Goldman Sachs Global Macro Research: Gen AI: too much spend, too little benefit [pdf] (2024)

Posted by u1hcw9nx 2 days ago

Counter38Comment20OpenOriginal

Comments

Comment by alexc05 2 days ago

can you post the correct link in the comments?

this appears to be a redirect away from static assets.

looks like you've got to go here or their systems will mess you around: https://www.goldmansachs.com/insights/top-of-mind/gen-ai-too...

Comment by 2 days ago

Comment by biophysboy 2 days ago

The recent JP morgan "Eyes on the Market" report on the AI boom is a better alternative to this.

Comment by deadbishop 2 days ago

From 2024

Comment by piker 2 days ago

June 2024

Comment by ares623 2 days ago

Phew. Things are completely different now.

Comment by lostmsu 2 days ago

I take Goldman Sachs reports like this as a strong signal to buy.

Comment by 2 days ago

Comment by gyanchawdhary 2 days ago

The banker wankers got it completely wrong

Comment by nateglims 2 days ago

The title is click bait. 3 of the 5 people interviewed were optimistic and one of the pessimists is an MIT professor, not a goldman analyst. The rest of it is market outlooks from 2024 for chips and power that don't seem that far off.

Comment by gyanchawdhary 2 days ago

yeah, no.

Here is the original quote from their head of equity stratigy ..

AI technology is exceptionally expensive, and to justify those costs, the technology must be able to solve complex problems, which it isn’t designed to do. - Jim Covello

“complex problem” isnt an objective standard, and some rando GS "stratigist of equities" certainly doesn’t get to define it .. reality already disagrees with him .. and plenty of real usecases show AI solving interesting non-trivial problems ..

Comment by nateglims 1 day ago

Reality doesn't disagree with him in that the ROI is still not here yet. Maybe the coming replacement of software engineers will start to ramp up equity gains due to productivity, but it's real hard to get there by replacing your receptionists and lowest tier of support personnel.

Comment by tliltocatl 2 days ago

> plenty of real usecases show AI solving interesting non-trivial problems ..

Such as? Generating tons of spam? Generating tons of boilerplate code (which shouldn't have been necessary if the industry haven't been valuing coders' fungibility above development time)? Adult content is probably the only usecase so far that isn't automating away what should never be done in the first place.

Comment by palmotea 2 days ago

> Adult content is probably the only usecase so far that isn't automating away what should never be done in the first place.

AI models are making porn. I there are "nudify" apps (basically undress social media pics) and a rush to ban them.

Comment by gyanchawdhary 2 days ago

you're conflating “usecases I don’t like” with “no usecases”

Comment by tliltocatl 1 day ago

It's not about whatever I like them or not. It is about "will it pay for itself"?

Comment by gyanchawdhary 1 day ago

Nah, that’s not how you started with your response. And regardless you and that articles thesis is playing out wrong ;)

Comment by badgersnake 2 days ago

Nah, seems increasingly accurate. Spend keeps going up, successful projects in production not so much.

Comment by gyanchawdhary 2 days ago

lack of visibility != lack of deployment

Comment by badgersnake 1 day ago

Companies all quietly agrees to stop talking about their successes. Right.

Comment by gyanchawdhary 1 day ago

Yes. Right.lol

Comment by 2 days ago

Comment by lotsofpulp 2 days ago

They’ve gotten it wrong since 2008. GS couldn’t even figure out the credit card business.