The Most Human Human
Apr. 17th, 2012 12:36 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
ivyficTitle: The Most Human Human
Author: Brian Christian
Genre: Non-fiction (Science and Philosophy)
Summary: Every year, the Loebner Prize tests Turing machines to see if any computer programs have yet reached the point of indistinguishability from real humans. To compare, of course, the judges evaluate a computer head-to-head with a human. The computer that gets the most votes is the Most Human Computer, and winner of the Loebner Prize. But there's a prize for the human interlocutors, too: The Most Human Human. Brian Christian signs up for the contest and launches on an interrogation of what the title "Most Human Human" even means.
Review: This is a fine book. Which is a huge disappointment, because it could have been excellent. It has one of the best premises--and best titles--of any book to come out recently. It got a lot of press, because the interest in the topic is immediate and obvious.
With all that, I wanted a story of the Loebner Prize and the author's quest for the Most Human Human award, along with some computer science and philosophy. I didn't get a story of the Loebner Prize--at all. He talks about leading up to it, makes a few references to the actual event randomly throughout, and then skips to the award ceremony, which was profoundly disappointing.
As for the computer science and philosophy, well... There's a lot of it.
Christian talks about human speech patterns, which, in spoken conversation, are overlapping, interrupting, digressing--everything but linear. Maybe it was his intent to write the book that way, but if it was, it's an impressive failure. There are some chapters that have a coherent direction. But most of them randomly wander off and never get to the point. You might find yourself suddenly reading about Aristophanes in Plato's Symposium and never quite figure out how that has anything at all to do with the Turing test. Sometimes he'll give a bit of a precis at the start of the chapter, but then he won't follow it.
And there are all these headers throughout. That contain no structural value at all. It seems like he wrote the thing and then just stuck a header in whenever a pun occured to him. Or whenever he realized that this page-long bit has nothing to do with anything in the text surrounding it. And they're all A-heads, so they don't help create a sense of hierarchy. And there are random epigraphs, sometimes in the middle of a section. I get it, dude, you studied philosophy. You know shit. But you need to work all of this material together into a coherent whole, instead of leaving all the lumps in the gravy.
So as you read, you find yourself in a forty-page long digression about data entropy and compression algorithms. Or not even a digression, really, because "digression" implies some sort of starting off point, and the only relation this seems to have to the purported topic of the book is that he read about it while researching the book. And there are some interesting things about how video compression works in there (and a reference to Kanye West's video "Welcome to Heartbreak" which intentionally uses compression artifacts as a style), but somewhere in the hour or so of reading this chapter, you start to wonder what the hell happened to the narrative.
The chapters that I liked best were the ones I read all in one sitting--I read about a hundred pages on Sunday, and it seems like you need to take this book in chunks that big in order for him to wander back to his topic often enough to figure out what's going on.
He has some really interesting things to say about Kasparov v. Deep Blue, and how Kasparov wasn't defeated by the computer but by the book.
And he talks (a lot, actually) about Mystery and pick-up artists, since they've basically turned their conversations into computer algorithms. The sections on conversation--when it is most like computers (arguments) and when it is least--were the best. And he has a little bit of philosophy in there about how humans can't lose their jobs to computers unless they are already performing their jobs like computers.
There's one idea he described about how conversation is about giving handholds to the listener. (Like rock climbing.) A good conversationalist is someone who gives two or three handholds, two or three potential topics to pursue, which gives the other person options about what they want to say. Or, you could conversely decide that you don't want people to interrupt a story till the end, and so strip out handholds so they can't cut in.
I hadn't thought of it this way, but I very intentionally pay attention to handholds when I converse. For example, I think a good email to someone on an online dating site is one that gives two or three potential conversational launchpoints. I try to mention a couple of topics, and ask at least one question, so it's easy for the other person to answer. Or, for another example, if I am trying to tell an anecdote at a crowded party, I will map out in my head how far I think I'm going to get before I get hijacked. And if I don't think I can get to a point of impact, I choose something else to say.
So there are interesting things in this book. But--honestly--the interesting things have already been mined. I've heard several radio stories based on this book, and a few more brushing the same topics (on This American Life and Radiolab). Those stories were a lot better than this book is. Which shows what a really talented journalist can do with the material. Christian, on the other hand, mostly squanders it.
Author: Brian Christian
Genre: Non-fiction (Science and Philosophy)
Summary: Every year, the Loebner Prize tests Turing machines to see if any computer programs have yet reached the point of indistinguishability from real humans. To compare, of course, the judges evaluate a computer head-to-head with a human. The computer that gets the most votes is the Most Human Computer, and winner of the Loebner Prize. But there's a prize for the human interlocutors, too: The Most Human Human. Brian Christian signs up for the contest and launches on an interrogation of what the title "Most Human Human" even means.
Review: This is a fine book. Which is a huge disappointment, because it could have been excellent. It has one of the best premises--and best titles--of any book to come out recently. It got a lot of press, because the interest in the topic is immediate and obvious.
With all that, I wanted a story of the Loebner Prize and the author's quest for the Most Human Human award, along with some computer science and philosophy. I didn't get a story of the Loebner Prize--at all. He talks about leading up to it, makes a few references to the actual event randomly throughout, and then skips to the award ceremony, which was profoundly disappointing.
As for the computer science and philosophy, well... There's a lot of it.
Christian talks about human speech patterns, which, in spoken conversation, are overlapping, interrupting, digressing--everything but linear. Maybe it was his intent to write the book that way, but if it was, it's an impressive failure. There are some chapters that have a coherent direction. But most of them randomly wander off and never get to the point. You might find yourself suddenly reading about Aristophanes in Plato's Symposium and never quite figure out how that has anything at all to do with the Turing test. Sometimes he'll give a bit of a precis at the start of the chapter, but then he won't follow it.
And there are all these headers throughout. That contain no structural value at all. It seems like he wrote the thing and then just stuck a header in whenever a pun occured to him. Or whenever he realized that this page-long bit has nothing to do with anything in the text surrounding it. And they're all A-heads, so they don't help create a sense of hierarchy. And there are random epigraphs, sometimes in the middle of a section. I get it, dude, you studied philosophy. You know shit. But you need to work all of this material together into a coherent whole, instead of leaving all the lumps in the gravy.
So as you read, you find yourself in a forty-page long digression about data entropy and compression algorithms. Or not even a digression, really, because "digression" implies some sort of starting off point, and the only relation this seems to have to the purported topic of the book is that he read about it while researching the book. And there are some interesting things about how video compression works in there (and a reference to Kanye West's video "Welcome to Heartbreak" which intentionally uses compression artifacts as a style), but somewhere in the hour or so of reading this chapter, you start to wonder what the hell happened to the narrative.
The chapters that I liked best were the ones I read all in one sitting--I read about a hundred pages on Sunday, and it seems like you need to take this book in chunks that big in order for him to wander back to his topic often enough to figure out what's going on.
He has some really interesting things to say about Kasparov v. Deep Blue, and how Kasparov wasn't defeated by the computer but by the book.
And he talks (a lot, actually) about Mystery and pick-up artists, since they've basically turned their conversations into computer algorithms. The sections on conversation--when it is most like computers (arguments) and when it is least--were the best. And he has a little bit of philosophy in there about how humans can't lose their jobs to computers unless they are already performing their jobs like computers.
There's one idea he described about how conversation is about giving handholds to the listener. (Like rock climbing.) A good conversationalist is someone who gives two or three handholds, two or three potential topics to pursue, which gives the other person options about what they want to say. Or, you could conversely decide that you don't want people to interrupt a story till the end, and so strip out handholds so they can't cut in.
I hadn't thought of it this way, but I very intentionally pay attention to handholds when I converse. For example, I think a good email to someone on an online dating site is one that gives two or three potential conversational launchpoints. I try to mention a couple of topics, and ask at least one question, so it's easy for the other person to answer. Or, for another example, if I am trying to tell an anecdote at a crowded party, I will map out in my head how far I think I'm going to get before I get hijacked. And if I don't think I can get to a point of impact, I choose something else to say.
So there are interesting things in this book. But--honestly--the interesting things have already been mined. I've heard several radio stories based on this book, and a few more brushing the same topics (on This American Life and Radiolab). Those stories were a lot better than this book is. Which shows what a really talented journalist can do with the material. Christian, on the other hand, mostly squanders it.