This video covers the topic very well (and I appreciate and recommend her channel overall).
The long and the short of it is, they fucked their product up. Most of us here are not terribly fond of AI, but let’s for a moment just take it for granted that it is a product that does a thing. OpenAI will swap models out on the backend based on your use case and if it detects certain patterns, also as a sort of protective guard rails in some ways. It’s also extremely unclear on billing and token usage (paraphrasing, I’m not intimately familiar with the details myself, watch the video). This is highly problematic if you’re trying to build a service on top of it. AI is extremely finicky and not having clear ideas on token usage limit or what models may be running on the backend can affect that. This has rightly made even the devout users upset at these choices.
And from what I understand in the video, this is not a case with other providers where you stick with exactly the specified model and token usage and billing are more clearly indicated.
So now not only are they fighting a losing battle against the profitability of the entire tech and service model, they have a spurned and dwindling customer base to contend with, too. They are turbo fucked. Still, no one can say how or when this all finally goes tits up, but at least it’s kind of funny.
House of El - AL: OpenAI’s $1 Trillion Bullsh*t Is Falling Apart
This video covers the topic very well (and I appreciate and recommend her channel overall).
The long and the short of it is, they fucked their product up. Most of us here are not terribly fond of AI, but let’s for a moment just take it for granted that it is a product that does a thing. OpenAI will swap models out on the backend based on your use case and if it detects certain patterns, also as a sort of protective guard rails in some ways. It’s also extremely unclear on billing and token usage (paraphrasing, I’m not intimately familiar with the details myself, watch the video). This is highly problematic if you’re trying to build a service on top of it. AI is extremely finicky and not having clear ideas on token usage limit or what models may be running on the backend can affect that. This has rightly made even the devout users upset at these choices.
And from what I understand in the video, this is not a case with other providers where you stick with exactly the specified model and token usage and billing are more clearly indicated.
So now not only are they fighting a losing battle against the profitability of the entire tech and service model, they have a spurned and dwindling customer base to contend with, too. They are turbo fucked. Still, no one can say how or when this all finally goes tits up, but at least it’s kind of funny.