The AI party is ending before it even began

[ergophobe]

GitHub Copilot started lowering usage limits mid May. Business Clients only receive 300 requests per month and Enterprise is 1,000. How exactly is AI supposed to take our jobs if we can barely even use it?

https://old.reddit.com/r/cscareerquestions/comments/1l799y0/the_ai_party_is_ending_before_it_even_began_the/

And in the comments:

OpenAI announced at their last earnings they would have to 40x their revenue in the next 4 years to become profitable.

And FTR because this drives me nuts, this comment is frequently stated and simply wrong:
"In a gold rush, the guy who gets rich is the one selling the shovels."

That wasn't true in California in the 1800s (see Bonanza King, by Greg Crouch) and it wasn't true in the internet Gold Rush (see Macromedia and Cisco).

So far it's true in the AI Gold Rush (see Nvidia), but the big winners will probably be pure AI-plays just like the big winners and losers in the internet gold rush (see Amazon, Google and MySpace)

[ergophobe]

Why OpenAI burns through billions

https://www.axios.com/2024/10/03/openai-investors-profit-money-costs

The interesting part (or the part that surprised me) is that OpenAI spends more money running the current model than building the new one. I had read that it was training the model that was so expensive, so I thought that what we would see would be an asymptotic improvement in models and that at a certain point upgrades would slow way down and companies would pull value out of the existing model once they had buried their competitors or achieved lock-in somehow with their users (non-exportable but critical info; see, "Enshitification")

But it doesn't look like that gets them to profit either.  That makes me think that over the short term, LLM chatbots get more expensive until the Linux version comes along, then development slows but it's good enough for what most people want.

[rcjordan]

Good, solid -but not sharpest- Local LLMs will eventually make it to retail use.