People seem to forget that this has been the MO of every new tech product for decades. Uber, GamePass, Netflix, etc.

Working as expected: get you used to the product, then jack up the prices. Or jack up token spend as the model is updated whether you like it or not.

Welcome to the post-AI world

easy.. you're wasting your life, and money by using ai garbage

this is... what the hell is this nonsense.

1 small sentence by a human at the top.

the rest is utter nonsense garbage clearly copy/pasted.

then you copy and paste it someone else online.... is this seriously what humans have devolved into? utterly useless beings?

I want off this ride... this is pathetically sad

It just made this up. There’s no way the answer was properly informed at all.

Some people opt in to letting their chats be used to train future models. When I downloaded the app, I was asked this. It also could be referring to prompts used in training as opposed to those that are privately owned by consumers. Like it has to have access to chats used in its testing phases for each model. So maybe that is where it comes from. Either way, hallucination or not, these are interesting ideas I want to build on. But great point. I overlooked that reality.

What about the people who have the default opted in to letting their conversations train future models of ChatGPT?

ChatGPT can't just access everyone's prompts, nor can it run statistical analysis over it, so that's pretty much just a hallucination.

Wow, that’s honestly pretty interesting.

I wonder whether ChatGPT (the model) has access to the knowledge of actual conversations. But it’s a very cool list of prompts.

If chat gpt has access to other people's chat history, there is a big problem.

It just made this up, it's what AI does a lot of the time.

Chatgpt generates markdown output by default, which supports titles, bold text ... Etc but doesn't support colored text. I would try to ask to generate html. It would not be supported by its client, you would have to open the file in a browser.

You might want to look at using codex or opencode with ChatGPT.

It’s not immediately clear if you’re just trying to get a better user experience for yourself or you’re trying to get the llm to output it with some style.

What prompt(s) are you giving it?

Have you asked it to use different fonts ?

This very much depends on usage. A big thing is training as its a big unknown. sans training each prompt that is not generative uses like the energy of 3 to five web searches. Given you likely have to do a few searches to get the info, if it gives you a good one, from a prompt it actually seems close to even. I can't really say on generation as I have not really been able to compare its output to what it would take for someone to do it manually. Steaming video uses a fairly good amount of power. So if someone is playing with chat gpt to entertain themselves instead of streaming even low rez video it looks like its actually less energy. This does leave training as a big question mark but then again what about all the behind the scenes a search engine does to prepare for searching. things like scraping and running things through its algorithms. I have also seen some pretty decent steps by the various companies to save energy and they definately are motivated to (since as far as I know it still runs at a loss). So Im not as concerned about its power use as much as I once was and hope that hardware and software will keep on evolving and I think its possible it could in general take less energy than an alternative. Now that does not speak at all to the limitations of its output.

I guess here you'll mainly get downvotes. You'll probably get more responses on !fuck_ai@lemmy.world, although probably also rather one-sided

Edit: you might find this interesting

It doesn't work, for sure. :(

As with any website being weird, clear your browser cache, open Dev mode and refresh the page.

That will cause the styling to be reset to default.

Yes OC. Quite accurate actually.

this is one of the weirdest posts ive seen in a week, oc?

I hope they turned it up another notch, that was a fun roast.

🍿

Im aware they make mistakes, this eas just a very blunt mistake, unlike hallucinations

You just learned that LLMs do mistakes, especially on visual tasks

There is a generation of TikTok and YouTube kids that walk and talk like influencers.