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

Claude's Cycles [pdf]

www-cs-faculty.stanford.edu

Post reply on HN
Print view View original

221–230 of 376 posts

Re: Claude's Cycles [pdf]

#221

Post by gravypod »

Aerroon wrote: post #3
mccoyb wrote: post #2

It's fascinating to think about the space of problems which are amenable to RL scaling of these probability distributions. Before, we didn't have a fast (we had to rely on human cognition) way to try problems - even if the techniques and workflows were known by someone. Now, we've baked these patterns into probability distributions - anyone can access them with the correct "summoning spell". Experts will naturally us…

A bit related: open weights models are basically time capsules. These models have a knowledge cut off point and essentially forever live in that time.

This is very interesting. I wonder if someone could create a future-sight benchmark for these models? Like, if given a set of newspaper articles for the past N months can it predict if certain world events would happen? We could backtest against results that have happened since the training cutoff.

Re: Claude's Cycles [pdf]

#222

Post by NewsaHackO »

CobrastanJorji wrote: post #216
OneManyNone wrote: post #159

Earlier quoted context omitted.

Claude did not find a proof, though. It found an algorithm which Knuth then proved was correct.

Yes, and his point is that finding that algorithm was, to Knuth, the interesting part. Getting from that to a proof was the boring bit.

Yeah, and I'm not sure what the other guy's argument is. It's Knuth, the primary researcher, who is giving the praise here. I don't see a possible motivation he would have to falsely give accolades to a AI for a problem he presented, then cleaned up to solve.

Re: Claude's Cycles [pdf]

#223

Post by famouswaffles »

konne88 wrote: post #120

I didn't expect such a misleading intro from Knuth. It reads like Claude solved Knuth's math problem. In reality, Claude generated various example solution, and Knuth then manually generalized that to a formal proof. What Claude did is certainly useful, but it would have been nice to be clear about the scope of the contribution in the intro.

Claude solved it, Knuth developed the proof for the solution.

Re: Claude's Cycles [pdf]

#224

Post by iterance »

OneManyNone wrote: post #159
bachmeier wrote: post #139

Earlier quoted context omitted.

My interpretation is that Claude did what Knuth considers to be the "solution". Doing the remaining work and polishing up the proof are not necessary to have a solution from this perspective.

Claude did not find a proof, though. It found an algorithm which Knuth then proved was correct.

The insight is the point of research. Proof isn't the desired product of research, it's simply an apparatus that exists for the purpose of verifying and demonstrating correctness of insight.

Re: Claude's Cycles [pdf]

#225

Post by charcircuit »

lmf4lol wrote: post #200
charcircuit wrote: post #41

Earlier quoted context omitted.

Memory is not just bolted on top of the latest models. They under go training on how and when to effectively use memory and how to use compaction to avoid running out of context when working on problems.

do you have some reading material to share on this matter? thanks already

I don't, but look into what the creators of Codex, Gemini CLI, Claude Code, Kimi CLI, etc have said about the models. While these harnesses are advertised as coding specific we know that coding ability correlates with reasoning ability.

Re: Claude's Cycles [pdf]

#226

Post by fourthark »

LPisGood wrote: post #205
kami23 wrote: post #197

Earlier quoted context omitted.

When the LLMs start compacting they summarize the conversation up to that point using various techniques. Overall a lot of maybe finer points of the work goes missing and can only be retrieved by the LLM being told to search for it explicitly in old logs. Once you compact, you've thrown away a lot of relevant tokens from your problem solving and they do become significantly dumber as a result. If I see a compaction c…

> I ask it to write a letter to its future self, and then start a new session by having it read the letter Is that not one kf the primary technologies for compactification?

I think the point is that you have a better idea of what you want it to remember and even a small hint can have big impact.

Just saying "write up what you know", with no other clues, should not perform any better than generic compaction.

Re: Claude's Cycles [pdf]

#228

Post by jeremyjh »

bonoboTP wrote: post #207
xlii wrote: post #201

Earlier quoted context omitted.

So the actors who portrait great thinkers are great thinkers?

No, actors recite a pre-written script. But scriptwriters do have to be great thinkers in order to know what the great thinker would actually say.

Which is why so many portrayals are unconvincing.

Re: Claude's Cycles [pdf]

#229

Post by saturnite »

pdntspa wrote: post #210
wang_li wrote: post #68

Earlier quoted context omitted.

Or you could have just said "they can't form new memories."

Sure, if you want to speak with the precision of a sledgehammer instead of a scalpel

All that needed to be conveyed was that there are humans who cannot create new memories. That is enough to pose the philosophical question about these models having intelligence. Anything more is just adding an anecdote that isn't necessary.

Re: Claude's Cycles [pdf]

#230

Post by j45 »

Aerroon wrote: post #3
mccoyb wrote: post #2

It's fascinating to think about the space of problems which are amenable to RL scaling of these probability distributions. Before, we didn't have a fast (we had to rely on human cognition) way to try problems - even if the techniques and workflows were known by someone. Now, we've baked these patterns into probability distributions - anyone can access them with the correct "summoning spell". Experts will naturally us…

A bit related: open weights models are basically time capsules. These models have a knowledge cut off point and essentially forever live in that time.

That's a nice way of putting it, appreciate you sharing.
Post reply on HN