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

How does GPT obtain its ability? Tracing emergent abilities of language models

yaofu.notion.site

Post reply on HN
Print view View original

61–70 of 205 posts

Re: How does GPT obtain its ability? Tracing emergent abilities of language models

#61

Post by esprehn »

resource0x wrote: post #43

Can anyone explain why the GPT response to the same question 4 days ago is different from today's: [ 4 days ago] > Can I learn to sing like Pavarotti? It is possible to learn to sing like Pavarotti, but it will require a lot of practice, dedication, and training. Pavarotti was a renowned opera singer known for his incredible vocal range, powerful voice, and technical mastery. He had many years of training and experie…

This is by design (and implemented as explained in the sibling comments). The reason for the randomness is to avoid returning the same thing all the time. For example if you ask it "write me a poem" you don't want it to return the exact same one each time.

You can also set the "temperature" parameter to 0 to remove the non determinism (though I'm not sure that's exposed in the web interface).

Re: How does GPT obtain its ability? Tracing emergent abilities of language models

#62

Post by AnthonyMouse »

aeternum wrote: post #59
hippich wrote: post #30

Earlier quoted context omitted.

There is some kind of recursion in here with authors names and "have no respect..." part:)

Fair point

Actually, it isn't.

Their arguments are of the form "this statement could be false."

If you evaluate it as a true statement, you have no problems. It could be false; you have to evaluate it for yourself instead of trusting some authority.

It's only if you assert that it's certainly false that you have a problem. Because then it's clearly true -- since otherwise these authorities would be telling you something false, which proves their assertion true.

Put another way, it can get you from the undesirable position of blindly trusting authorities to the desirable position of questioning them, but not the other way around. Which is the intended result.

Re: How does GPT obtain its ability? Tracing emergent abilities of language models

#63

Post by LelouBil »

resource0x wrote: post #6
jimmaswell wrote: post #3

Imagine Wolfram's system with GPT integrated. I hope they're working on that.

Current GPT version is not very good at math, and it's not at all clear whether this can be fixed within a language model. Q: I have 2 cats and 1 horse. How many legs do they have in total? A: If you have two cats and one horse, then the total number of legs they have is 8. Cats typically have four legs, while horses have four legs as well. So two cats would have a total of 8 legs, and when you add the four legs of t…

I'd say it struggles with pure logic.

Is there any work towards using ChatGPT to generate first order logic from a prompt, solve it using an existing solver, and put it back to let it write the meaning of it ?

I think it could be interesting, as it would make some prompts work better, and only use got for actual language to language tasks but leave reasoning to another software.

Re: How does GPT obtain its ability? Tracing emergent abilities of language models

#64

Post by resource0x »

skybrian wrote: post #58
resource0x wrote: post #48

Earlier quoted context omitted.

Sometimes the meaning of the response is totally different. [ 4 days ago ] The son of my father, but not my brother. Who is he? If a person is the son of the speaker's father but is not the speaker's brother, then that person is the speaker's nephew. A nephew is the son of a person's sibling, so if the speaker's father has a son who is not the speaker's brother, that person is the speaker's nephew. For example, if th…

Yes. GPT doesn't really deal in meanings. Much like autocomplete, it doesn't know what the end of a sentence will be when it starts it. If it randomly chooses different words at the start of a sentence, it may pretend to have a different belief by the end. In reality it doesn't have beliefs any more than a library does. It implicitly contains beliefs (since it's been trained on them) and it can imitate them, but whic…

I see. The assumption here is that one can simulate intelligence without formalizing the notion of meaning. (And if "meaning" is not defined, then the notion of "truth" is impossible to define either). Is my understanding correct?

Re: How does GPT obtain its ability? Tracing emergent abilities of language models

#65

Post by ricopags »

kelseyfrog wrote: post #19
srajabi wrote: post #2

Amazing insight, particularly section 6. "- The two important but different abilities of GPT-3.5 are *knowledge* and *reasoning*. Generally, it would be ideal if we could *offload the knowledge part to the outside retrieval system and let the language model only focus on reasoning.* This is because: - The model’s internal knowledge is always cut off at a certain time. The model always needs up-to-date knowledge to an…

See REALM[1] for some older(2 years) work on this idea. 1. https://arxiv.org/abs/2002.08909

Or Meta's Atlas[0] for more recent work

[0]https://arxiv.org/abs/2208.03299

Re: How does GPT obtain its ability? Tracing emergent abilities of language models

#66

Post by int_19h »

resource0x wrote: post #6
jimmaswell wrote: post #3

Imagine Wolfram's system with GPT integrated. I hope they're working on that.

Current GPT version is not very good at math, and it's not at all clear whether this can be fixed within a language model. Q: I have 2 cats and 1 horse. How many legs do they have in total? A: If you have two cats and one horse, then the total number of legs they have is 8. Cats typically have four legs, while horses have four legs as well. So two cats would have a total of 8 legs, and when you add the four legs of t…

It doesn't have to be good at math. It just needs to know when the question is math, so that the answer can be delegated to a system that can easily answer it. Some people have already managed to do it with Jupyter notebooks (with manual copy-pasting of expressions generated by the model, and results of their evaluation).

Re: How does GPT obtain its ability? Tracing emergent abilities of language models

#67

Post by sdenton4 »

kelseyfrog wrote: post #19
srajabi wrote: post #2

Amazing insight, particularly section 6. "- The two important but different abilities of GPT-3.5 are *knowledge* and *reasoning*. Generally, it would be ideal if we could *offload the knowledge part to the outside retrieval system and let the language model only focus on reasoning.* This is because: - The model’s internal knowledge is always cut off at a certain time. The model always needs up-to-date knowledge to an…

See REALM[1] for some older(2 years) work on this idea. 1. https://arxiv.org/abs/2002.08909

Or RETRO, from deep mind, last year.

https://www.deepmind.com/publications/improving-language-mod...

Re: How does GPT obtain its ability? Tracing emergent abilities of language models

#68

Post by hcarvalhoalves »

imiric wrote: post #25
page_severian wrote: post #23

Earlier quoted context omitted.

These errors are common and important to note. Though what I find fascinating is how often I can prompt engineer around these errors by telling GPT to just be smarter and think more carefully. Just tried the following for instance. Q: You are a master logician, who can solve all manner of language problems and have excellent math skills. I have 2 cats and 1 horse. How many legs do they have in total? Show your reason…

That's interesting. It seems it's not great at raw calculation, but if you ask it to explain its reasoning, it derives steps from previous results, and arrives at the correct answer. It's similar to the sibling comment about asking it to tell a story.

This seem to work because in the end it’s a markov chain. So the probability of the next step of a long answer being correct is greater than the probability of jumping to a correct conclusion.

Re: How does GPT obtain its ability? Tracing emergent abilities of language models

#69

Post by int_19h »

resource0x wrote: post #64
skybrian wrote: post #58

Earlier quoted context omitted.

Yes. GPT doesn't really deal in meanings. Much like autocomplete, it doesn't know what the end of a sentence will be when it starts it. If it randomly chooses different words at the start of a sentence, it may pretend to have a different belief by the end. In reality it doesn't have beliefs any more than a library does. It implicitly contains beliefs (since it's been trained on them) and it can imitate them, but whic…

I see. The assumption here is that one can simulate intelligence without formalizing the notion of meaning. (And if "meaning" is not defined, then the notion of "truth" is impossible to define either). Is my understanding correct?

The assumption here is that one can produce useful outputs in this manner. Whether it constitutes "simulating intelligence" is a philosophical question.

Re: How does GPT obtain its ability? Tracing emergent abilities of language models

#70

Post by nos482 »

mensetmanusman wrote: post #40

Is there anything like ChatGPT happening from China? They have almost 10x’d AI investment compared to the west, and I would be somewhat surprised if that investment combined with ubiquitous math skills failed to produce.

AI is a pretty diverse field. I suspect a lot of their investment is targeted at image/video instead of NLP and generative text models in particular.
Post reply on HN