codebase is the memory. A sufficiently capable LLM can ingest it, figure out what changes to make, and make them.

Post reply on HN