Comment by dgfitz
2 months ago
I am so sincerely amused that “we” figured out how to monetize LLMs from the jump using tokens.
It isn’t tech for techs sake, it’s a money grab. Reminds me of paying to send a text message or buying minutes for a phone plan. Purely rent-seeking.
Can you explain how this is rent seeking? It seems to be straightforwardly not rent seeking.
1. Company develops model, invests in research, hardware, and software.
2. Company sells access to the model.
(1) is the step that makes this not rent seeking.
Rent seeking is when you profit from something you didn't earn - land rent, monopoly profits, protectionism.
That’s fair. My thought was, when there is an interesting new technology, it usually takes time to figure out how to monetize it. Figuring out how to monetize LLMs took no time at all.
"GPT 1.0" was released in 2018, I think that's a decent amount of time.
2 replies →
It's likely that no one who makes base models is currently making money from LLMs. More likely losing it at a crazy rate.
These prices are almost certainly "introductory offer" prices to get people/devs to integrate AI into their lives/workflow/product.
In a few years is when we will see what the actual cost is.
I don't think it's obvious that any of these model providers are even profitable right now. I'm also not sure what there is to "figure out" - it's an expensive technology where the cost scales per token, so they charge per token? would you rather they burned even more money giving it away for free until everyone was dependent on it and then hyper enshittified to try and not go broke like so much of the rest of tech?
My point, poorly made, was that I can run it myself for “free” without caring about tokens at all. Tokens are an artificial construct.
That's not remotely correct- how would you be able to ignore tokens? That's literally what defines the context size of an llm, the larger the more memory. Generating each token is literally the compute you're doing. Your hardware limits how many tokens per second you can produce with a particular model. It's literally what you're consuming electricity to produce. That's like saying you can drive your own car for free without worrying about how many miles you've driven.
So by that logic all VPS providers are just a money grab because you can run your software yourself for “free” without having to pay for that artificial construct these greedy people call “compute?”
I don’t understand your point. You’re using a resource. You’re wasting time on the GPU of someone else. That chunk is called a token. And that’s what you’re being billed.
1 reply →