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

Cargo Cult AI

queue.acm.org

Post reply on HN
Print view View original

111–120 of 191 posts

Re: Cargo Cult AI

#111

Post by sebzim4500 »

tjr wrote: post #92
loandbehold wrote: post #89

Earlier quoted context omitted.

The issue of hallucinations is overblown. I use GPT4 all the time and don't see any hallucinations at all. It's a big problem with Google BARD and GPT3 and earlier models. But GPT4 fixed the issue of hallucinations completely.

I wonder if it's "fixed" or if it's just less obvious. It seems that an LLM would "hallucinate" a bogus answer if it didn't actually have a good answer somewhere in its training. Is GPT4 so much more trained that it rarely encounters something it doesn't have a reasonable answer for? In which case, it would still "hallucinate" if cornered on some more obscure matter? I mean, like, what would it mean to actually solve…

In my experience, GPT-4 is equally willing to make things up if it doesn't know something but it has so much more knowledge than GPT-3.5 that this happens less often in practice.

Re: Cargo Cult AI

#112

Post by neatze »

_yb2s wrote: post #79

The whole premise of this article hinges on the idea that LLMs have fundamental limitations that they clearly don’t have if you’ve looked at lots of gpt4 examples. For example, it can do scientific thinking if you specifically ask it to, and it can reason about totally new situations outside of the training data based on generalizable models of reality it creates to predict training data. If you are certain these lim…

> clearly don’t have if you’ve looked at lots of gpt4 examples

for example, can you fine tune GPT to play chess at ELO 1600 ?

If you don't know answer, you are in for surprise.

Re: Cargo Cult AI

#113

Post by neatze »

fragsworth wrote: post #91
_yb2s wrote: post #79

The whole premise of this article hinges on the idea that LLMs have fundamental limitations that they clearly don’t have if you’ve looked at lots of gpt4 examples. For example, it can do scientific thinking if you specifically ask it to, and it can reason about totally new situations outside of the training data based on generalizable models of reality it creates to predict training data. If you are certain these lim…

The article was clearly written by someone who hasn't used GPT-4 extensively. "Current methods will not achieve AGI unless fundamental algorithmic innovations are introduced that enable AI to ask and answer questions of why." This is complete nonsense. GPT-4 is already close to being able to do basically everything. All you need is the obvious improvements - better prompts, multi-shotting, bigger context, and access…

This claim does not make sense, transformer networks in my limited experience are limited in there learning ability (fine tuning), furthermore there planning abilities are non-existent.

Re: Cargo Cult AI

#114

Post by Jtsummers »

esafak wrote: post #107
Jtsummers wrote: post #66

Earlier quoted context omitted.

Probably, I mean I first studied ANNs over two decades ago and had conversations with people about them prior to starting college in the 90s who had developed solutions with them in the 80s (obviously severely computationally constrained in those days compared to today). So still not magic. To be very blunt: If you believe that GPT and the like are magic, then you're not thinking clearly. You're blinded by the result…

ANNs in the 90s were nothing like this. They were not even like this a few years ago. And neither were HMMs. There is an emergent human quality to them because they have approached our abilities; the comparison is tenable, whereas before it was not.

Scale and topology. That's the difference between ANNs of the 90s and 00s and today. They're still based on the same fundamental principles and doing (modulo scale) the same fundamental things: classification, prediction, generation.

It is not magic, it has never been magic.

Re: Cargo Cult AI

#115

Post by RC_ITR »

sandworm101 wrote: post #61
RC_ITR wrote: post #45

Earlier quoted context omitted.

An interesting fallacy I see emerging is 'this system doesn't have animal-style intelligence, therefore it is lesser' The entire point of neural nets is removing the biases of animal intelligence and letting the computer brute force solutions during training. We are now learning the early stages of the amazing things that can lead to. From an outside perspective, it's not crazy to argue that symbolic reasoning is a c…

Why do you assume that the animal is also not brute forcing the problem? Brute forcing is the basis of evolution. The algorithm running inside the head of the mouse is the survivor of a million iterative generations as the species brute-forced the entire "get to the food" survival problem. I encourage those touting computers as something new to comprehend the concept of deep time, that no matter how many times you ru…

> Brute forcing is the basis of evolution.

The goal function of animal intelligence is very different from the goal function of model training.

> The algorithm running inside the head of the mouse is the survivor of a million iterative generations as the species brute-forced the entire "get to the food" survival problem.

That’s my entire point. There are problems beyond “get the food.” Said differently, the fact that computers aren’t good at “get the food” is not a strong criticism of computers. It just means their goal functions are different.

I didn’t say computers are “better,” just that they have a fundamentally different approach to problem solving that may end up being better. It will probably end up being complimentary! This isn’t either or; AI is a tool built by humans to expand their own capabilities.

> I encourage those touting computers as something new to comprehend the concept of deep time, that no matter how many of times you run simulations, the natural world has almost certainly run more.

And yet no animal figures out how to evolve wheels to move faster.

The entire point I’m making is that, sure it’s hard to make computers do certain things (like walk on two legs), but there are plenty of different avenues to achieve things.

I encourage evolution maximalists to think a lot more about how engineered and complex the world around them is.

Re: Cargo Cult AI

#116

Post by gyrovagueGeist »

Jtsummers wrote: post #66
esafak wrote: post #54

Earlier quoted context omitted.

Magic is that which can not be explained. Could you explain GPT-4's results if you saw them a few years ago?

Probably, I mean I first studied ANNs over two decades ago and had conversations with people about them prior to starting college in the 90s who had developed solutions with them in the 80s (obviously severely computationally constrained in those days compared to today). So still not magic. To be very blunt: If you believe that GPT and the like are magic, then you're not thinking clearly. You're blinded by the result…

"It doesn't stop being magic just because you know how it works" - Terry Prachett

Re: Cargo Cult AI

#117

Post by pmoriarty »

kerkeslager wrote: post #93
albertzeyer wrote: post #82

Earlier quoted context omitted.

> the idea that larger models can do more than smaller models, which should be a surprise to no one. Actually this was quite a surprise to a lot of people, since the whole race to scaling up began, with GPT2 or so. It was totally not obvious that you can scale up the model (and also training) and it would improve the performance. Many (most?) people thought there would be some limit, and we were close to that limit w…

You didn't read the post you're responding to, and shouldn't be responding. In fact, you didn't read the part of the post you quoted, where I said it should be a surprise to no one. But, unsurprisingly, the sort of people who stop reading at the first chance they see to correct something, are easily surprised, since actually understanding LLMs would require actually doing some nuanced reading.

Please consider HN's Guidelines[1] when replying.

In particular:

- Please don't comment on whether someone read an article. "Did you even read the article? It mentions that" can be shortened to "The article mentions that".

- Be kind. Don't be snarky. Converse curiously; don't cross-examine. Edit out swipes.

- Please don't fulminate. Please don't sneer, including at the rest of the community.

[1] - https://news.ycombinator.com/newsguidelines.html

Re: Cargo Cult AI

#118

Post by gatkinso »

celestialcheese wrote: post #15
gatkinso wrote: post #13

I think a lot of AI stuff is really cool and promising, but am dismayed by how impressed people are by it sometimes, especially the visual output of systems like DALL-E. Seems like machines are testing our intelligence, rather than the other way around.

Simple things that don't seem impressive to you, but are impressive to others, may be because you haven't experienced the first hand difficulty of doing that thing pre-transformers/GPT. For example, I get _unbelievably_ excited with knowledge extraction and question answering demos on PDFs. Why? Because i've built similar systems for over a decade and know how difficult it is to build on top of messy archival data. N…

fair, mostly I'm unimpressed with visual outputs - AI making 'art' etc. PDF extraction is extremely cool and useful and the results are fantastic. Fully agree re influencers. Seems endemic these days.

Re: Cargo Cult AI

#119

Post by gatkinso »

mrbungie wrote: post #30
gatkinso wrote: post #13

I think a lot of AI stuff is really cool and promising, but am dismayed by how impressed people are by it sometimes, especially the visual output of systems like DALL-E. Seems like machines are testing our intelligence, rather than the other way around.

Have you seen how the outputs of image generation tools evolved during mere 2-3 years? It went from the stuff of nightmares to actually pasable and consumable images pretty fast.

It's certainly gotten better since deep dream et al

Re: Cargo Cult AI

#120

Post by gatkinso »

burnished wrote: post #20
gatkinso wrote: post #13

I think a lot of AI stuff is really cool and promising, but am dismayed by how impressed people are by it sometimes, especially the visual output of systems like DALL-E. Seems like machines are testing our intelligence, rather than the other way around.

People are impressed because its black fucking magic my friend.

it does seem that way, and maybe it's better thought of that way. surely plenty of hn commenters are here to explain why its actually not magic...
Post reply on HN