I had a subscription for 3 months or so.
I haven't found ChatGPT to be all that useful. My primary use is in TTRPGs. Story hooks and very generic things sure. But if I ask for anything specific to the rules for one specific TTRPG, even giving it the text to reference, it still comes up with absolutely unreliable nonsense. A mashup of things from several different TTRPGs. Even when asked a direct question "what's the stat-block for a zombie in the PDF I just gave you?" And it responds with something completely made up.

I have both Proton and Kagi. Kagi offers access to most of the major LLMs. Proton has their own. They all work more less the same. ChatGPT is nothing special realy.

I wouldn’t mind paying a small subscription to cover compute costs and retain neutrality.

If Netflix, Amazon Prime, etc. are anything to go by, big tech has found it much more profitable to serve you ads (and sell your data) than take a pittance from you to avoid them. Don't hold out hope for anything like what you're describing.

I would be seriously surprised if it wasn’t happening already because it’s the only way I can imagine them making money. Say, Disney pays OpenAI so that overall output is 3% more positive which scales quickly with the number of ChatGPT users that mistake it for a search engine. That kind of marketing benefits from not being blatant.

I think people are letting their "Fuck AI" mindset (which I agree with) interfere with the meaning of your post.

AI (all of them) are currently in the "growth at any cost" phase. This is where they burn mountains (and apparently oceans) of cash to get their product in use everywhere they possibly can, and get people hooked on it. Most people around here think that it will never get beyond this point, but let's play it out anyway.

At some point, everyone selling it needs to start showing a profit. It's not a cheap venture, and it sure as fuck isn't being done as some service to humanity. It needs to pay massive dividends to recoup the massive investments.

You're thinking advertising and the like. I don't think that's big enough.

AI, collectively, is a multi-trillion dollar investment. This is built entirely on the promise of eliminating payroll for a fuck-ton of people, more than the trillions it's costing. Adopters (users) are also expecting to keep that "profit".

Doctorow's process of enshittification is that the current phase would be to get businesses hooked at the low price, and fire all of the people that currently do it. Of course, many will, since that is in their short-term interest.

The next phase is to then extract that value as well. I would expect AI to develop complex and expensive licensing models. If it's saving your company the salary of an $80k/year graphics artist, OpenAI/Anthropic/Microsoft/etc is going to want, and feel entitled, to charge $40k/year for it. And those same businesses will pay, because it's still cheaper than the real person.

TL;DR: The future of AI is AutoCAD.

when? its like smartphones and have never been that great.

I just run local and ignore the subscription tools.

claude.ai for text, krea.ai for media, and huggingface.co for everything else. Until the tides change again, anyway. The wave of enshittification is ever swelling.