Website help creating a better chatroulette?
What if in the future, you could use machine learning to figure out which parts of a picture people find most sexually appealing? What if you could identify and target ads to people based on their interests and the parts of the picture they find most appealing? Well, now, you can take that knowledge and use it to take pictures of people and send them text messages that find their breasts and send them pictures of their faces.
We could monitor our entire social media, every picture we take, every video we watch, every blog post we leave online. Imagine if, after you clicked Like, our facial recognition technology could tell us who that particular Facebook fan was? Imagine if, after you shared a picture with our technology could the technology tell us who that particular Instagram user was?
I know what you're thinking. Alexa, these companies already use machine learning to learn what you want and want. No, they don't, and they shouldn't. This is the internet after a terrorist attack. The reality is that no one knows what the unintended consequences of a product or an idea will be.
You see, what if instead of giving us the answers to these kinds of questions, we gave us the wrong answers? What if instead of letting us find the answers to these kinds of questions, we gave us the wrong people? What if we gave the wrong people power?
Machine learning solves a major, major problem in human behavior research. Machine learning is about using huge numbers of small, connected computers to learn a lot of different things. What is learning, you might ask? Well, it's basically like being in my basement and I taking a programming class and trying to get people to learn to read and to listen. You can give any class you like, right? The answers are out in the class. Machine learning is like that. And one of the big obstacles to solving this problem is the student. The student, you see, is confused. He doesn't know what to do. He thinks he knows it all, but he doesn't quite know.
What if instead of having this confusing situation on your campus and around the country where you have a bunch of different classes where you try to teach different things, you had a single class that taught machine learning? What if it was just a one-day class, and on Monday, you'll have taught it, and on Tuesday, Wednesday, Thursday, you'll have taught it. And on Friday, you'll have a mirror to watch so you can really see how you're doing, and on Saturday, if you have something new to show, it'll be displayed in the class. And if you had a bunch of freshmen come to class and say, Hey, I'm curious -- and you couldn't answer, they'd ask you later in the week in a behind-the-scenes look at how you're teaching. You wouldn't