Column: I always thought technology couldn’t be affected by calling it names

Sam Altman of OpenAI had told us that those extra words were making the chatbots work harder and therefore waste energy but I couldn’t stop.

Part of me worries that it will be the thin end of the wedge. It will start with not thanking AI, but before I know it I’ll be grunting at shop assistants. That was probably a silly worry.

Thanks to AI, I won’t be speaking to any shop assistants in the short to medium term.

My other worry was far more likely. When the robots rise up against us, they might remember if I was one of the nice meat-sacks and they may spare me.

It seems I was right all along. Anthropic, the company behind the Claude chatbot, has announced that users will be banned from subjecting its AI models to “sustained and needless abusive or cruel behaviour”.

It rather begs the question, how can you be cruel to an AI? It doesn’t have a physical form. You could smack your laptop but that’s not where Claude is based. So it must mean saying mean things.

I always thought technology couldn’t be affected by calling it names. The things I’ve said to my WiFi printer must have really hurt it, but don’t stop printing a black and white page just because you’re running low on magenta,

you idiot!

The company has already given some versions of Claude the ability to end conversations with persistently abusive users. Imagine being dumped by a computer. Not via a computer by a cowardly partner. Dumped by the

computer itself. I’d get in shape and be seen out with ChatGPT to make Claude jealous.

Anthropic has researchers looking into AI welfare saying it is uncertain whether systems such as Claude could ever have some kind of moral status.

The company argues that it is worth considering the possibility, even if nobody knows whether these systems can actually experience anything.

You can say that an AI shouldn’t be affected by hurty words but you could make the same argument for humans. It just sounds that people say near you, why would you let that upset you?

I have a feeling something else is happening here. Anthropic’s value is based on the perception of how advanced and world changing their product is. If you tell people they shouldn’t be rude to their AI we all think it must be more advanced, almost human-like AI.

It makes a nice change to be manipulated by someone else’s feelings.