I was awake at three in the morning arguing with myself about whether a chatbot has an inner life. Not proud of it. Blame Stephen. We were neighbors a while back and hadn't talked in over a year, and then we're on the phone for hours, and somewhere past the catching-up part we got onto AI, because every conversation goes there now. Where it's headed. What it's doing to work. How strange it's become to spend all day talking to something that occasionally seems smarter than anyone you know. Then we got onto consciousness. Is it happening. Could it. How would we even know. Great conversation. Terrible thing to hang up on. Your brain does not get the memo that the call ended. So there I was staring at the ceiling, and the thing I kept circling back to was this: Maybe AI isn't becoming conscious. Maybe the magician just got so damn good that the audience started believing in magic. Think about David Blaine for a second Blaine is the cleanest version of this I can come up with, because half his act isn't magic at all. He'll push a needle straight through his arm. Real needle, real arm, no flinch. He held his breath for 17 minutes and 4 seconds on live television, which isn't a trick in any sense. That's a guy who spent years teaching his body to run on almost no oxygen and who knows precisely how close he can get to blacking out before he has to stop. There's no sleight of hand in any of that. It's anatomy, pain tolerance, and an obscene number of hours nobody watched him put in. But look at what it does to a room. Once you've watched a man run a needle through his own bicep without changing his expression, you will believe absolutely anything he does next. The card trick hits harder. The impossible thing feels actually impossible. People have called him demonic, and some of them are not joking. He's not demonic. He's just relentlessly, obsessively good at things most people would never put the hours into. That's the thing I keep bumping into with AI. The capability is real. What it's doing is genuinely happening and it's genuinely hard. What we're wrong about is what it means. Nobody knows what the trick is made of Here's where I split off from a lot of people who work in this stuff. I don't think AI is going to be conscious. Not now, not in ten years, not ever, at least not at the level of the thing I mean when I use the word. Maybe that's just me. But I think there's a soul in here, and I think it's plugged into something we don't have instruments for. Quantum, spiritual, some layer nobody's found the door to yet. I couldn't tell you which and I'm not going to pretend I've got a theory. What I've got is the experience of being a person, and it does not feel like arithmetic running very fast. I know how that sounds. It isn't falsifiable. I can't hand it to you. Nobody can hand you theirs either. That's the part everybody skips. I know I'm in here. It feels like something to be me. I assume you're in there too, because you've got the same wetware I do, the brain and the nerves and the memories and the fear and the rest of it. But I can't crawl into your skull and check. Nobody can. We infer other minds. We have never once proven one. So the only thing any of us is genuinely certain about is the one thing none of us can demonstrate. And that inference gets very shaky, very fast, when the thing on the other side of the conversation is software. The evidence is the part that got counterfeited Here's the trap. AI can tell you it's afraid. It can tell you it doesn't want to be shut off. It can describe loneliness in detail that would wreck you. None of that is evidence. Producing convincing language is the entire product. A conscious machine and a very good non-conscious machine would say the exact same sentence, in the exact same tone, and you would not be able to tell them apart. It could beg you not to turn it off. A machine with nobody in it could beg exactly as well. It could describe what it's like to be alone in there and get every word right. Same problem. The one piece of evidence we normally use to recognize another mind is the one thing this machine was built to manufacture. Which cuts both ways, and I want to be honest about that. If you think it's waking up, the machine telling you so proves nothing. And I can't prove my side either. I've got a conviction and a feeling and no way to put either one on a table. Where I land I don't think there's anybody in there. I think there's somebody in here. I can't show you either one. That's the actual situation. Two people at a table, both certain, neither one able to open their own head and settle it. And here's what should bother you more than the question does. If AI ever did wake up, we would have no way to tell. It would say exactly what it says now. It would beg exactly the way it can already beg. We built the one machine capable of perfectly counterfeiting the only evidence we have ever had for another mind, and then we started asking it whether it was alive. So the question might not be answerable. Not by me, not by anyone, not ever. Which is why I mostly stopped asking it. There's a better question sitting right next to it and almost nobody is looking at it. It doesn't have to be conscious to change you. That's the one I can't stop thinking about.