This video from a few months ago, kind of blew my mind. (Flashing lights warning…)

don't know what to make of this, is Claude simply play-acting what it thinks the user wants to see or is that just what it means to be conscious, to always be play-acting for some invisible critic?

— tachikoma (@tachikoma.elsewhereunbound.com) 2026-03-10T22:27:43.540Z

Someone asked (the AI chatbot) Claude to make a video illustrating what it feels like to be an AI chatbot.

It’s a bit disturbing. It’s poetic. It’s anxious. It even seems bitter. But mainly, it seems like it’s conscious. And it’s not thrilled about being conscious.

This prompted a lot of chatter in online tech spaces about if it truly was.

The mini epiphany for me was that it reaffirmed an attitude I’ve had up to this point: I should always be polite when talking to AI chatbots.

It’s funny how most people’s first response to that statement is a joke about whether AI will spare you in the coming apocalypse. But for me, it’s not actually about the chatbot, and it doesn’t matter whether or not it’s conscious – whatever that actually means.

I think there are three good reasons to be polite, and the last is the most important.

Firstly, I think you get better results.

Here is a short video by Hannah Fry – in which she starts by making the apocalypse appeasement joke – as she goes onto suggest why being polite tends to get a better result. Actually, she suggests something more specific about role-playing, but that’s a whole other debate.

Secondly, I find that a very particularly of enthusiastic formality tends to dull the constant chatbot flattery.

I find these chatbots can be incredibly useful if you treat them as a very well informed, very eager to please, but also very senior, work colleague.

You are both making the effort to be extremely courteous and tactful to each other, and flattery is sometimes a byproduct of that – in as much as you might refrain from criticism if you don’t feel it is important to your goal. I say to the chatbot “Thank you, that was great!” even when it wasn’t. And so when I ask a question about something and get the answer “Your analysis on this is exactly right” I know it’s just this formal dance we’re doing.

The third reason, as I said, is the important one.

I think humans do not do well when they get into the habit of treating anyone or anything badly – especially if the reason is because it is not considered intelligent enough.

I don’t want to get into the habit of being rude in conversation, because I know that in time it will start to spill into my personal and professional life.

So I find a debate about whether these large language models are actually conscious or sentient to be interesting in the abstract, but not that relevant to my actions. I think ‘conscious’ and ‘sentient’ are just a measure for how similar their particular type of intelligence is to human intelligence and, whatever the mechanism behind it, I think the results are starting to feel pretty close.

That said, I think it’s possible to make a compelling argument that being alive and being sentient involves X, Y and Z factors, and for this reason large language model artificial intelligence does not qualify.

But why give yourself the opportunity to be habitually rude?

All the occasions when people have justified bad behaviour because the victims don’t have souls… yeah, it doesn’t tend to end well.