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

521–530 of 659 posts

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

#521

Post by adriand »

mlsu wrote: post #514
noisy_boy wrote: post #432

Earlier quoted context omitted.

I feel like it almost always starts well, given the full picture, but then for non-trivial stuff, gets stuck towards the end. The longer the conversation goes, the more wheel-spinning occurs and before you know it, you have spent an hour chasing that last-mile-connectivity. For complex questions, I now only use it to get the broad picture and once the output is good enough to be a foundation, I build the rest of it m…

This is the way. I've had this experience many times: - hey, can you write me a thing that can do "xyz" - sure, here's how we can do "xyz" (gets some small part of the error handling for xyz slightly wrong) - can you add onto this with "abc" - sure. in order to do "abc" we'll need to add "lmn" to our error handling. this also means that you need "ijk" and "qrs" too, and since "lmn" doesn't support "qrs" out of the bo…

Some good suggestions here. I have also had success asking things like, “is this a standard/accepted approach for solving this problem?”, “is there a cleaner, simpler way to do this?”, “can you suggest a simpler approach that does not rely on X library?”, etc.

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

#522

Post by TeMPOraL »

troupo wrote: post #442
KoolKat23 wrote: post #393

Earlier quoted context omitted.

you're both on the wrong wavelength. No one has claimed it is better than an expert human yet. Be glad, for now your jobs are safe, why not use it as a tool to boost your productivity, yes, even though you'll get proportionally less use than others in other perhaps less "expert" jobs.

In order for it to boost productivity it needs to answer more than the regular questions for the top-3 languages on Stackoverflow, no? It often fails even for those questions. If I need to babysit it for every line of code, it's not a productivity boost.

If you need to babysit it for every line of code, you're either a superhuman coder, working in some obscure alien language, or just using the LLM wrong.

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

#523

Post by markk »

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…

Stupid question: Why can't models be trained in such a way to rate the authoritativeness of inputs? As a human, I contain a lot of bad information, but I'm aware of the source. I trust my physics textbook over something my nephew thinks.

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

#524

Post by 015a »

littlestymaar wrote: post #519
015a wrote: post #511

Earlier quoted context omitted.

So what you're saying is, LLMs are good enough to do something that humans are already capable of doing, in a timeframe that a human would be reasonably capable of doing it in, and its unrealistic to believe that LLMs will ever be able to do something truly superhuman. Got it :+1:

Being able to do “stuff a human is capable of doing” used to be the definition of “artificial intelligence” and until very recently it was seen as a dream that may never happen. And it hasn't completely happened yet BTW, there are still plenty of trivial stuff LLM can't do just because there's no available training data for that. Also their ability to do “reasoning” or few-shot-learning is overhyped (even if impressi…

> AI isn't going to do “faster than the speed of light” MPEG decoding ever, all physical limits apply to it.

This simply isn't a good faith take, because you're straw-manning the implementation of the query that the original poster put forward. They aren't asserting that the AI would need to do supernatural super-real time decoding of MPEG encoded files. What if the AI had already seen them? And was able to encode in the typically-compressed way LLMs do the information it needs to answer questions like that without re-decoding the original movies?

This raises many valid questions on topics like the structuring of data within an LLM, how large LLMs may eventually become, what systems should orbit around the LLM (does it make more sense for LLMs to watch YouTube videos, or have already watched YouTube videos?).

My definition of AI is the same definition that Nick Bostrom talks about in his 2014 book Superintelligence. There's no moving goalposts. Goal posts have been set in cement since 2014. Achieving human-level parity has obviously only been a "goal" insomuch as its a 10 millisecond stop on the gradient toward superintelligence. OpenAI is not worth $150 billion dollars because it purports to be building a human-and-nothing-more in a box.

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

#525

Post by bonoboTP »

ImHereToVote wrote: post #349
bonoboTP wrote: post #322

Earlier quoted context omitted.

This illustrates a different point. This is a variation on a well known riddle that definitely comes up in the training corpus many times. In the original riddle a father and his son die in the car accident and the idea of the original riddle is that people will be confused how the boy can be the doctor's son if the boy's father just died, not realizing that women can be doctors too and so the doctor is the boy's mot…

My codebases are riddled with these gotchas. For instance, I sometimes write Python for the Blender rendering engine. This requires highly non-idiomatic Python. Whenever something complex comes up, LLM's just degenerate to cookie cutter basic bitch Python code. There is simply no "there" there. They are very useful to help you reason about unfamiliar codebases though.

For me the best coding use case is getting up to speed in an unfamiliar library or usage. I describe the thing I want and get a good starting point and often the cookie-cutter way is good enough. The pre-LLM alternative would be to search for tutorials but they will talk about some slightly different problem with different goals etc then you have to piece it together, and the tutorial assumes you already know a bunch of things like how to initialize stuff and skips the boilerplate and so on.

Now sure, actually working through it will give a deeper understanding that might come handy at a later point, but sometimes the thing is really a one-off and not an important point. Like as an AI researcher I sometimes want to draft up a quick demo website, or throw together a quick Qt GUI prototype or a Blender script or use some arcane optimization library or write a SWIG or a Cython wrapper around a C/C++ library to access it in Python, or how to stuff with Lustre, or the XFS filesystem or whatever. Any number of small things where, sure, I could open the manual, do some trial and error, read stack overflow, read blogs and forums, OR I could just use an LLM, use my background knowledge to judge whether it looks reasonable, then verify it, use the now obtained key terms to google more effectively etc. You can't just blindly copy-paste it and you have to think critically and remain in the driver seat. But it's an effective tool if you know how and when to use it.

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

#526

Post by TeMPOraL »

jiggawatts wrote: post #145
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 Oooh... oohhh!! I just had a thought: By now we're all familiar with the strict JSON output mode capability of these LLMs. That's just a matter of filtering the token probability vector by the output grammar. Only valid tokens are allowed, which guarantees that the output matches the grammar. But... why just data grammars? Why not t…

I still fail to see the overall problem. Hallucinating non-existing libraries is a good programming practice in many cases: you express your solution in terms of an imaginary API that is convenient for you, and then you replace your API with real functions, and/or implement it in terms of real functions.

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

#527

Post by Salgat »

MPSimmons wrote: post #106
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 failure is in how you're using it. I don't mean this as a personal attack, but more to shed light on what's happening. A lot of people use LLMs as a search engine. It makes sense - it's basically a lossy compressed database of everything its ever read, and it generates output that is statistically likely - varying degrees of likeliness depending on the temperature, as well as how many times the particular weights…

Except that it sometimes does do those tasks well. The danger in an LLM isn't that it sometimes hallucinates, the danger is that you need to be sufficiently competent to know when it hallucinates in order to fully take advantage of it, otherwise you have to fallback to double checking every single thing it tells you.

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

#528

Post by petesergeant »

MPSimmons wrote: post #106
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 failure is in how you're using it. I don't mean this as a personal attack, but more to shed light on what's happening. A lot of people use LLMs as a search engine. It makes sense - it's basically a lossy compressed database of everything its ever read, and it generates output that is statistically likely - varying degrees of likeliness depending on the temperature, as well as how many times the particular weights…

This is demonstrably wrong, because you can just add "is this real" to a response and it generally knows if it made it up or not. Not every time, but I find it works 95% of the time. Given that, this is exactly a step I'd hope an advanced model was doing behind the scenes.

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

#529

Post by accountnum »

jfengel wrote: post #516
accountnum wrote: post #331

Earlier quoted context omitted.

It literally is a riddle, just as the original one was, because it tries to use your expectations of the world against you. The entire point of the original, which a lot of people fell for, was to expose expectations of gender roles leading to a supposed contradiction that didn't exist. You are now asking a modified question to a model that has seen the unmodified one millions of times. The model has an expectation o…

Phrased as it is, it deliberately gives away the answer by using the pronoun "he" for the doctor. The original deliberately obfuscates it by avoiding pronouns. So it doesn't take an understanding of gender roles, just grammar.

My point isn't that the model falls for gender stereotypes, but that it falls for thinking that it needs to solve the unmodified riddle.

Humans fail at the original because they expect doctors to be male and miss crucial information because of that assumption. The model fails at the modification because it assumes that it is the unmodified riddle and misses crucial information because of that assumption.

In both cases, the trick is to subvert assumptions. To provoke the human or LLM into taking a reasoning shortcut that leads them astray.

You can construct arbitrary situations like this one, and the LLM will get it unless you deliberately try to confuse it by basing it on a well known variation with a different answer.

I mean, genuinely, do you believe that LLMs don't understand grammar? Have you ever interacted with one? Why not test that theory outside of adversarial examples that humans fall for as well?

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

#530

Post by SeanAnderson »

franze wrote: post #402

Just coded this this morning using chatgpt o1 - it is the reimplementation of an old idea now music, multiple dots, more and more bug fixes honestly, chatgpt is now a better coder than i ever was or will be https://lsd.franzai.com/

Neat idea. The ball frequently passes through solid lines though.
Post reply on HN