[dead]
LLMs, Theory of Mind, and Cheryl's Birthday
141–150 of 150 posts
Re: LLMs, Theory of Mind, and Cheryl's Birthday
#142Gp1-o1 preview solves this puzzle correctly in 13 seconds and has a thorough logical deduction in the comments and explanation. I think it’s a bit unfair on llm to ask it to retrieve the puzzle definition from its training data. I posted the info on the puzzle from his notebook. https://chatgpt.com/share/670103ae-1c18-8011-8068-dd21793727...
Disappointing that Norvig didn’t use the model that OpenAI states is their best model for programming. Also using himself as the programmer seemed like a convenient choice. I’d much rather see him grab a random professional programmer for the task.
gpt-o1 was released Sept. 12th and Norvig ran his tests Sept 25th... I don't understand how Norvig didn't think to test gpt-o1, it actually irritates me lol
Re: LLMs, Theory of Mind, and Cheryl's Birthday
#143The problem with evaluating LLMs is that there's a random component, and the specific wording of prompts is so important. I asked Claude to explain the problem, then write python to solve it. When it ran there was an exception, so I pasted that back in and got the correct answer. I'm not sure what this says about theory of mind (the first script it wrote was organized into steps based on who knew what when, so it see…
This test plainly shows that even with the real solution in the training data, the wrong answer is written as though it's the correct answer. A human would say, "I'm not sure, I want to test it." The current AI summer is heaving with breathless claims of intelligence, comprehension, reasoning, etc.
I think these claims need to be balanced with a cold shower of reality. Personally, I find LLMs very impressive at what they do well; generating and summarizing and translating. People apologizing for LLMs' performance at straight-forward reasoning and programming tasks, suggesting various crutches and head-starts, gives me the creeps. It's not the Messiah. It's a very naughty computer program.
Re: LLMs, Theory of Mind, and Cheryl's Birthday
#144It's important to remember that modern LLMs are trained on bloody everything. They know every common logic problem, at least when stated the way they would have seen it.
If you want to test an LLM, always make up a new problem. It can be the same idea as an existing problem, but change all names and numbers.
I tested if GPT 3.5 could recognize chaos theory. If I stated it as the typical "butterfly flaps its wings" it instantly recognized it as the chaos theory example. If I totally changed the problem statement, it correctly identified that weather isn't correlated with a single action by a single person, but it didn't associate it with chaos theory.
Re: LLMs, Theory of Mind, and Cheryl's Birthday
#145Earlier quoted context omitted.
Disappointing that Norvig didn’t use the model that OpenAI states is their best model for programming. Also using himself as the programmer seemed like a convenient choice. I’d much rather see him grab a random professional programmer for the task.
gpt-o1 was released Sept. 12th and Norvig ran his tests Sept 25th... I don't understand how Norvig didn't think to test gpt-o1, it actually irritates me lol
Not everybody follows GPT releases so closely. I work implementing software using LLMs and this is the first I've heard of this.
Re: LLMs, Theory of Mind, and Cheryl's Birthday
#146The approach is fundamentally flawed. You can’t query an LLM as to whether it has a theory of mind. You need to analyze how its internal logic works.
Imagine the opposite result had occurred, and the LLM had outputted something which was considered a theory of mind… Does that prove it has one, or that it was trained on some data that had something it used which made it sound like it has a theory of mind?
Re: LLMs, Theory of Mind, and Cheryl's Birthday
#147This seems like a terrible test case since python examples are readily available in the training data: https://rosettacode.org/wiki/Cheryl%27s_birthday It's interesting that so many of the model's fail to retrieve this, but any thta do solve it should clearly be able to do so with no reasoning/theory of mind.
I agree this is not a great test. What's good about it is that it is a constraint satisfaction problem, and I would expect LLMs to be pretty bad at unknown problems of this kind. Simple reason, an LLM only has a a finite number of layers and it cannot do arbitrary long searches.
I don't understand what you're saying - the idea is that we're asking the LLM to generate code to perform the search, rather than run an arbitrarily long search on its own, right? So why should the number of layers it has matter?
Re: LLMs, Theory of Mind, and Cheryl's Birthday
#148Deducing things from the inability of an LLM to answer a specific question seemed doomed by the "it will be able to on the next itteration" principle. It seems like the only way you could systematic chart the weaknesses of an LLM is by having a class of problems that get harder for LLMs at a steep rate, so a small increase in problem complexity requires a significant increase in LLM power.
... and such that the same increase in problem complexity requires a smaller increase in human effort to solve.
This was the idea with the Winograd schema challenge [0] and now the ARC benchmark [1], but human-level performance on the former was achieved in 2019, and very strong progress is being made over the last few months on the latter. But at the current point in time, it seems that we're pretty much reaching the limit of such challenges that are relatively easy for humans to solve in a single sitting, and we'll have to start switching to benchmarks which rely on extensive work over time, such as SWE-Bench [1], and even there it seems that state of the art AI agents are already doing better than the "average" human developer.
Re: LLMs, Theory of Mind, and Cheryl's Birthday
#149Earlier quoted context omitted.
For me the simplest way to test would be to first ask specific knowledge, and then ask where it learnt that knowledge, and check the reference. Currently they fail spectacularly, and the most useful next step would be to use source-aware training
Why would I know where I learned a thing, much less be expected to produce a valid URL off the top of my head?
I am not trying to explain your specific brain. Next time people play the game werewolf in real life, join it for a couple of rounds, and tell the players you're not too familiar with the game, and ask them to discuss mistakes after each round. You will notice they pay a lot of attention to who said what. If you don't pay attention you become like a villager lynching random people, while if the villagers play enough attention they can prevent the wherewolf from killing the whole village most of the time.
Re: LLMs, Theory of Mind, and Cheryl's Birthday
#150Earlier quoted context omitted.
Why would I know where I learned a thing, much less be expected to produce a valid URL off the top of my head?
I am not trying to explain your specific brain. Next time people play the game werewolf in real life, join it for a couple of rounds, and tell the players you're not too familiar with the game, and ask them to discuss mistakes after each round. You will notice they pay a lot of attention to who said what . If you don't pay attention you become like a villager lynching random people, while if the villagers play enough…
Where did you learn that the capital of Virginia is Richmond?