No, but the GP wasn't satisfied with that, and had to put in a snide "even on inference" parenthetical. The leaks showed inference having positive margins.
The Zitronites will say that the data is fraudulent, and OpenAI must have classified some of their inference as marketing, or R&D, or some other wacky theory of the week. But the actual data does not show that. It is made up.
If you want to cherry-pick the worst parts from the leak and disbelieve the more positive ones, it feels like you're not in a great place epistemically...
The inference part requires training the models to keep up to date as far as I am aware(I am open to being convinced otherwise if you have data showing so). Everything I’ve read about it implies that within a few years of reality updating or even just software if we constrain it to models for coding, won’t be able to handle all of the updates within the context windows available.
Given that I don’t know how you(royal you, not you specifically) can constrain an analysis of inference profitability to just the literal running of inference and not include training costs. That’s where it all breaks down for profitability.
> The leaks showed inference having positive margins.
They don't. They show that OpenAI need to people to draw that conclusion, because of course they do.
> The Zitronites will say that the data is fraudulent, and OpenAI must have classified some of their inference as marketing, or R&D, or some other wacky theory of the week. But the actual data does not show that.
It doesn't?
It shows a marketing budget so absolutely mahoosive that it's almost completely implausible, which does make you think — have a percentage of marketing-driven free plan tokens been hidden in there? If not, what the hell is in there? Because it's an insane figure for a company that has benefited from a level of word of mouth that makes
"ChatGPT" broadly synonymous with "AI".
Fraudulent is a big claim, of course. I didn't say it.
> Zitronites will say that the data is fraudulent, and OpenAI must have classified some of their inference as marketing, or R&D, or some other wacky theory of the week
Which, look, could be true! But it's currently speculation only.
How do you explain OAI spending $6B on “sales and marketing” in a year. More than Coca Cola?
I think it’s reasonable to draw the conclusion that they are folding inference subsidies (for both paying and non-paying users) under this category. Frankly I think occam’s razor demands it because where else would all that money have gone? Fancy trips for enterprise clients?
The Zitronites will say that the data is fraudulent, and OpenAI must have classified some of their inference as marketing, or R&D, or some other wacky theory of the week. But the actual data does not show that. It is made up.
If you want to cherry-pick the worst parts from the leak and disbelieve the more positive ones, it feels like you're not in a great place epistemically...