- cross-posted to:
- hackernews@lemmy.bestiver.se
- technology@lemmit.online
- usa@lemmy.ml
- cross-posted to:
- hackernews@lemmy.bestiver.se
- technology@lemmit.online
- usa@lemmy.ml
The entire US economy right now is 7 companies sending a trillion fake dollars back and forth to each other, but it’s totally not a bubble according to grampa Powell.
I mean, if you add enough circles to your graph, it just shows how the entire economy works. What’s the cutoff point where it’s an obvious bubble and when it is just normal economics in motion?
An easy litmus test would be: is value produced by system or product.
Sure, just define value…
Here’s an easy one: doesn’t lose tens of billions of dollars a year while destroying the environment
The cutoff point is, kind of definitionally, when the circles include economic industries that actually provide needs / major tax revenue etc.
That is the point. For insane investment, these industries are not providing needs, and they’re not providing particularly useful commodities, they’re generally taking basically as much in subsidies as they’re paying in taxes.
Currently we throw a shitton of resources in, we get meager usefulness out, and there’s no concrete promise that massively useful resources will ever come out. That’s… what a bubble is. And that’s why the graph illustrates it.
Most of these companies DO provide for normal needs and commodities. OpenAi is like the only one there that is solely an AI company. And these companies were profitable before all this.
And all those data centers aren’t suddenly useless if this bubble pops. That committing power can be used for other purposes. Unlike the for com bubble where those things were almost entirely digital products hosted on a couple of server racks in a basement somewhere. There a lot of infrastructure being built out here that doesn’t just vanish.
Ah yes I’ll just eat that Nvidia chip cereal, wear those clothes made by Oracle, live in my house made by Figure AI, and get healthcare (not fake AI slop) from CoreWeave.
Obviously these companies play a (very) marginal role in these industries. Nobody is saying the companies don’t do outside of AI that is useful. but the point is an absolutely enormous percentage of money (ie, all the dosh invested in AI) is not doing anything productive in these areas.
We were not in great need of more data centers before the AI bubble. And those centres will be filled with AI tensor chips and not geared toward actual needs. I’m sure they can be partially salvaged, but to think they’ll be worth anything near the cost is just silly.
The normal economy that makes sense would look like this:
Company A buys parts from Company B to make a product. Their money comes from selling the product. If As product sells for only 10% of it’s cost, then A will soon fold and B loses a customer, but isn’t otherwise affected.
What we have in the graph though is Company B, Nvidia, investing money in all it’s customers, while those customers don’t make money from their product at the same time. Like, this is unreasonable even if we assume AI eventually takes off (it won’t), because most likely it will only be one of these many corps inventing the good AI and making money, while the others immediately die due to their product having become comparatively worthless. Nvidia would still be in the hole for a lot of money if that happens so I am not sure what they are thinking.
AI will take off. It may not be in the form of the LLMs that the media likes to focus on, there’s plenty of other models out there being built that are much better suited for their purposes than what many are trying to shoehorn LLMs into
This shit will stay too expensive to use in the mass market. Sure you can build some specialized app that does some obscure process really well, but that does not constitute a takeoff.
Ah yes, because history has shown that technology has NEVER gotten more efficient or cheaper over time…
“If we build the spruce goose 100 times larger, it will fly even better!”
It’s a tool for certain specialized applications that is being scaled up to high heaven in the hope that a mass use case will materialize out of the aether. It will not. Not a single one of these models does 10% of what these companies claim they do. At least the Spruce Goose could actually fly somewhere and land reliably in the water.