Just four weeks after my last episode, here is a new (63 min.) Blogging Heads TV episode, this time with Brian Christian, author of the new celebrated (NYT, New Yorker) book The Most Human Human: Christian describes competing in a Turing test to be seen as more human than rival chat-bots, and he offers his approach in that contest as a general philosophy of life: we should all try to live our lives to be hard for machines to mimic.
Sheesh, this guy just wrote a whole book on this stuff and he seems put in an uncomfortable place by your relatively simple questions. Every time you pose something to him, he says "Ehh, I dunno, maybe like... I'm trying to think... I guess it's like..." And then comes out with some weak response that doesn't sound good.
"But I just can’t see why one should try especially to be hard for machines to mimic, over and above the other usual reasons for our actions."
--Agreed!
I am reminded of Jaron Lanier's 'You Are Not a Gadget' here, where he argues that in order to make machines seem smarter (because we want to think of them as 'thinking', and we want to analogize our minds to machines), we 'dumb ourselves down' in the process. And we stop telling the truth about machines.
For instance, he argues, Google Search is a dumb machine that you have to manipulate to get it to do what you want it to do, he thinks. We forget this because we get caught up in the good things it does, and use language that suggests it is smarter than it is.. like 'knows', Google 'knows' what I am about to type... and so on... 'You are not a gadget!' he asserts, and thinking that you are does a disservice to humanity, and distorts our perception on how well machines are actually doing.
I agree that we would all be better off if we all were well-versed in rationality, constantly seeking truth through a cold analysis of facts. But in a world where most people reflect very little on what they do and why they do it, I don't have a problem with Brian Christian entertaining people while giving them an idea to think about.
The more predictable you are, the easier it is for others to manipulate you for their own purposes.
This can be a good thing, as mutual predictability facilitates cooperative enterprises, but it can also be a bad thing if you focus on achieving some minor goals, via this efficient "robotic" method, and you lose sight of the bigger picture in which perhaps you're just being used and taken for a fool. (E.g. the husband with a fixed schedule offers more opportunities for being cheated, than the one with an unpredictable schedule.)
This isn't an argument against being robotic per se, but just a potential large cost that needs to be taken into consideration when one is moving in that direction. But I'm guessing that in general it is strategic error to be overly predictable in all contexts, hence the worry most feel about this sort of thing.
You're right. I'm sure that chatbot authors are working on that. I think that they will also come with a much larger playbook of standard conversation topics. I had an incredibly charismatic friend who worked in a leadership position that led him into many personal conversations with smart people. He consciously worked on skills of remembering what people said and directing the conversation in a way that seemed most compatible with what he learned through this. He also explicitly worked on developing a playbook. The way he put it: "I want to be able to say one interesting thing about every possible topic." He would seek out interesting, knowledgeable people and harvest information from them. With me, he'd just ask something like: "So tell me an interesting observation about the NHL playoffs this year." Then he'd ask some follow-up questions, and that was enough. It might sound mercenary, but he was one of the most charming people I've ever met, because he honestly loved people, loved chatting and loved information. Sometimes I wonder how hard it would be to make a Stuart-bot.
The real irony is that a large part of the "machine-like" behavior is memoryless behavior:
just responding to the last thing the other person said
customer service systems that make agents follow scripts, won’t give agents much discretion, and keep switching who each customer deals withBut computer systems have huge amounts of memory. At this point even consumer-level systems have far more memory (if I include disk space) than their human operators have. The real question is when programmers will learn the right set of tricks to take full advantage of the memory hardware we already have.
Not much, usually. Reading is entertainment for 99% of us.
I realize you might be a real truth seeker, but most Iibertarian econ profs confound me much the way pop science writers confound you.
My cynical side questions why so many of them, so quick to spot the signalling nature of education and to deride the waste of the state, chose to nestle in like ticks under a fold of panniculus at Universities, adding to the world's bountiful heap of peer-reviewed knowledge and MBAs.
Fuck it, I got me a kindle full of Gladwell, a fishing pole and a boogie-board I can paddle out past the territorial waters and rename "Truthistan." See you chumps later!
Both your argument and his argument remind of a book I just read: "The Rebel Sell: why the culture can't be jammed" by Joseph Heath and Andrew Potter. Basically it's a long critique of the "counter cultural" idea that by striving to be more original, more individual, we are striking a blow against the system. In fact such efforts (a) totally misunderstand the system and (b) in as much as they inspire you to differentiate yourself by purchasing things, feed into the most hated aspect of the system: consumerism.
I've love to see you review the book, I think it'd be right up your alley :-)
I feel your time is to precious to be paired with this type of blogging heads. I think a pairing on wasteful domestic policy and signalling with an expert, or on the future with Prof. Verner Vinge would be more productive.
This was the exact puzzle piece I was missing, thanks!
I agree.
Sheesh, this guy just wrote a whole book on this stuff and he seems put in an uncomfortable place by your relatively simple questions. Every time you pose something to him, he says "Ehh, I dunno, maybe like... I'm trying to think... I guess it's like..." And then comes out with some weak response that doesn't sound good.
"But I just can’t see why one should try especially to be hard for machines to mimic, over and above the other usual reasons for our actions."
--Agreed!
I am reminded of Jaron Lanier's 'You Are Not a Gadget' here, where he argues that in order to make machines seem smarter (because we want to think of them as 'thinking', and we want to analogize our minds to machines), we 'dumb ourselves down' in the process. And we stop telling the truth about machines.
For instance, he argues, Google Search is a dumb machine that you have to manipulate to get it to do what you want it to do, he thinks. We forget this because we get caught up in the good things it does, and use language that suggests it is smarter than it is.. like 'knows', Google 'knows' what I am about to type... and so on... 'You are not a gadget!' he asserts, and thinking that you are does a disservice to humanity, and distorts our perception on how well machines are actually doing.
I agree that we would all be better off if we all were well-versed in rationality, constantly seeking truth through a cold analysis of facts. But in a world where most people reflect very little on what they do and why they do it, I don't have a problem with Brian Christian entertaining people while giving them an idea to think about.
"What's wrong with being robotic?"
The more predictable you are, the easier it is for others to manipulate you for their own purposes.
This can be a good thing, as mutual predictability facilitates cooperative enterprises, but it can also be a bad thing if you focus on achieving some minor goals, via this efficient "robotic" method, and you lose sight of the bigger picture in which perhaps you're just being used and taken for a fool. (E.g. the husband with a fixed schedule offers more opportunities for being cheated, than the one with an unpredictable schedule.)
This isn't an argument against being robotic per se, but just a potential large cost that needs to be taken into consideration when one is moving in that direction. But I'm guessing that in general it is strategic error to be overly predictable in all contexts, hence the worry most feel about this sort of thing.
I would be happy to let the mechanizable things I do be done by a machine. The other stuff (in my job) is more fun anyway.
You're right. I'm sure that chatbot authors are working on that. I think that they will also come with a much larger playbook of standard conversation topics. I had an incredibly charismatic friend who worked in a leadership position that led him into many personal conversations with smart people. He consciously worked on skills of remembering what people said and directing the conversation in a way that seemed most compatible with what he learned through this. He also explicitly worked on developing a playbook. The way he put it: "I want to be able to say one interesting thing about every possible topic." He would seek out interesting, knowledgeable people and harvest information from them. With me, he'd just ask something like: "So tell me an interesting observation about the NHL playoffs this year." Then he'd ask some follow-up questions, and that was enough. It might sound mercenary, but he was one of the most charming people I've ever met, because he honestly loved people, loved chatting and loved information. Sometimes I wonder how hard it would be to make a Stuart-bot.
Beware of acting like a machine at work - or one will soon take your job.
The real irony is that a large part of the "machine-like" behavior is memoryless behavior:
just responding to the last thing the other person said
customer service systems that make agents follow scripts, won’t give agents much discretion, and keep switching who each customer deals withBut computer systems have huge amounts of memory. At this point even consumer-level systems have far more memory (if I include disk space) than their human operators have. The real question is when programmers will learn the right set of tricks to take full advantage of the memory hardware we already have.
"What more could a reader want?"
Not much, usually. Reading is entertainment for 99% of us.
I realize you might be a real truth seeker, but most Iibertarian econ profs confound me much the way pop science writers confound you.
My cynical side questions why so many of them, so quick to spot the signalling nature of education and to deride the waste of the state, chose to nestle in like ticks under a fold of panniculus at Universities, adding to the world's bountiful heap of peer-reviewed knowledge and MBAs.
Fuck it, I got me a kindle full of Gladwell, a fishing pole and a boogie-board I can paddle out past the territorial waters and rename "Truthistan." See you chumps later!
Both your argument and his argument remind of a book I just read: "The Rebel Sell: why the culture can't be jammed" by Joseph Heath and Andrew Potter. Basically it's a long critique of the "counter cultural" idea that by striving to be more original, more individual, we are striking a blow against the system. In fact such efforts (a) totally misunderstand the system and (b) in as much as they inspire you to differentiate yourself by purchasing things, feed into the most hated aspect of the system: consumerism.
I've love to see you review the book, I think it'd be right up your alley :-)
I feel your time is to precious to be paired with this type of blogging heads. I think a pairing on wasteful domestic policy and signalling with an expert, or on the future with Prof. Verner Vinge would be more productive.