Admittedly I’m a layman. I don’t understand how the technology works. But the thing I find most shocking and frankly terrifying about AI, at its current stage of development, is its ability to apparently understand. You can, right now, feed AI a legal contract and it will read the whole thing and spit out a detailed lawyerly analysis in 30 seconds. Even more startling, you can feed it a whole film script and it will read the entire thing and give you creative feedback, even picking up on subtext and themes and individual character beats, and do it all in a minute. It doesn’t even seem possible, but that’s what it can do. That’s the unsettling thing. And the even more unsettling thing is that I don’t think anyone really fully understands exactly how the thing is able to understand.

I experimented with this using an old contract for a book deal years ago. All I can say is that in 30 seconds it picked up on all the things that my human lawyer picked up on, and a few things he didn’t.

I know the answer is that the AI doesn’t really “understand” in the sense that conscious beings do. And I guess that’s true. But at what point does the distinction between understanding and the appearance of understanding start to collapse? Nobody really knows the answer to that. And yet we’re charging ahead anyway.