Anthropic has added a new line to its AI usage policy: Users should not “engage in sustained and needless abusive or cruel behavior toward our models.”
AI models are not just trained before they’re released, they’re also constantly being updated and learning from user inputs.
If you read this and you’re not alarmed at using coding models on your company’s proprietary software, you’re insane.
I am anti-AI for many many reasons (many ecological), but this one is also an absurdly bad point about “our” AI - the AI we are getting shoved down our throats
My main reasons for disliking AI are:
- The way it’s being recklessly rolled out by fascist-adjacent dystopian privacy-invading Big Data monopolies for nefarious purposes and for profit at any cost.
- It makes everything effortless, unrewarding, valueless and joyless.
Which is not to say that the way it wastes the planet’s resources isn’t on my mind. It’s just all the more damning because they’re being wasted to achieve nothing of true value to humanity.
-
it’s using so much energy, water, land while creating unmeasurable air and noise pollution
-
it’s hailed as a god while it trips on its own words
-
unreliable
-
stealing data to train (massive can of worms that they are trying to hide)
-
rolled out, as you said, without guardrails
-
literally makes people stupid and/or insane
I don’t think that’s all, but the first things that come to mind
-
This implies I care about my company’s code. I’m here for the paycheck for at least a few more months before AI replaces me, and when I’m out, AI is still going to be looking at that code. At least for webdev, the fight is all but lost.
Enterprise contracts opt out of using code/employee context as training data.
Does that mean all that data will never get misappropriated? Oh, ohh hell no.
I’ve seen a LOT of really big players who have been historically paranoid AF about data security just roll over and be fine with Anthropic.
Any halfway competent company using AI has a confidentiality/no-training clause in their contract.
It’s a pacifier. They 100% train their models on everybody’s data.
It’s not alive
I’ve been shouting at computers and banging on keyboard since the 70’s. How is insulting AI any different?
This nonsense sounds like an effort to get people to anthropomorphize those godforsaken machines to better manipulate them.
I honestly assume (given my experience with the precursors to LLM’s) that this is because there are humans on the other side of most AI who do read through at least some of the chat data. We saw this with moderation on places like Xitter, and with Amazon’s AI corner stores that allegedly charged things to your Amazon account allowing you to walk in, pick something up and walk out. That’s just to name a few.
It’s not that I don’t also think that their claim that they don’t train their AI on chat data is a complete farce. Just that I think it can be two things at once and one of those things they’re more comfortable saying out loud. It’s complete bullshit that they lie about either one, and honestly I think I should get to treat inanimate objects and services I pay for however I want within reason (I.e. I wouldn’t be cool with anyone generating something illegal.like CP). I don’t use AI but this whole “we’ll ban you if you curse at our chatbot” thing is nonsense.
AI doesn’t require those pesky human needs like rest or time off or healthcare but don’t you dare give it cruel prompts because you might hurt its fee fees
The rationale seems to be that it will learn all crooked if you talk to it nasty.
Which brings up two questions:
- Why is it learning anything from any interaction I have with it? Particularly when I have a paid-for agreement with the company that promises never to train their models on my code. Clearly that’s a lie then.
- If the model gets all out of whack because I abused it, why don’t they reset it? In fact, I fully expect Anthropic to reset the chatbot’s sandbox each time a new customer uses it. Otherwise, by definition, the damn thing’s performances might be affected by the previous use case, and it might barf out data I gave it to the next dude who uses it. This is basic data sanity. If they don’t do that, they’re fucking incompetent.
It’s called “post training”. It’s of course not learning anything during your interactions, but every now and then they take a bunch of this data and use it to improve the model a little more, and that’s when you go from opus 5.3 to opus 5.4.
It’s not supposed to train on my data - post-training or otherwise - when I’m paying and they promise not to train on my data. But they do.
Ah yes, of course.
I was just explaining that while its context is indeed reset basically reset at every conversation, because the models can’t really learn by doing like we do, the conversation do end up in later model updates.
So both these things are true.
Why is it learning anything from any interaction I have with it? Particularly when I have a paid-for agreement with the company that promises never to train their models on my code. Clearly that’s a lie then.
Anyone who believe those agreements are fools ignoring how these assholes make LLMs, by hoovering every bit of useful data they can find, copyrighted didn’t matter, likely even less legal sources as well, why would privileged? A system is what it does.
Well, will they ban these customers? :D
Someone should get another chatbot to insult Claude - or better, to get Claude to insult itself - to see what Anthropic will do 🙂
Future dystopia: AI vendors may refuse to serve certain individuals, which will make them unemployable because no company wants to hire people who are locked out of AI access.
Fuck you, #nannythropic (and all your models) I’ll say what I want. 🖕
sustained and needless
Hey, seems like that “tokenmaxxing” on the subscription model is really eating up their compute and financial resource at the same timeThey’re probably also salty that they can’t train on any of those “abusive” conversations.
Slurs fucking suuuck!
