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

Notes on OpenAI's new o1 chain-of-thought models

simonwillison.net

Post reply on HN
Print view View original

41–50 of 659 posts

Re: Notes on OpenAI's new o1 chain-of-thought models

#41

Post by tsunamifury »

mensetmanusman wrote: post #38
layer8 wrote: post #23

Earlier quoted context omitted.

My experience is likely colored by the fact that I tend to turn to LLMs for problems I have trouble solving by myself. I typically don't use them for the low-hanging fruits. That's the frustrating thing. LLMs don't materially reduce the set of problems where I'm running against a wall or have trouble finding information.

LLMs are not for expanding the sphere of human knowledge, but for speeding up auto-correct of higher order processing to help you more quickly reach the shell of the sphere and make progress with your own mind :)

Definitely. When we talk about being skilled in a T shape LLMs are all about spreading your top of T and not making the bottom go deeper.

Re: Notes on OpenAI's new o1 chain-of-thought models

#42

Post by CuriouslyC »

deepsquirrelnet wrote: post #27

It kind of seems like they just wrote a generalized DSPy program. Can anyone confirm? This has been a very incremental year for OpenAI. If this is what it seems like, then I’ve got to believe they’re stalling for time.

DSPy doesn't do that, you could describe it as a langchain style agent that evaluates its own output though it's better/faster than that.

OpenAI is definitely trying to run a hype game to keep the ball rolling. They're burning cash too quickly given their monetization path though, so I think they're going to end up completely in Microsoft's pocket.

Re: Notes on OpenAI's new o1 chain-of-thought models

#43

Post by lukev »

It's interesting to note that there's really two things going on here:

1. A LLM (probably a finetuned GPT-4o) trained specifically to read and emit good chain-of-thought prompts.

2. Runtime code that iteratively re-prompts the model with the chain of thought so far. This sounds like it includes loops, branches and backtracking. This is not "the model", it's regular code invoking the model. Interesting that OpenAI is making no attempt to clarify this.

I wonder where the real innovation here lies. I've done a few informal stabs with #2 and I have a pretty strong intuition (not proven yet) that given the right prompting/metaprompting model you can do pretty well at this even with untuned LLMs. The end game here is complex agents with arbitrary continuous looping interleaved with RAG and tool use.

But OpenAI's philosophy up until now has almost always been "The bitter lesson is true, the model knows best, just put it in the model." So it's also possible that the prompt loop has no special sauce and that the capabilities here do come mostly from the model itself.

Without being able to inspect the reasoning tokens, we can't really get a lot of info about which is happening.

Re: Notes on OpenAI's new o1 chain-of-thought models

#45

Post by w4 »

CSSer wrote: post #40
w4 wrote: post #29

Earlier quoted context omitted.

It depends on what you're doing. If you're writing something specific to your particular problem, or thinking through how to structure your data, or even working on something tough to describe in words like UI design, it probably is easier to just code it yourself in most high-level languages. On the other hand, if you're just trying to get a framework or library to do something and you don't want to spend a bunch of…

An abstraction machete. Heh.

This is a wonderful term for it!

Re: Notes on OpenAI's new o1 chain-of-thought models

#46

Post by tptacek »

layer8 wrote: post #23
tptacek wrote: post #20

Earlier quoted context omitted.

I've had the opposite experience with some coding samples. After reading Nick Carlini's post, I've gotten into the habit of powering through coding problems with GPT (where previously I'd just laugh and immediately give up) by just presenting it the errors in its code and asking it to fix them. o1 seems to be effectively screening for some of those errors (I assume it's just some, but I've noticed that the o1 things…

My experience is likely colored by the fact that I tend to turn to LLMs for problems I have trouble solving by myself. I typically don't use them for the low-hanging fruits. That's the frustrating thing. LLMs don't materially reduce the set of problems where I'm running against a wall or have trouble finding information.

I use LLMs for three things:

* To catch passive voice and nominalizations in my writing.

* To convert Linux kernel subsystems into Python so I can quickly understand them (I'm a C programmer but everyone reads Python faster).

* To write dumb programs using languages and libraries I haven't used much before; for instance, I'm an ActiveRecord person and needed to do some SQLAlchemy stuff today, and GPT 4o (and o1) kept me away from the SQLAlchemy documentation.

OpenAI talks about o1 going head to head with PhDs. I could care less. But for the specific problem we're talking about on this subthread: o1 seems materially better.

Re: Notes on OpenAI's new o1 chain-of-thought models

#47

Post by cma »

knicholes wrote: post #32
ldjkfkdsjnv wrote: post #5

The theory is that this solves the data shortage problem, they can generate a ton of chain of reasoning data from what we already have. True iterative improvement, like out of a science fiction novel These models are going to get embedded deeply into IDE's, like cursor has, and essentially end software development as we know it. A properly written requirements spec, and an engineer, can do the work of 5. Software eng…

I dunno man. I just spent a couple hours trying to get it to write functioning code to read from my RTSP stream, detect if my kid is playing piano, and send the result to HomeAssistant. It did not succeed.

How many hours without it?

Re: Notes on OpenAI's new o1 chain-of-thought models

#48

Post by ttul »

mrtesthah wrote: post #25
layer8 wrote: post #18

The o1-preview model still hallucinates non-existing libraries and functions for me, and is quickly wrong about facts that aren't well-represented on the web. It's the usual string of "You're absolutely correct, and I apologize for the oversight in my previous response. [Let me make another guess.]" While the reasoning may have been improved, this doesn't solve the problem of the model having no way to assess if what…

It begs the question of whether we can supply a function to be called (e.g., one that compiles and runs code) to evaluate intermediate CoT results

It seems OpenAI has decided to keep the CoT results a secret. If they were to allow the model to call out to tools to help fill in the CoT steps, then this might reveal what the model is thinking - something they do not want the outside world to know about.

I could imagine OpenAI might allow their own vetted tools to be used, but perhaps it will be a while (if ever) before developers are allowed to hook up their own tools. The risks here are substantial. A model fine-tuned to run chain-of-thought that can answer graduate level physics problems at an expert level can probably figure out how to scam your grandma out of her savings too.

Re: Notes on OpenAI's new o1 chain-of-thought models

#49

Post by ttul »

OutOfHere wrote: post #37
jumploops wrote: post #3

> the idea that I can run a complex prompt and have key details of how that prompt was evaluated hidden from me feels like a big step backwards. As a developer, this is highly concerning, as it makes it much harder to debug where/how the “reasoning” went wrong. The pricing is also silly, because I’m paying for tokens I can’t see. As a user, I don’t really care. LLMs are already magic boxes and I usually only care abo…

> As a user, I don’t really care. Tell me: Just how is it fair for a user to pay for the reasoning tokens without actually seeing them? If they are not shared, the service can bill you anything they want for them!

OpenAI’s answer to this would be, “Okay then, don’t use it.”

Re: Notes on OpenAI's new o1 chain-of-thought models

#50

Post by faangguyindia »

layer8 wrote: post #18

The o1-preview model still hallucinates non-existing libraries and functions for me, and is quickly wrong about facts that aren't well-represented on the web. It's the usual string of "You're absolutely correct, and I apologize for the oversight in my previous response. [Let me make another guess.]" While the reasoning may have been improved, this doesn't solve the problem of the model having no way to assess if what…

>The o1-preview model still hallucinates non-existing libraries and functions for me, and is quickly wrong about facts that aren't well-represented on the web. It's the usual string of "You're absolutely correct, and I apologize for the oversight in my previous response. [Let me make another guess.]"

After that you switch to Claude Soñnet and after sometime it also gets stuck.

Problem with LLM is that they are not aware of libraries.

I've fed them library version, using requirements.txt, python version I am using etc...

They still make mistakes and try to use methods which do not exist.

Where to go from here? At this point I manually pull the library version I am using and go to its docs, I generate a page which uses the this library correctly (then I feed that example into LLM)

Using this approach works. Now I just need to automate it so that I don't have to manually find the library, create specific example which uses the methods I need in my code!

Directly feeding the docs isn't working well either.

Post reply on HN