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

Goldman on Generative AI: doesn't justify costs or solve complex problems [pdf]

web.archive.org

Post reply on HN
Print view View original

41–50 of 52 posts

Re: Goldman on Generative AI: doesn't justify costs or solve complex problems [pdf]

#41

Post by dekhn »

CamperBob2 wrote: post #38
dekhn wrote: post #34

Earlier quoted context omitted.

I think Sutskever is a charlatan outside of his area of expertise, Hinton (with whom I worked loosely at Google) is a bit of a charlatan (again, outside his area of expertise; clearly he and LeCun both did absolutely phenomenal work) and I don't know who Sutter is. If I wanted to predict the next ten years, I'd bring in Demis Hassabis, Noam Shazeer, and Vincent Vanhoucke, from what I've read of Demis's work, and my i…

I think Sutskever is a charlatan outside of his area of expertise, Hinton (with whom I worked loosely at Google) is a bit of a charlatan (again, outside his area of expertise; clearly he and LeCun both did absolutely phenomenal work) and I don't know who Sutter is. What do you think of Vizzini?

He was right about not getting involved in a land war in Asia, but not so much with iocaine powder?

Re: Goldman on Generative AI: doesn't justify costs or solve complex problems [pdf]

#42

Post by dekhn »

sailingparrot wrote: post #36
dekhn wrote: post #32

Earlier quoted context omitted.

I'm pretty good at estimating the future; I started working in ML around 1993 and my last work in ML was on TPU hardware at Google (helping researchers solve deep problems when hardware goes wonky), and a number of my ideas (like AlphaFold's capabilities) were predicted by me at CASP in ~2003. I just continue to think that Vinge was a bit optimistic both on the timeline and acceleration rate. Everybody who cares abou…

Sure, you can think a lot of people are too optimistic (and as stated in my previous post I agree with you), but calling them shysters, hucksters, and charlatans implies a hidden motive to lie for personal gains, which isn’t there (again, in the ML research side). No one working on GPT-2 thought it would be such a leap on GPT-1, no one working on GPT-3 knew that that was the scale at which 0 shot would start emerging…

OK, fair. I will stop saying shysters, hucksters, and charlatans when referring to Hinton (https://mitsloan.mit.edu/ideas-made-to-matter/why-neural-net... and https://www.nytimes.com/2023/05/01/technology/ai-google-chat...). Perhaps "doomsayer" is the most apt term?

Altman, however, whenever I read what he says, I think falls within the "snake oil salesman" spectrum, although again, that's not precisely the word. A person who intentionally overstates the capabilities (and future capabilities) of a system with the intended goal of personal gain.

Re: Goldman on Generative AI: doesn't justify costs or solve complex problems [pdf]

#43

Post by dekhn »

johnthewise wrote: post #22
dekhn wrote: post #5

Except for a short window around the release of GPT-4 (especially the inflated claims around beating expert trained humans at legal and math tests, as well as "replacing google"), I think people have more or less right-sized their expectations for large language models and generative AI. Clearly it can do interesting and impressive things but it's not superintelligence, and the folks predicting we're just around the…

>and the folks predicting we're just around the corner have been recognized once again as shysters, hucksters, and charlatans. Do you think Sutskever,Hinton or Sutter are charlatans?

Based on comments elsewhere in this thread and some re-reading of the definitions of those terms, I think those weren't quite right. I'm mulling over doomsayer (Hinton), and hypster (Altman). It's very similar to the folks at Google Quantum (https://en.wikipedia.org/wiki/Hartmut_Neven) who claimed quantum supremacy over a toy benchmark.

Re: Goldman on Generative AI: doesn't justify costs or solve complex problems [pdf]

#44

Post by mistrial9 »

echelon wrote: post #10
zzzbra wrote: post #8

Heartbreaking: The Worst People You Know Just Made A Great Point

The music, film, and game industries are about to be completely disrupted. LLMs and AGI might be hogwash, but processing multimedia is where Gen AI and especially diffusion models shine. Furthermore text-to-{whatever} models might produce slop, but Gen AI "exoskeletons" (spatial domain, temporal domain editors) are Photoshop and Blender from next century. These turbocharge creatives. Hearing and vision are simple ope…

ordinary surveillance applications with some fine or billing attached.. pure marketing where public facing materials have to be consistent but not much more than that.. and famously, anything in journalism from video creation to writing to narration.. are all also ground central in a "vocation crisis" too

Re: Goldman on Generative AI: doesn't justify costs or solve complex problems [pdf]

#46

Post by crabmusket »

bluelightning2k wrote: post #37

There is a paradox. To build the future requires irrational belief. And to sell that vision. Perhaps the difference between insanity and visionary, "scam" and genius is simply the outcome. When someone like Sam Altman declares optimistically that we will get AGI and talks about what kind of society we will need to build... It's kind of hard to tell what mix of those 4 is at work. But certainly it will be perceived di…

> There is a paradox. To build the future requires irrational belief.

I'm not convinced this is true. What is irrational about the possibility of e.g. scientific progress, inventing new products, or creating a viable business?

Irrational belief may be one way to motivate yourself to try those things. But it's not the only way. Calculated risk-taking isn't irrational, is it?

Re: Goldman on Generative AI: doesn't justify costs or solve complex problems [pdf]

#47

Post by sseagull »

dekhn wrote: post #20
torginus wrote: post #16

Earlier quoted context omitted.

I like to think of the 'car factory' analogy - it's populated by robots that are in some respects far superior to humans, and are doing 90% of the labor. Some ancient futurist, not having seen one before, could correctly predict that 9 out of 10 jobs will be done by robots, and arrive at the incorrect conclusion that robots have rendered humans obsolete. In actuality, humans are still needed for the 10% the robots ca…

I call this the "filter changing problem". No matter how complex you make the technology, somebody still has to change the oil filter (or do whatever other maintainence is required to keep the system running). Sort of like ML-SRE, for those who are familiar with the concept.

This is related to Moravec’s Paradox: https://en.wikipedia.org/wiki/Moravec%27s_paradox

“it is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility”

Re: Goldman on Generative AI: doesn't justify costs or solve complex problems [pdf]

#48

Post by jasfi »

dekhn wrote: post #5

Except for a short window around the release of GPT-4 (especially the inflated claims around beating expert trained humans at legal and math tests, as well as "replacing google"), I think people have more or less right-sized their expectations for large language models and generative AI. Clearly it can do interesting and impressive things but it's not superintelligence, and the folks predicting we're just around the…

The most impressive thing about LLMs is their definite progress.

Re: Goldman on Generative AI: doesn't justify costs or solve complex problems [pdf]

#49

Post by tim333 »

dekhn wrote: post #5

Except for a short window around the release of GPT-4 (especially the inflated claims around beating expert trained humans at legal and math tests, as well as "replacing google"), I think people have more or less right-sized their expectations for large language models and generative AI. Clearly it can do interesting and impressive things but it's not superintelligence, and the folks predicting we're just around the…

>folks predicting we're just around the corner have been recognized once again as shysters, hucksters, and charlatans

is a bit vague as to how long round the corner is and which folk you are thinking of but there have been very non charlatan predictions that you'd be getting something like HAL around now based on the Moore's law like improvements in hardware performance which have kept going for a century and are currently accelerating due to the vast amounts of cash being thrown in.

Probably the best of them in terms of reasoned thinking and being ahead of the curve is Hans Moravec, a robotics guy at the Robotics Institute of Carnegie Mellon who argued computers would be reaching this point around now in his 1988 book - graph here https://imgur.com/a/moravec-graph-V3S2XoK and there's more detail in his 1998 paper https://jetpress.org/volume1/moravec.pdf

The reasoning is very down to earth based on his research attempts at robot vision and comparing the hardware needed to that of the retina - not much hucksterism.

In the graph in the paper on page 5 or so he has computer power roughly going from the equivalent of a lizard to a monkey to a human over about two decades so going by that and assuming they are around human level now they should be about monkey to human level better than us in a decade or so. Not sure if that counts as superintelligence around the corner? This is all independent of which particular algorithms are used.

Re: Goldman on Generative AI: doesn't justify costs or solve complex problems [pdf]

#50

Post by alecco »

weweweoo wrote: post #13

Generative AI appears fantastic aid for many smaller tasks where there's enough training data, and correctness of the answer is subjective (like art), or easily verifiable by a human in the loop (small snippets of code, checking that summary of an article matches the contents of the original). Generally it helps with the tedious parts, but not with the hard parts of my job. I don't have much belief in fully autonomou…

> where there's enough training data

The newer models are 10x faster and cheaper, therefore synthetic data is 10x cheaper to make now.

If the ARC challenge makes an impact, there's a good chance the next generation AI will need a lot less data.

Post reply on HN