Raw LLM Responses

10:15 is where I broke down laughing. It's very confusing, because he has some valid concerns about the use and control of AI - and he sounds smart, when he describes these problems. But he clearly has no real conception of the software he's been testing. Current AI is a function approximator. The output is literally white noise being fitted to an extremely complex set of patterns that resemble "text with by a person". When you're having a "dialog", it's actually using the text of the entire "conversation" as input every time, because it's just a single function - text in, text out. The AI doesn't recall what you just said - it has no memory. You definitely can't hurt its feelings or change it's mood, because the only place those things exist is in your interpretation of the text, which isn't even stored or learned. If you dump the AI's neural network to a disk, it's identical, before and after your input. Part of me thinks he does understand that, or he's not a very good engineer. I expect he's deliberately being facetious about it, in order to raise awareness of other ethical issues. But then he uses words like "sentient" and "spiritual", and it's really hard to know if he's being serious or not. If you think a large enough spreadsheet will become "sentient", then sure, current AI might be "sentient" - but then we have VERY different definitions of what that word might mean. I think there is more consciousness in a potato than there is in current AI.