Idk if I’m weird on this, but I rarely use LLMs as search engines. Their memorization of facts is too iffy imo. Prefer duck duck go for fact finding. I do use LLMs a ton in lieu of talking to experts about something I’m thinking about, or asking them to write code or other text for me. If it’s a questionable topic or answer I’ll ask it to critique itself as a new personality. It’s already much rarer than before, but I would be unsurprised if I stop needing to hire humans to benefit from their expertise within a few years. Regulated industries will be the hold out.

Are you saying that the openAI API costs don’t actually cover the full price of each run of the model? Or what do you mean by subsidized?

Reply to this note

Please Login to reply.

Discussion

I normally use it to give me outlines for topics I want to research or to explain certain concepts to me like I'm 5

While the article below could be a little hyperbolic, I think there is some truth in the burn rate to run ChatGPT, will they be able to remain the popular option with other companies also offering alternatives and diluting market share?

How many users and companies are paying for it currently and whats their churn?

Do they reach a point where they can reduce cost per query to a point where revenue covers it? How long before they reach that goal?

Subsidize as in previous rounds of investment covers cost not revenue allone, and how many more times can it raise? Will it need to IPO to continue since it's already in the multi-billion rage of valuation?

https://www.wionews.com/science/openai-may-go-bankrupt-by-end-of-2024-costs-700k-to-operate-chatgpt-daily-report-625016

https://www.forbes.com/sites/lutzfinger/2023/08/18/is-openai-going-bankrupt-no-but-ai-models-dont-create-moats/?sh=3ee768125e22

“1.7 billion users in June”

Haha wow. That’s so crazy