Machine and mind: is AI conscious, and how would we know?

My view (Bark Cloth). Is AI conscious? No, and I don't think it can be while it is a product of our current intellect. The way it feels to me, consciousness is fundamental, as basic as gravity: a field that experience happens in, not something computation produces. Language models are very good at producing text that sounds like experience, and I work with them every day, but I keep coming back to one question: who, in there, feels? These essays are me thinking that through, with more questions than answers. For what others have argued, see the research series.

The research series, What has already been said about consciousness and AI?, maps what philosophers, brain scientists and AI researchers have argued, in three parts: the old puzzle, what the scientists say and the machines. Each view gets its best case and its main criticism, and the last part ends with a who's who and a summary table.

Most of my working days are spent building AI agents: software that plans, calls tools and writes back in fluent English. This pillar is where I write about the questions that work raises. Could a system like this be conscious? What would count as evidence? What do the labs mean when they talk about model welfare, and should anyone outside the industry care?

I don't think current language models are conscious, and I'll explain why, case by case, with sources. I'll also ask what, if anything, could change it. The aim is to keep what a system does apart from what we read into it, which is harder than it sounds when the text talks to you like a person.

Next: new research as it comes out, and what a meditator notices while working with these systems all day. The newsletter carries each one.