Has Enshittification Finally Reached OpenAI?
Real examples of how OpenAI biting off more than they can chew is affecting users.
It has been gradual and often celebrated as progress, but there’s less and less trace of that AGI spark that Microsoft touted during their GPT-4 evaluation before it reached us all. We have new models, more tools available to everyone, but those of us who were subscribed, have we possibly ended up on the losing end?
Look, I’ve been a paying OpenAI user practically since the option became available. And I was truly delighted. It’s true that GPT-3.5 gave me the impression of being a student who didn’t know the answer but knew how to waffle, but GPT-4 seemed to read my intentions and, despite a few errors and a somewhat limited quota, was excellent. On the other hand, I started with StableDiffusion on my own machine, but when DALL·E 3 came out, just like GPT-4, it also seemed to read my intentions, with impressive quality and detail.
And yes, recently, GPT-4o came out, overcoming the GPT-4’s quota limitation. And although many say that GPT-4o is just as good —or even better according to LMSYS, — I feel that we don’t…
