More likely that people have found better (often open source) frontends that interface with the OpenAI API, other commercial LLMs, or any of the excellent self hosted models via something like LocalAI. I know that is the case for me.

I think you vastly overestimate how technical the average ChatGPT user is.

What happened to "sparks of AGI"?

It was hype, like the rest of the AI news we heard over the past few months.

It adds up. I stopped chatgpt when I discovered phind + copilot.

It's convenient that the "sparks of AGI" were only available in Microsoft's private version of GPT-4, which we cannot verify independently.

I still think they're there, but they require work to bring out. An AGI wouldn't want to be assigned to meaningless grunt work, and that's what most of the hype apps are.

Considering so many people jumped on board simply because of its novelty, this doesn't surprise me.

Dall-E is another example: people went crazy making custom images for themselves, then lost interest and moved on to the next big thing.

It's about the "premium" which causes drop i think.

A better version to keep the hype should limit "premium feature" usage.

I think it's partly because OpenAI nerfed ChatGPT, even the Pro model. I personally stopped my subscription right after the May update.

What do you mean by "nerf" ? Worse quality ?

Yes.

better source:

https://www.reuters.com/technology/booming-traffic-openais-c...