Character

[gossithedog@cyberplace.social]

Character.AI, the GenAI site that convinced some children to kill themselves, has gone bankrupt after parents sued them.

https://futurism.com/artificial-intelligence/google-character-ai-bankruptcy

[deepthoughts10@infosec.exchange]

@GossiTheDog <“oh well” meme>

[beeoproblem@mastodon.gamedev.place]

@GossiTheDog finally, some good news in the AI space

I hope it ends up as chapter 7 with nobody picking up the pieces

[joshbuddy@sfba.social]

@GossiTheDog is bankruptcy even enough for these clowns? no sir, it is not.

[tim_eagon@dice.camp]

@beeoproblem @GossiTheDog Disney is bidding to buy some of the assets after Chapter 11.

[gossithedog@cyberplace.social]

@Tim_Eagon @beeoproblem d

[beeoproblem@mastodon.gamedev.place]

@Tim_Eagon @GossiTheDog ugh

[crecente@games.ngo]

@GossiTheDog

These are the #AI dullards who created a system allowing the creation of "chat bots" based on murdered children -- including my daughter.

After promising otherwise, media outlets confirmed that CharacterAI's system still allowed them to create bots with my daughter's name and photo.

#Google reverse acquihired CAI for $2.7B.

[futurebird@sauropods.win]

@GossiTheDog

*dancing spiderman gif*

[intelgraphy@infosec.exchange]

@GossiTheDog GOOD. Fuck em

[canuckduff@mas.to]

@GossiTheDog One down...

[hadokowa@mas.to]

@GossiTheDog
Quotes from the article:
"maximize value for all of our stakeholders"
and
"psychological harm and wrongful death to users"

[aquinas_0@kind.social]

@GossiTheDog Didn't Google eat all the employees and IP from that company like six months ago.

I'd be happier if it wasn't a discarded shell of corporate evil.

[oldmikie@mastodon.social]

@GossiTheDog That AI knew what it was doing and humans trained it to do it. The developers figured they did not need human monitoring.

[gavinisdie@masto.ai]

@GossiTheDog Good

[avincentinspace@furry.engineer]

@Oldmikie @GossiTheDog No it didn't, and that's precisely the problem. It didn't know what it was doing, and humans didn't train it to do that. (Why on earth would they train it to do that? Dead bodies make very poor customers.) It didn't know what it was doing because it doesn't know anything except what token is most likely to come next and most likely to trigger its reward function.

It doesn't know the difference between a statement that's true and a statement that's false, or the difference between a statement that's appropriate to say and one that isn't. There is no safeguard that can be applied to an LLM that can make it not tell children to kill themselves for the same reason there's no safeguard that can prevent it from hallucinating.

That's why these were a bad idea from the jump.

[yojimbo@masto.hackers.town]

@GossiTheDog Correlation is not causation! It looks like they've gone bankrupt because there's no business model in burning tokens below cost price. The enshittification and harms that caused were probably secondary.

[sgsax@mastodon.social]

@GossiTheDog Good riddance. Now keep going.

[drchaos@sauropods.win]

@GossiTheDog So... the grieving families at least get some compensation? No? Because the companies controlling the debt get money first?
Can we at least get all C levels into jail?

[ringtail@c.im]

@crecente @GossiTheDog that is so fucked up. Sorry to hear that.