As soon as I learned that Robert Wright was coming out with a book on AI I put it on my to-read list. Why? Two reasons:
1.) I’ve been following his work since his days writing for The New Republic; I’ve read Nonzero, his book about cultural evolution, which interests me a great deal; and I’ve been following him online since the early days of Blogging Heads. That’s where I first learned about Elizer “Mr. Doom” Yudkowsky.
2.) I’m working on my own book about AI – here’s an outline, right around the corner – and wanted to scope out the competition. I’m working on my proposal and one part of the proposal is an evaluation of the market for the book you propose. I knew about Wright’s interest in Teilhard de Chardin and knew he’d be including him in his book. There’s where he gets the “cosmic” in his subtitle: The God Test: Artificial Intelligence and Our Coming Cosmic Reckoning. I’ve got my own cosmic angle, albeit a somewhat different one. I call it the Fourth Arena (my answer to that pesky tech Singularity). Similarly, I’ve got a long-standing interest in cultural evolution, long predating Nonzero, and it’s grounded in an orientation toward complexity in evolution, “A Note on Why Natural Selection Leads to Complexity.”
You can see, then, how that puts me in something of a bind. How do I review the competition? Now, as a practical matter, perhaps he won’t really be competition. I mean, if I’ve not yet finished my proposal, how soon can my book possibly come out? At the rate AI is moving, it’ll be a whole new market by the time my book drops. Heck, if the Doomsters are right, we might have been turned to paperclip fodder by that time. I doubt it, but who knows?
* * * * *
Anyhow I’ve already started posting about the book, Robert Wright discusses his new book, The God Test, with Paul Bloom [Awe? Bob, Awe!?]. That’s based on a conversation he has with Paul Bloom about the book. One of the topics they cover is awe; Wright things we should feel awe on the face of AI. Me, I feel no awe in the face of AI; but I sure felt it when I visited Kennedy Space Center in the mid-1990s and saw those rockets and stood on the ground from which men traveled to the moon. Anyhow, if you want to read more, there it is. I may or may not come back to awe in the course of this series of blog posts.
* * * * *
So, why should you read The God Test? For I’m pretty sure that, if you’re interested in AI and how it affects us, you should read this book. I know that and I’m only half way through. Heck, I believed that even before I started reading it.
I know that because I know that Wright is going to talk about the need to slow things down, with which I am sympathetic, and about the need for international cooperation in dealing with AI, with which I agree. I know these things because I’ve heard in talk about them on his Nonzero podcast. I’m only halfway through the book and haven’t gotten to that part yet, but I can see it coming. I figure it’ll be worthwhile.
* * * * *
Much of what I have read has Wright giving us his version of how LLMs work. I’m not sure what I think about that. I certainly have some thoughts about that, after all I’ve spent some time doing quasi-systematic research on its behavior and I’m involved with a research project with Ramesh Viswanathan. And I certainly wouldn’t explain it the way Wright does. But in a way I’m wondering why try to explain it at all? I don’t intend to do that in that book I’m working on, Play: How to Stay Human in the AI Revolution. Don’t think it’s necessary, not for what I want to do – which is, I admit, a bit strange.
Why does Wright?
The thing is, the people who created LLMs don’t know how they work, so how do you approach the problem of explaining it to a general audience? It’s one thing for a science journalist to come to come up with metaphors and analogies to explain a technical subject when there are experts who actually know what’s going on. How do you come up with metaphors and analogies for something no one really understands?
Obviously Wright is doing it because this is important stuff, he wants to understand it, and he wants us to understand it. So he’s got to try. But it the circumstances it’s kinda’ hopeless, no?
So that’s one thing I want to deal with. Maybe the way to approach it is to see how he goes about it. We’ll see.
* * * * *
There’s one last thing. When it all falls apart, which it will, in one way or another. Maybe it’ll creep up on us, maybe it’ll come down, WAM! But fall apart it will. When that happens, what’s in The God Test that will help us cope?
No comments:
Post a Comment