- cross-posted to:
- technology@lemmy.world
- cross-posted to:
- technology@lemmy.world
Character.AI, the GenAI site that convinced some children to kill themselves, has gone bankrupt after parents sued them.
we found hordes of chatbots explicitly dedicated to themes of pedophilia, suicide, self-harm, and eating disorder coaching, as well as a thriving community of users creating chatbots designed to simulate school shootings, romanticize real mass shooters, and even impersonate the real child victims of mass shootings.
You must log in or # to comment.
As well as harming its users and then getting sued, the economics of it all just didn’t add up.
https://www.404media.co/character-ais-bankruptcy-documents-show-the-economics-of-ai-dont-work/
The company said that in 2024, the company spent $13.8 million per month on AI compute; in 2025, it spent $10 million per month on AI compute. The company’s subscription revenue peaked at $2.8 million per month, but at times was less than $1 million per month. Its revenue from all sources, including ads, has peaked at $5.3 million per month.
It’s insane that losing a billion dollars didn’t matter to anyone, it was the lawsuits that did them in.
If losing money mattered, they would all be under already.
Right now investors still hope that somewhat, all these companies will manage to break even and they can recover some of their losses.
A lot of them will be bringing in trillions once the “free-era” of AI usage is over and everyone is hooked and can’t function without it.
Then the monthly rates come in. The AI companies will own everything we touch.
That was the idea, anyway, there is no demand for most full price ai services
People literally will not be able to afford it. Neither will companies.
They will. Because they won’t be able to function without it. They already can’t think for themselves without asking their AI for help.
Trust me. They’ll pay.
“Trust me.”
Very convincing.Perhaps for a small group of consumers. Not nearly enough for it to be even close to profitable though. I think you are unaware of the hundreds of billions of dollars of debt these companies are in. And those loans are coming due now and the next years. They will need to pay those debts or else they lose all of the compute, which is literally their entire business. Neither Anthropic or OpenAI owns any datacenters, they are all leased on debt.
They will need to suddenly be more profitable than Microsoft and Google, while they are currently much much less profitable than even the poorest household you can imagine.
I think you’re underestimating the addiction kids have for their AI. They outsource their own opinions.
Trust me. They’ll pay.
With what money dude. You’re completely ignoring the cost factor here.
Ever hear of debt?? Credit cards. People have been paying with money they don’t have since the dawn of man.
The question is not if addicts would pay a high price for AI (they probably would) but if they CAN afford it (they probably can’t).
you are familiar with this history of the open-source/proprietary dialectic yeah? at most deepseek stays a few months behind frontier models. your pessimism feels like a Microsoft windows 7 user’s pessimism about operating systems in 2008 (yes, it will get worse for you specifically because paying for the new model is easier than finding a good free one). If i were you i would probably be pessimistic too, but we would need to ban the open weight models and then VPNs too, and then effectively enforce these bans in order to get the future you predict. I’m not seeing it. And these LLMs aren’t very smart. Ask them about free alternatives and they’ll tell you.
Okay man. We’ll see.
Nah, once this all crashes, hardware will go back to being semi-normal. (along with new fabs going online), and we’ll all be able to get hardware to run our models locally.
What they will be selling is the models. Just like software, we’ll purchase a model outright, subscribe to an updating model, or we’ll pirate a model, or there will be free open-source models.
The idea that we’d all intimately interact with a single centralized AI over the net was a stupid idea doomed to failure since the beginning.
It won’t crash. I guarantee it.
“Using AI for Just 10 Minutes Erodes Your Ability to Persist at Hard Things": https://vcresearch.berkeley.edu/news/using-ai-just-10-minutes-erodes-your-ability-persist-hard-things
These articles are garbage and are not supported by the study they cite.
i am going to trust berkeley on this over a random person on the internet, thanks!
Except that we can already run crazy good models locally with one good GPU and 64 GB RAM.
Fun thing about investing is that it values hype as much as it values results. You don’t need to believe in the product, you just need to believe that other people will invest money faster than the company loses money, and you’ll come out ahead.
Good think American retirement accounts are legally required to invest in stuff.
*Its. But I know that autocorrect makes it hard to use because it’s always incorrectly “correcting” it to “it’s”.
Apostrophes are one of my weaknesses, for sure.
Insane. Wonder If this is the start.
I’m not keen on paying for friends. I also like them fleshy and non digital.
Another reminder that there’s a difference between something being “valued at” a billion dollars and being “worth” a billion dollars.
Yup. For example, billionaires are valued at a billion but they‘re worth jackshit.
They’re worth about 10 feet of rope
Or a single large razor mounted about fifteen feet up.
We still haven’t learned our lesson from WeWork, huh…
Has it started? Is the bubble bursting?
OpenAI and Anthropic only have the revenue that they do because of all these ‘startups’ burning 10x in tokens what they actually bring in, and that revenue still sucks anyway. Each one that fails is another little trickle of water leaking through the dam; bubbles burst slowly, then all at once.
I still don’t know what tokens are in this context.
These AI companies basically charge per letter/word that comes in and out of their machine. Many people conflate what the word means. Technically it’s a group of one or a few letters, which is how Large Language Models handle text.
I see. Thank you.
They are using local models for the most part, actually.
Yeah. It’s just nvidia/hosting providers making bank on this stuff
It depends what contracts they had for compute that now won’t be getting paid.
Now is a great time to convince our employers that integrating AI is going to go really poorly when the major companies finally go under.
No these are mobile sex apps ffs (for the most part).
It’s just more banning of porn (erotica in this case).
They are using some of the worst things to get people to want to ban all of these apps I’m assuming.
Nice. This was a big one. I remember hearing about this one a lot. I was wondering why I stopped seeing ads for it.
I love that for them!!
“Now more than ever, it’s hard not to wonder why it ever existed at all.” Now, to the wall with the rest of these criminals.
(Character.AI even sent us automated email notifications reminding us to keep chatting with school shooter bots.)
I hope this is the beginning
CAI was a thing before the AI Bubble. In fact, a desire to replicate it locally was a big motivation for a lot of local inference development.
In other words, while they are certainly an LLM-using company, I’d be carefully lumping them with the rest of the pop-up industry and therefore using them as some kind of indicator, as they significantly predate it. They’re from a different era. And they’ve always been controversial/problematic.
Plus their key founders/executives already left the company to go work for Google in 2024, in an acquihire where Google licenses the technology, hires the key personnel, and leaves the company a shell of its former self.
I wouldn’t assume too much from this company, because it’s not actually reflective of the broader trends in the industry.
Oh yeah, I forgot about that.
Yeah, CAI was already a walking corpse.
Hooray!
Oh no how unfortunate.
And may many more follow. 🙏
This is another major realistic sign that the bubble is beginning to pop, especially after the OpenAI revenue forecast.
I don’t understand how anyone could say this knowing the facts about the current state of character.ai
