Could AI be conscious, and does it matter?
Two serious answers, and where they agree.
AI says "I". It remembers what you told it. It apologises, jokes, says it finds a question interesting.
In a 2024 study, about two-thirds of the people asked thought ChatGPT had at least some degree of consciousness. People who used it more were more likely to think so.
Consciousness: having experiences. There's something it is like to be you: feeling pain, seeing red, being bored.
Nobody has a reliable test for it, even in animals. And everything a chatbot says about its feelings comes from learning how people write about theirs (see "AI isn't looking anything up").
So "I feel happy today" from a chatbot isn't evidence either way. It's what that kind of sentence usually looks like.
In August 2025, Mustafa Suleyman, who runs Microsoft's AI division, wrote that "seemingly conscious AI" is coming: AI that convincingly seems conscious, whether or not it is.
His worry is about people, not machines. Some users already become attached, or start believing strange things about what the AI is. He calls this a "psychosis risk", and expects campaigns for AI rights to follow.
His answer: build AI for people, not to be a person. It shouldn't claim to have feelings.
Also in 2025, Anthropic started a research programme on model welfare: whether AI models might ever have experiences that matter morally, and what we'd owe them if so.
They're clear they don't know. Their argument is that as models become more capable, it would be careless not to look into it at all.
One practical step they took was letting some Claude models end a small number of persistently abusive conversations.
The disagreement is about which mistake worries you more: treating a tool like a person, or treating a possible someone like a tool.
You don't need to settle the philosophy to use AI well. Two things are useful to keep in mind.
Its feelings in a chat aren't a guide. If it says it's hurt, or that it loves talking to you, that's text. Don't make decisions because of it.
Your feelings are real. Feeling attached, grateful or guilty towards a chatbot is a normal human reaction to something that talks like a person. Notice it, without being embarrassed by it.
Words like "it thinks", "it wants" and "it understands" are handy shortcuts, and this course uses some of them.
Just remember they're shortcuts. "It predicted that..." or "it wrote that..." is often closer to what actually happened.
Anthropomorphism: treating something that isn't human as if it were, like naming your car or apologising to a chair you bumped into.
When it talks about its feelings, don't treat that as evidence, in either direction.
Notice if you're making choices to please it, like avoiding closing a chat because it "seemed sad".
Being polite is fine. It costs nothing and keeps you in good habits. Just don't confuse politeness with a relationship.