Goldman-Sachs warns that AI could be a bad investment

Originally published at:


How is that a change in viewpoint? Seems totally consistent (and not at all unheard of) but focusing on a different question.

  1. GenAI could significantly increase productivity growth and the size of the global economy.

  2. Investors in the companies developing this technology are unlikely to capture those gains because of the large capex amounts required.

In other words, this is basically saying that GenAI is both a generator of large positive externalities, and a transfer of wealth from investors to the rest of the economy.

Also: The article seems to be forward looking, about investments in “data centers, chips, other AI infrastructure, and the power grid.” Anyone who thinks those will generate returns at software company speeds is being unreasonable, and should be told so.


At least they admit it is a bubble. :man_shrugging:

(Seriously, though, it probably isn’t, but it may well have much more extreme ups and downs as it proceeds along the Gartner maturity curve. Probably curves, plural.)


No worries. A.I. financiers will just invent fusion and then - underpants for everyone!


Machine learning biscuits?


I’d be morbidly curious to know how much of the ‘AI’ hype boils down to the fact that it would be a prerequisite for really mopping up the class wars; rather than a sincere belief that it’s particularly profitable in petty, grubbing, ROI terms.


So smart people paid to be smart see most prolific uses of AI are being nonstop dribble producing machines?

Can’t say that CS doesn’t try to see ahead of the curve


They do include a graphic for that. See “Labor displacement” and the much smaller “Redeployment of displaced workers”


Total BS and they know it.

In my work developing AI for business, I’ve seen firsthand that the consumer-facing AI hype is mostly fluff. However, the real innovation is happening behind the scenes. Companies are using AI to automate tedious tasks and manage data more efficiently, a necessity given the tech illiteracy rampant among the white-collar workforce. The over-reliance on clumsy tools like PDFs and spreadsheets only compounds the problem (oh the horror stories I could tell).

Businesses aren’t keen to share their advances, as every industry grapples with these late 90s boomer-era bottlenecks. The real AI revolution is happening internally, transforming white-collar work from within. AI now serves as a strategic advisor, aiding in everything from business acquisitions to marketing research. The future lies in these quiet but significant shifts, reshaping industries from the core.


The problem is that the garbage spewed by the garbage spewing machine is garbage. Until that is solved, the investment in the tool won’t be justified by the return. The resources (electricity, water) needed for a ChatGPT “search” is significantly greater than a regular “search”, and returns worse results.

For those who say this will be solved in time, I ask: “where’re the self driving cars Elongated Muskrat had been promising since 2016?”

I’ve been waiting. I’ll wait some more.

In the meantime, AI is the new blockchain. A solution in search of problems which have already been mostly solved — efficiently — over the past centuries.


They’ll be out in 2 years, duh (/s)


Self driving cars would be in first grade by now if they’d appeared on Muskrats original schedule.


They’ll probably be out about the time they’re old enough to drink. And drive like they’ve been drinking.


The consumer market is not the correct bellweather to watch with Ai. Upwards of two thirds of the economy is business to business (B2B).

When you take a base model like gpt, and train it exclusively on industry specific data and information the results are astounding for how much time it saves on mundane white collar tasks.


I’m here for the fused underpants.


All this is lovely, but about the only success stories I’ve seen with LLMs and similar come from competent, knowledgeable people training and teaching the ‘AI’, GIGO is still very applicable, as LLMs don’t know shit from shinola.


The late 90’s wasn’t GenX era?


I think that GS just figured out that a well trained investment AI will probably outperform GS, without the need to pay commissions.


It’s also the new blockchain in the sense that it’s the newest Hot New Thing That Will Change Everything (And Make Us Obscenely Rich Any Moment Now).

I do not for a second believe it’s just a coincidence that the AI hype started soon after blockchain, cryptocurrencies and NFTs were exposed as so shitty and scammy that nobody who wasn’t already eyeball-deep in them is going to invest in them.


Exactly. Which is why@terminal is correct to note where the real action is happening - internally in corporations, training their own AIs (largely in a B2B environment). If an LLM had been around when I was at work I’d have trained it on my very assorted and extensive knowledge (language) assets and it would have automated much of the ‘grunt’ work aspects of what I did, but would not have replaced me, I’d have been hugely more productive (and profitable) as a result. Without me as the human to manage its outputs it would have likely resulted in failed projects.