Rodney Brooks on limitations of generative AI • Hacker News Board

Re: Rodney Brooks on limitations of generative AI

#21

Post by lispm » Sun, Jun 30, 2024, 9:11 AM UTC

aurareturn wrote: post #15

fxtentacle wrote: post #13

To me, this reads like a very reasonable take. He suggests to limit the scope of the AI problem, add manual overrides in case there are unexpected situations, and he (rightly, in my opinion) predicts that the business case for exponentially scaling LLM models isn't there. With that context, I like his iPod example. Apple probably could have made a 3TB iPod to stick to Moore's law for another few years, but after they…

I’m still waiting for SalesForce to integrate an LLM into Slack so I can ask it business logic and decisions long lost. Still waiting for Microsoft to integrate an LLM into outlook so I can get a summary of a 20 email long chain I just got CCed into. I don’t think the iPod comparison is a valid one. People only have so much time to listen to music. Past a certain point, no one has enough good music they like to put i…

> However, the more data you feed into an LLM, the smarter it should be in the response.

Is it that way? For example if it lacks a certain reasoning capability, then more data may not change that. So far LLMs lack useful ideas of truth, it will easily generate untrue statements. We see lots of hacks how to control that, with unconvincing results.

Re: Rodney Brooks on limitations of generative AI

#22

Post by aurareturn » Sun, Jun 30, 2024, 9:15 AM UTC

richrichie wrote: post #20

aurareturn wrote: post #15

Earlier quoted context omitted.

I’m still waiting for SalesForce to integrate an LLM into Slack so I can ask it business logic and decisions long lost. Still waiting for Microsoft to integrate an LLM into outlook so I can get a summary of a 20 email long chain I just got CCed into. I don’t think the iPod comparison is a valid one. People only have so much time to listen to music. Past a certain point, no one has enough good music they like to put i…

> the more data you feed into an LLM, the smarter it should be in the response This is not obvious though.

It’s in theory. The more information you have, the better the decision in theory.

Re: Rodney Brooks on limitations of generative AI

#23

Post by albert_e » Sun, Jun 30, 2024, 9:15 AM UTC

roenxi wrote: post #9

> He uses the iPod as an example. For a few iterations, it did in fact double in storage size from 10 all the way to 160GB. If it had continued on that trajectory, he figured out we would have an iPod with 160TB of storage by 2017, but of course we didn’t. I think Brooks' opinions will age poorly, but if anyone doesn't already know all the arguments for that they aren't interested in learning them now. This quote see…

> we would have an iPod with 160TB of storage by 2017, but of course we didn’t.

160 Tera Bytes to store what?

I probably would never listen to more than a few GBs of high fidelity music in my life time.

Why would anyone keep investing in exponential growth or even linear growth beyond a point of utility and economic sense.

This can be extrapolated to miniaturization also ... why is a personal computer not smaller than my fingertip already?

Re: Rodney Brooks on limitations of generative AI

#24

Post by aurareturn » Sun, Jun 30, 2024, 9:16 AM UTC

lispm wrote: post #21

aurareturn wrote: post #15

Earlier quoted context omitted.

I’m still waiting for SalesForce to integrate an LLM into Slack so I can ask it business logic and decisions long lost. Still waiting for Microsoft to integrate an LLM into outlook so I can get a summary of a 20 email long chain I just got CCed into. I don’t think the iPod comparison is a valid one. People only have so much time to listen to music. Past a certain point, no one has enough good music they like to put i…

> However, the more data you feed into an LLM, the smarter it should be in the response. Is it that way? For example if it lacks a certain reasoning capability, then more data may not change that. So far LLMs lack useful ideas of truth, it will easily generate untrue statements. We see lots of hacks how to control that, with unconvincing results.

That has not been my experience with GPT4 and GPt4o. Maybe you’re using worse models?

The point is that the more context an LLM or human has, the better decision it can make in theory. I don’t think you can debate this.

Hallucinations and LLM context scale are more engineering problems.

Re: Rodney Brooks on limitations of generative AI

#25

Post by richrichie » Sun, Jun 30, 2024, 9:18 AM UTC

globalnode wrote: post #12

i dont know much about machine learning but what i think i know is that its getting an outcome based on averages of witnessed data/events. so how's it going to come up with anything novel? or outside of normal?

ML is (smooth) surface fitting. That's mostly it. But that is not undermining it in any way. Approximation theory and statistical learning have long history, full of beautiful results.

The kitchen sinks that we can throw at ML are incredibly powerful these days. So, we can quickly iterate through different neural net architecture configurations, check accuracy on a few standard datasets and report whatever configuration that does better on Archiv. Some might call it 'p-hacking'.

Re: Rodney Brooks on limitations of generative AI

#26

Post by threeseed » Sun, Jun 30, 2024, 9:19 AM UTC

aurareturn wrote: post #22

richrichie wrote: post #20

Earlier quoted context omitted.

> the more data you feed into an LLM, the smarter it should be in the response This is not obvious though.

It’s in theory. The more information you have, the better the decision in theory.

It's quality not quantity.

You need to have accurate, properly reasoned information for better decisions.

Re: Rodney Brooks on limitations of generative AI

#27

Post by michaelt » Sun, Jun 30, 2024, 9:22 AM UTC

globalnode wrote: post #12

i dont know much about machine learning but what i think i know is that its getting an outcome based on averages of witnessed data/events. so how's it going to come up with anything novel? or outside of normal?

Hypothetically?

Training data gives the model an idea what "blue" means, and what "cat" means.

It can now generate sensible output about blue cats, despite blue cats not being normal.

Re: Rodney Brooks on limitations of generative AI

#28

Post by akerr » Sun, Jun 30, 2024, 9:24 AM UTC

aurareturn wrote: post #24

lispm wrote: post #21

Earlier quoted context omitted.

> However, the more data you feed into an LLM, the smarter it should be in the response. Is it that way? For example if it lacks a certain reasoning capability, then more data may not change that. So far LLMs lack useful ideas of truth, it will easily generate untrue statements. We see lots of hacks how to control that, with unconvincing results.

That has not been my experience with GPT4 and GPt4o. Maybe you’re using worse models? The point is that the more context an LLM or human has, the better decision it can make in theory. I don’t think you can debate this. Hallucinations and LLM context scale are more engineering problems.

ChatGPT says, “Generally, yes, both large language models (LLMs) and humans can make better decisions with more context. … However, both LLMs and humans can also be overwhelmed by too much context if it’s not relevant or well-organized, so there is a balance to be struck.”

Re: Rodney Brooks on limitations of generative AI

#29

Post by akerr » Sun, Jun 30, 2024, 9:26 AM UTC

aurareturn wrote: post #24

lispm wrote: post #21

Earlier quoted context omitted.

> However, the more data you feed into an LLM, the smarter it should be in the response. Is it that way? For example if it lacks a certain reasoning capability, then more data may not change that. So far LLMs lack useful ideas of truth, it will easily generate untrue statements. We see lots of hacks how to control that, with unconvincing results.

That has not been my experience with GPT4 and GPt4o. Maybe you’re using worse models? The point is that the more context an LLM or human has, the better decision it can make in theory. I don’t think you can debate this. Hallucinations and LLM context scale are more engineering problems.

“Yes, it is debatable. Here are some arguments for and against the idea that more context leads to better decisions…”

Re: Rodney Brooks on limitations of generative AI

#30

Post by latexr » Sun, Jun 30, 2024, 9:28 AM UTC

roenxi wrote: post #9

> He uses the iPod as an example. For a few iterations, it did in fact double in storage size from 10 all the way to 160GB. If it had continued on that trajectory, he figured out we would have an iPod with 160TB of storage by 2017, but of course we didn’t. I think Brooks' opinions will age poorly, but if anyone doesn't already know all the arguments for that they aren't interested in learning them now. This quote see…

> I think Brooks' opinions will age poorly, but if anyone doesn't already know all the arguments for that they aren't interested in learning them now.

Or they are too young still, or they just got interested in the subject, or, or, or…

https://xkcd.com/1053/

Don’t dismiss someone offhand because they disagree with you, they may really have never heard your argument.

> AFAIK, looking at Wikipedia, they're discontinuing iPods in favour of iPhones where we can get a 1TB model and the disk size trend is still exponential.

iPods were single-purpose while iPhones are general computers. While music file sizes have been fairly consistent for a while, you can keep adding more apps and photos. The former become larger as new features are added (and as companies stop caring about optimisations) while the latter become larger with better cameras and keep growing in number as the person lives.