Sam Altman:
A lot of people want local inference, and somehow they assume that the problem is solved now because we have open-source models, we’re going to be able to run them on our computers, and so we’re not going to need all the stuff in the data center.
But from a global allocation perspective, actually it’s quite terrible to have everybody, like, have a bunch of the world’s limited supply of DRAM sitting in their own computer and being used five percent of the time.
Like you want maximum utilisation in the cloud if you’re trying to deal with this very supply-constrained environment,
And somehow many people’s intuition is the opposite of that
TL;DR Guy selling product: “Here’s why owning your own product instead of renting it from me is a terrible idea”
It’s true what he says and not just about RAM but about anything you own.
My car isn’t driving most of the time. My drill isn’t drilling most of the time and my oven is off most of the time.
That’s a huge waste of resources in a macro economic kind of way.
You can’t argue with the correctness of his statement. You need other arguments. These arguments exist.
See, the thing is though, all of these things are designed with some kind of running hours in mind. If a car is used more, it’ll wear our faster so it’s either using it a lot in a short time or less in a long time, its utility ends up being the same so there’s no direct gain from not letting it sit idle. That’s why passive robotaxis are a pretty dumb idea.
But yes there are things that really only wear out from time spent existing. A pencil’s lifespan is purely defined from its use, a cast iron pan can pretty much be used indefinitely. Cars and drills are more like pencils imo. RAM might be more like a cast iron pan though, although trace erosion is still a thing.
Right, so the logical response would be to collectively own them. But that’s not what this parasite wants.
I like the idea of collectively owning the computing power!
Would be nice to be able to buy part of a datacenter instead of renting it.
Defending that an argument CAN be true in an extremely narrow subset is an example of a really stupid argument, and only highlights the stupidity of the original argument.
Here’s a counterargument: The resources wouldn’t be scarce if his company didn’t buy all of it for exetremely questionable gain. (Three years into the bubble, and AI compute is still being sold at a subsidized rate. You’d think if enough genuine demand existed to justify the buildout, you’d have to pay dearly for any AI use.