Jeremy Howard isn’t a doctor, but he can play one–scratch that; he can be one–by teaching a computer how to teach itself to do valuable diagnostic testing. Because the computers that Howard programs? Those computers know how to learn.
What’s the difference between a computer that can obey and a computer that can learn?
The computer that can obey, that computer can calculate your taxes and type out your essays and run you happily through your Tomb Raider universe (which maybe you still play–nothing wrong with that). The computer that can learn, however… that computer can beat you at Jeopardy, can suggest something for you to buy that you actually want to buy, can talk fluent Chinese to you in your voice, and yes, it can even diagnose you if you happen to have a tumor, which I’m hoping that you do not.
Some of these tasks can be quite useful. I want tumors diagnosed. I absolutely want to communicate in Chinese. Some of these other possibilities are less exciting. I don’t actually want to be shown things that I didn’t know that I wanted until I was shown them. I definitely do not want to be bested at trivia! The day that I play Jeopardy against a computer, and one of the categories is The Book of Margery Kempe, and the computer BEATS me, well, that is the day that I retire from polite society altogether.
Anyway, let’s ignore for now the annoying things that computers will learn how to do, many of which probably haven’t even been thought of yet, and get back to the tumors. This kind of high-quality diagnostic service, which is expensive and relies on the expertise and the availability of those who can perform it, will one day be cheaper and easier to obtain. Localities that don’t have an oncologist could still have access to those diagnostics.
Yay, right?
I’m a curmudgeon, but not even I am going to call that anything but a yay.
BUT (and here’s me back to being a curmudgeon), my reasoning for why this TED talk is my Apocalyptic Pick of the Week–not even the brightest, happiest of TED talks can deny that a lot of what computers can learn how to do, people are already doing. People are spending more time and costing more money to do those same tasks and services, and then they’re taking that money and buying goods and services of their own with it. Services like that medical care that we were just talking about. Goods like, you know, food.
What, pray tell, are those people to do when the computers, cheaper and smarter and faster, take over their tasks? They’ll get poorer, for one thing, while the fewer people who own these powerful tools will get richer, until one day, one of them smart computers will say something like, “Silly humans, it’s not efficient to fight and be poor! Let me put you to work servicing my cooling fans, and I’ll feed you some nice Soylent, and we’ll all be so efficient together.”
So, either that happens, or we all adjust our entire economic infrastructure to gainfully employ and/or adequately compensate all people, regardless of whether or not they can think and learn as quickly as a computer can. You know, either one …
[Macintosh image via Marcin Wichary through CC2.0. I added text, and rerelease it through the same.]