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

Understanding large language models: A cross-section of the relevant literature

magazine.sebastianraschka.com

Post reply on HN
Print view View original

21–30 of 32 posts

Re: Understanding large language models: A cross-section of the relevant literature

#21

Post by jonathanstrange »

two_in_one wrote: post #4
cs702 wrote: post #2

This is a good intro for anyone who already has at least some background in ML and wants to get up to speed on LLMs relatively quickly. Props to the author for giving credit to Bandanau et al (2014), which I believe first proposed the concept of applying a Softmax function over token scores to compute attention, setting the stage for the original transformer by Vaswani et al (2017).

The big question here is what's next? So far my coding experiments with GPT-4 shows it's a shallow thinker. It has ready answers for many questions. One step further and it fails miserably. Much better but not a full replacement for google search.

I was wondering why nobody combines LLMs with well-known symbolic AI like CYC[1]. Or, perhaps I'm wrong and plenty of companies are working on it.

[1] https://cyc.com/

Re: Understanding large language models: A cross-section of the relevant literature

#22

Post by RandomLensman »

atleastoptimal wrote: post #9

I think LLMs are gonna be the "jumpstart" of more general AGI prototypes over the next couple years. The large corpus of text gives them a general basis of logical patterns, which can then be pruned iteratively in simulated environments.

My experience with ChatGPT is that it cannot reason from the data it has, i.e., it cannot take abstract concepts it can write about and use them on something it doesn't have the solution for. So not sure how the logical patterns really can evolve into something closer to AGI there. Maybe other LLMs? The inability to do math properly is really limiting, I think.

Re: Understanding large language models: A cross-section of the relevant literature

#23

Post by simonh »

jonathanstrange wrote: post #21
two_in_one wrote: post #4

Earlier quoted context omitted.

The big question here is what's next? So far my coding experiments with GPT-4 shows it's a shallow thinker. It has ready answers for many questions. One step further and it fails miserably. Much better but not a full replacement for google search.

I was wondering why nobody combines LLMs with well-known symbolic AI like CYC[1]. Or, perhaps I'm wrong and plenty of companies are working on it. [1] https://cyc.com/

It depends what you mean by combine. They work in fundamentally different ways. It's a bit like combining an internal combustion engine with a ramjet.

You might be able to compose systems that use LLMs for some sub-components and symbolic AI for other functions. So for example if ChatGPT gets asked a question that LLMs are poor at but symbolic systems are good at, it could switch to using a symbolic system, but that's not really combining the actual technologies.

Re: Understanding large language models: A cross-section of the relevant literature

#24

Post by fredgrott »

two_in_one wrote: post #4
cs702 wrote: post #2

This is a good intro for anyone who already has at least some background in ML and wants to get up to speed on LLMs relatively quickly. Props to the author for giving credit to Bandanau et al (2014), which I believe first proposed the concept of applying a Softmax function over token scores to compute attention, setting the stage for the original transformer by Vaswani et al (2017).

The big question here is what's next? So far my coding experiments with GPT-4 shows it's a shallow thinker. It has ready answers for many questions. One step further and it fails miserably. Much better but not a full replacement for google search.

The next thing is how do we model quantitative thinking as every facet of what makes us human is involved in quantitative thinking.

Re: Understanding large language models: A cross-section of the relevant literature

#25

Post by lucubratory »

RandomLensman wrote: post #22
atleastoptimal wrote: post #9

I think LLMs are gonna be the "jumpstart" of more general AGI prototypes over the next couple years. The large corpus of text gives them a general basis of logical patterns, which can then be pruned iteratively in simulated environments.

My experience with ChatGPT is that it cannot reason from the data it has, i.e., it cannot take abstract concepts it can write about and use them on something it doesn't have the solution for. So not sure how the logical patterns really can evolve into something closer to AGI there. Maybe other LLMs? The inability to do math properly is really limiting, I think.

I haven't yet seen an example of an LLM failing at math that cannot be very easily solved by having the LLM use a calculator, much as all humans do with math of any significance. It needs to use a calculator more often, but that's a completely negligible compute cost.

Re: Understanding large language models: A cross-section of the relevant literature

#26

Post by RandomLensman »

lucubratory wrote: post #25
RandomLensman wrote: post #22

Earlier quoted context omitted.

My experience with ChatGPT is that it cannot reason from the data it has, i.e., it cannot take abstract concepts it can write about and use them on something it doesn't have the solution for. So not sure how the logical patterns really can evolve into something closer to AGI there. Maybe other LLMs? The inability to do math properly is really limiting, I think.

I haven't yet seen an example of an LLM failing at math that cannot be very easily solved by having the LLM use a calculator, much as all humans do with math of any significance. It needs to use a calculator more often, but that's a completely negligible compute cost.

I meant more abstract math, e.g., construct something that needs math to construct it. Take a concept like risk neutral pricing and get it to construct a replicating portfolio for something not totally trivial (i.e., with not a lot of solved examples on the web). Fails for me.

Re: Understanding large language models: A cross-section of the relevant literature

#27

Post by pohl »

simonh wrote: post #23
jonathanstrange wrote: post #21

Earlier quoted context omitted.

I was wondering why nobody combines LLMs with well-known symbolic AI like CYC[1]. Or, perhaps I'm wrong and plenty of companies are working on it. [1] https://cyc.com/

It depends what you mean by combine. They work in fundamentally different ways. It's a bit like combining an internal combustion engine with a ramjet. You might be able to compose systems that use LLMs for some sub-components and symbolic AI for other functions. So for example if ChatGPT gets asked a question that LLMs are poor at but symbolic systems are good at, it could switch to using a symbolic system, but that'…

I’ve had similar thoughts, wondering why LLMs aren’t just a component of a larger system, perhaps used to generate “thoughts” that are then tested by some other subsystems, which in turn may re-fire the LLMs to generate more.

I guess I’m imagining that testing the LLM output might be a problem that some other technique might be suited for, and smells like what we need to mitigate LLM weaknesses.

Re: Understanding large language models: A cross-section of the relevant literature

#28

Post by mrbungie »

atleastoptimal wrote: post #12
notnaut wrote: post #10

Earlier quoted context omitted.

So you think the logical patterns found in human language might also be similar enough to the logical patterns found in other systems that these LLMs have a jumpstart in figuring those out? We already have such a poor idea of how these things seem to understand so much… it’ll be a wild day when cancer is cured and we have absolutely no idea why.

IMO a sufficiently advanced AI should be able to do a full analysis of its neural architecture and explain it and break down its functionality.

That would be even more intelligent than the majority (or even all, YMMV) of humans.

Re: Understanding large language models: A cross-section of the relevant literature

#29

Post by agileAlligator »

rahmeero wrote: post #8
agileAlligator wrote: post #6

Earlier quoted context omitted.

Try https://www.phind.com/ . I find it's much better optimized for coding purposes than regular GPT-4.

Just tried it. I like that it lists the information sources in the right-hand panel, and provides the digested information in the main panel

I've found that it is reasonably competent at most coding tasks; especially on Expert mode.

I rarely have to ask it to alter the code it gives me.

Re: Understanding large language models: A cross-section of the relevant literature

#30

Post by lucubratory »

RandomLensman wrote: post #26
lucubratory wrote: post #25

Earlier quoted context omitted.

I haven't yet seen an example of an LLM failing at math that cannot be very easily solved by having the LLM use a calculator, much as all humans do with math of any significance. It needs to use a calculator more often, but that's a completely negligible compute cost.

I meant more abstract math, e.g., construct something that needs math to construct it. Take a concept like risk neutral pricing and get it to construct a replicating portfolio for something not totally trivial (i.e., with not a lot of solved examples on the web). Fails for me.

That's fair, I've seen that it can really struggle coming up with novel algorithms. I am curious on if there is more improvement on that front in future models, because even its current performance at algorithmic manipulation is far, far better than e.g. GPT-3.
Post reply on HN