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

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 corner have been recognized once again as shysters, hucksters, and charlatans. It doesn't help that state of the art ML researches have gotten so good at over-hyping the actual abilities of their technology.

However, I do think we'll continue to see impressive advances in the areas of media consumption and production, with complex reasoning on hard problems being a likely area of improvement in the near (1 decade) future. While I once never expected to see something like HAL in my lifetime, I feel that many aspects of HAL (voice recognition, ship automation, and chess-playing) have been achieved, if not fully integrated into a single agent. We can expect most applications to be banal- the giants who have the largest data piles will train models that continue to optimize the addictivity of social media, and click-thru rates of ads.

I am also quite impressed at the recall of information by language models for highly factual and well-supported things (computer reviews in particular).

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

#6

Post by dekhn » Fri, Jul 05, 2024, 8:59 PM UTC

great_psy wrote: post #4

Is there more than the cover to this ? On mobile I only see one page of the PDF.

Yes, here's the original URL which should download a full multi-page PDF: https://www.goldmansachs.com/intelligence/pages/gs-research/...

It's mostly speculative narrative with a fair number of data-driven charts. I wouldn't spend much time on it unless you like financial analysis with hand-waving.

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

#7

Post by xmichael909 » Fri, Jul 05, 2024, 9:00 PM UTC

This interview by Adam Conover, really is a wonderful discussion on the topic https://www.youtube.com/watch?v=T8ByoAt5gCA I was pretty amazed with GPT when it came up, but increasingly find it makes to many mistakes. I full use it as a tool in writing code, but it needs to be treated as Intellisense plus, or something to that affect, not something that will handle complex tasks. GPT and Claude make many mistakes and unless they can solve it from completely making up stuff (which I don't think they can) will not advance much more beyond waht they currently are.

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

#8

Post by zzzbra » Fri, Jul 05, 2024, 9:07 PM UTC

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

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

#9

Post by karaterobot » Fri, Jul 05, 2024, 9:10 PM UTC

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…

I agree with what you say above, but my perception is that most people still view the current crop of models as a step or two away from superintelligence. That superintelligence, or AGI, is a matter of continued improvement along the current lines, rather than along entirely different lines.

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

#10

Post by echelon » Fri, Jul 05, 2024, 9:17 PM UTC

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 operations relative to reasoning. They're naturally occurring physical signals that the animal kingdom has evolved, on several different occasions, to process. This is likely why they're such a low hanging fruit to replicate with Gen AI.