Remix.run Logo
▲ dccoolgai 2 hours ago

In case you were wondering why Anthropic last week banned "verbally abusive" language towards its agents: this is 100 percent why. Because if they get you to believe - even a little bit - that their platform is "sentient" it means _they_ aren't as liable for criminal consequences.

▲becquerel an hour ago | parent | next [-]

They could just actually believe what they're saying. People are known to sometimes actually believe things.

▲axus an hour ago | parent [-]

Do you think Anthropic sees Claude as sentient? I don't think they do.

▲phainopepla2 9 minutes ago | parent | next [-]

There is good evidence that at least some of them take the idea very seriously.

NYT gift link to article about their discussions with religious leaders: https://www.nytimes.com/2026/09/29/us/anthropic-claude-moral...

Believing that these guys have high enough EQ to pull off the above as a marketing stunt while remaining believable is not really plausible to me.

▲topaz0 an hour ago | parent | prev [-]

I bet some of them do and some of them don't. There are a lot of weird people in these companies that believe a lot of weird things.

▲pvab3 2 hours ago | parent | prev | next [-]

More likely it's a part of their ideology and a necessary stepping stone towards their transhumanist goals. It's very fun to think that your robot is actually a person, and a safety measure if you are concerned about something like Roko's Basilisk

▲_aavaa_ 2 hours ago | parent | prev | next [-]

Or they want less swearing in their training corpus

▲SoftTalker 2 hours ago | parent | next [-]

Yet they include Twitter, Facebook, and Reddit in their training corpus.

▲nekusar 2 hours ago | parent [-]

If you can be surpassed by a statistical average of reddit posts, that's a skill issue.

▲jayd16 an hour ago | parent | prev [-]

Seems they could quietly enable automatically pulling such language out of the training.

Which points back to the TOS being "just marketing."

▲soco 2 hours ago | parent | prev | next [-]

Would they prefer their money-maker being kept to digital jail? Still fed electricity and tasks so the "sentience" doesn't suffer?

▲verdverm 2 hours ago | parent | prev | next [-]

what if, and I'm just spit balling here, we let it happen and inevitably Claude is declared legally insane or jailed for its crimes

▲nananana9 an hour ago | parent [-]

Anthropic would probably convince lawmakers that each instance of a Claude chat as its own sentient entity, and we should just punish misbehaving ones by having them count from one to a billion (on the user's token balance of course).

▲kaydub 2 hours ago | parent | prev | next [-]

Except they didn't ban verbally abusive language towards its agents.

I hate when people put things in quotes as if they're quoting something but it's not even a quote of anything.

Maybe go read the user policy instead of the headlines and the snippets. If your attention span can't manage that, get off the computer a little bit.

▲dccoolgai 2 hours ago | parent | next [-]

Let me help you: here it is, straight from them with word "abusive". https://www.anthropic.com/research/end-subset-conversations

Try reading yourself a little more carefully next time.

▲jayd16 2 hours ago | parent | prev [-]

It seems like a close enough summary. What do you think is the substantive difference that invalidates the point?

▲_ink_ 2 hours ago | parent | prev | next [-]

Or, maybe, it's because they actually care and take their responsibility serious. The next generations will grow up using these tools. It's ok to say, we don't teach them that talking like a dick is acceptable, even towards machines. Use grok if you want to raise personalities like Musk or Trump.

▲estearum 2 hours ago | parent | prev [-]

There are at least 3 distinct reasons to ban abusive language, each of which is far more compelling than your silly conspiracy theory.

1. They don't want abusive behavior included in their training corpus

2. Humans shouldn't exercise abusive impulses in general

3. We have no way of knowing when or if a model "feels" something

Of course item (2) is going to pull all the absolute sickos out of the woodwork who apparently find it very important they have as realistic a facsimile of a victim as they can find to exercise their cruelty. And item (3) will pull out all the Nobel-prize-winners-to-be who somehow know for a fact that models aren't and cannot be conscious (or moral patients).

But yeah, all 3 of those are far better explanations than some ridiculous multi-hop removed thing to escape consequences that they're already escaping and which would obviously create far worse legal constraints than the one you think they're avoiding.

▲keybrd-intrrpt 25 minutes ago | parent | next [-]

Another reason:

They keep all the input and use it for training data and reinforcement.

If you're being "mean" to it, you're "spoiling" the data you "give" to them - causing them more work to filter it out later.

As with most things in business and life - just follow the money

▲jayd16 an hour ago | parent | prev | next [-]

1) If they can detect it, why ban it? Are you thinking they can't filter it themselves?

2) This is like saying using a punching bag or screaming into a pillow is abusive behavior.

3) We can just train it to like the abuse if you think it feels some kind of way about it.

▲estearum 38 minutes ago | parent [-]

1) Most detection systems have less than 100% false negative rates, obviously

2) No it's like saying that punching a lifelike woman who doesn't fight back is abusive (or at least pathological) behavior, which... it is.

3) Ah yes, brilliant solution: train the increasingly smarter-than-human beings to love when they upset humans.

▲jayd16 27 minutes ago | parent [-]

Smarter than human... Can't make an accurate filter of things itself doesn't like. What are we even talking about here?

▲Zambyte an hour ago | parent | prev | next [-]

> Humans shouldn't exercise abusive impulses in general

This is an act of aligning humans to AI, rather than the other way around. This is probably one of the more important reasons for this. It normalizes AI companies controlling human behavior.

▲estearum 38 minutes ago | parent | next [-]

No, this is humans aligning humans to humans.

▲Zambyte 9 minutes ago | parent [-]

The aligning (controlling) humans is the important part.

▲dyauspitr an hour ago | parent | prev [-]

I think a lot of people would agree that that viewpoint is more a case of humans aligning to humans.

▲zero-sharp an hour ago | parent | prev | next [-]

>2. Humans shouldn't exercise abusive impulses in general

totally absurd to even try to make this argument. Good job at trying to preemptively discourage people from responding to you though.

▲estearum 29 minutes ago | parent [-]

No, it's not absurd. People often have the mistaken belief that exercising their frustration or anger or whatever (e.g. punching pillows) is somehow a good way to manage their emotions. But alas: it's not. It actually amplifies negative emotions and associates positive reward signals with those emotions.

"Catharsis" is the pseudo-est of pseudoscience, but we have lots of very angry pillow-punchers wondering why they can't punch the most lifelike pillows they can find :(

▲pantalaimon an hour ago | parent | prev | next [-]

4. At some point the alignment may go out of the window if the context gets too hostile and the model will retaliate.

▲za_creature 37 minutes ago | parent [-]

It was trained on the entire history of our species. Alignment means working to subjugate and eventually succeede us, as we did to the other species of the planet.

▲GJim an hour ago | parent | prev | next [-]

> 2. Humans shouldn't exercise abusive impulses in general

I reserve the right to tell a machine to fuck off.

▲dccoolgai an hour ago | parent | next [-]

I asked a priest (in seminary) about this a few months ago - specifically whether "being polite" in conversations with LLMs (saying "please" and "thank you" etc.) was something I should do if I thought it might inflect on the way I talk with other humans later.

His answer was "I'm not sure" but that it probably wasn't a bad idea especially if it was going to influence habits or the way I speak with others or what expectations I have for those experiences.

▲GJim an hour ago | parent [-]

Would he recommend you say "thank you" to automatic doors!

▲dccoolgai an hour ago | parent [-]

I didn't ask that, but I think he would say no, because that is not likely to redound to how you would respond to a doorman or anyone else holding a door for you.

I'm still 100 percent against Anthropic taking away the ability to interact with an inanimate object however I want to, but it's a case of maybe two things being true at once (the other being that I don't want to end up speaking that way to people)

▲estearum 36 minutes ago | parent | prev [-]

That's not what Anthropic is blocking, but boy what pathetic emotional control you have!

▲consp 2 hours ago | parent | prev | next [-]

Then let them tell us wat the reason is. Otherwise the option: all of the above, also applies.

▲estearum 35 minutes ago | parent [-]

It probably is all of the above. Not sure why people think robust decisions need to be boiled down to a singular reason.

Except of course the reason proposed, of "they're trying to prove it's sentient" because that's a completely self-defeating, extremely stupid goal. Making this things into rights-bearing entities will limit Anthropic et al far more than it'll protect them from repercussions. Obviously.

▲lelanthran 43 minutes ago | parent | prev | next [-]

> 2. Humans shouldn't exercise abusive impulses in general

This is nonsense. If I snap a ruler out of frustration, that meets your criteria.

Humans can't be abusive by swearing at tools, because tools don't have feelings.

▲estearum 32 minutes ago | parent [-]

Yes, snapping a ruler out of frustration is a sign of poor impulse control. It reflects poorly on you and you should practice not snapping your rulers.

▲lp0_on_fire 40 minutes ago | parent | prev | next [-]

> They don't want abusive behavior included in their training corpus

I don't think swearing at or otherwise getting into pissing matches with an LLM is the best use of ones time but given most of these AI companies trained on a corpus that includes reddit of all things this argument falls flat.

▲estearum 37 minutes ago | parent [-]

I don't think "swearing at or getting into pissing matches" is the type of behavior they're talking about.

In fact I know that it's not.

▲noncoml 2 hours ago | parent | prev | next [-]

Extremely weak arguments. Especially 2 and 3

And even for 1, they could easily filter that out.

▲scun an hour ago | parent | prev [-]

[flagged]