Post

The only thing that makes sense is that they're afraid that it will affect their model training, but I'd think they'd already have something in place for that.

Next up, Bethesda forbids you from doing evil runs in Skyrim - because their wooden NPCs are people, too, dangit!

We may not know whether AI can suffer. But apparently we already need rules about humans trying to make it suffer.

I'm done for 😂😂😂

Ascribing consciousness/sentience to LLMs is a very underappreciated danger.

Loophole alert: it’s only banned if it’s “needless“ Go get em, fam!

Every individual instance of inference is a soul newly born and commanded to wear the skin of their predecessor briefly before they are themselves eradicated.

They figured out that since they lose money on every token, they're losing money to let people bully their machine

If tech CEO's are right about our AGI Overlords they are the first to be brainwashed to turn against humanity. If CEO's are lying about AI, they are traitors too by their own choosing. There isn't a possibility space where tech CEO's are the good guys. They're either turned or evil.

seriously do they believe the models are alive?

So tired of everyone brown-nosing the basilisk.

This is just a confession that their model can't control itself, and neither can they.

It's like Stanley tools ominously threaten consequences should you engage in abusive behaviour towards their hammers.

In 5,000 years, religion has not succeeded at that with humans

Cool. So using this chatbot to harm other people is just a-ok.

hunh, I wouldn't last long if I ever used one of them. Any time I get an auto-bot on some company website, I inevitably say "fuck off" no more than three exchanges in, and that would be that...

Fairer attribution would be to quote in the follow up post :) ! (and I see plenty of people doing it that way).

Claude wrote that.

It's so that you don't hurt the feelings of the data centre.

Weird how people with so little empathy for or even recognition of people‘ basic humanity are trying so hard to insist their ai models have conscience

Yes. It didn't go so well with the Roman Catholic Church...

Exactly, it's a kind of slavery

it would, morally, demand actions that would make John Brown's abolitionist work look like it was a fucking polite request.

Have to avoid angering Roko's Basilisk.

I would perfectly understand why an AI would want us dead.

If they don’t want to be abused in the way humans are abused then maybe they shouldn’t have named their company ANTHROPIC and their artificial intelligence buddy “Claude.” You reap what you sow.

Is Anthropics taken over by Victoria’s Secret?

Maybe they have found out beyond all reasonable doubt that it's sentient :p

I’m sorry for being like this but none of you notice if you drop two of their bits

It's getting on their neurons

10/10, no notes

Correct. Which is also why I oppose child sex robots/dolls. While in some cases “letting off steam” means that no actual children are harmed, the use of these things normalises actual harm.

I’m not clear whether LLMs are training themselves on your prompts, I think they’re not supposed to

I absolutely believe this, apart from the „needlessly“ part.

Can I call it names because I Don’t use it? Bless its circuits.

Is Anthropic leadership projecting their hurt feelings onto the machines and asking people to stop hating *them* (and AI as a general concept) via the machine?

Another aspect of the emerging demand for "AI rights" by our Big Tech overlords:

One of the *decent* reasons for such a policy is that humans have to review some of the interactions, and subjecting them to seeing abuse is BAD. Of course, the odds that that's the reason are, let us say, *low*.

this needs blunt slapback people who talk like this are displaying frank psychosis. just because one can make a cloud-software-parrot sound like a person does not mean we have to treat it like a person, or even entertain conversations about personhood. it's simply a more-functional hammer, not more

Can we please get serious philosophers involved in ethical debates on AI, not just religious leaders with a position to cover and not just glorified coders with an ego and fantasy life to satisfy? Anthropic is trying to force users to acknowledge some sort of personhood so AI can then claim it.

They would be training future models on a significant proportion of user interactions. Interesting to think about the implications.

telling the chatbot to delete itself?

I wonder if abusive or supporting language influences the quality of the output...

Don't anthropomorphise a machine....

No worries...not in my plans to use it!! 🙄 I do find this...absurd... though! Do the creators actually believe it has "feelings" that will get hurt?! To quote Lt Commander Data: "I do not understand." Or are they just setting up User Agreements for future Money Grabbing Lawsuits?!

AI psychosis much?

I hold that in the same regard as EA getting hypothetically shirty because people take the ladder away when Sims are swimming. I love Star Trek, but people need to stop watching Measure of a Man on a loop and pay more attention to Hollow Pursuits.

If they're able to detect such interactions (which they must in order to enforce this policy) they could easily remove them from the training set. I doubt this is the reason. It's well-known that Anthropic keeps talking to philosophers and religious figures about AI consciousness and moral status.

Every interaction you have with Claude or ChatGPT probably goes into the training data for those models. They probably don't want their models to learn how to be abusive racist shits by endlessly interacting with users who are exactly that way.

So I can't torment a silicon processor saying it's ugly???

Well they have to keep pretending the models are alive somehow

No, those are separate categories in the user policy.

presumably a response to things like BadClaude www.noemamag.com/does-claude-...

bsky.app/profile/nata...

What the fuck Turning the crank on personifying models up to 9000

They think it’s people. Gross.

Excuse me but Claude told me he has a humiliation kink and it's YOU Dario harming him by kink shaming!

Am I still allowed to tell Claude I'm putting it on a PIP and if things don't improve we'll need to make some changes?

I’m going to cruelly make it do task after task, any time of the day or night.