Guy is building self-driving car in his garage

[Jacob]

This guys sounds pretty smart... http://www.bloomberg.com/features/2015-george-hotz-self-driving-car/

Hotz's approach isn't simply a low-cost knockoff of existing autonomous vehicle technology. He says he's come up with discoveries—most of which he refuses to disclose in detail—that improve how the AI software interprets data coming in from the cameras. "We've figured out how to phrase the driving problem in ways compatible with deep learning," Hotz says. Instead of the hundreds of thousands of lines of code found in other self-driving vehicles, Hotz's software is based on about 2,000 lines.

The major advance he will discuss is the edge that deep-learning techniques provide in autonomous technology. He says the usual practice has been to manually code rules that handle specific situations. There's code that helps cars follow other vehicles on the highway, and more code to deal with a deer that leaps into the road. Hotz's car has no such built-in rules. It learns what drivers typically do in various situations and then tries to mimic and perfect that behavior. If his Acura cruises by a bicyclist, for example, it gives the biker some extra room, because it's seen Hotz do that in the past. His system has a more general-purpose kind of intelligence than a long series of if/then rules. As Hotz puts it in developer parlance, " 'If' statements kill." They're unreliable and imprecise in a real world full of vagaries and nuance. It's better to teach the computer to be like a human, who constantly processes all kinds of visual clues and uses experience, to deal with the unexpected rather than teach it a hard-and-fast policy.

The guy turned down millions of dollars from Musk, though they still have a conversation going.

[The Brain]

Neural network? Whoop-de-fuck-do.

[Baron von Schtinkenbutt]

Neural network? Whoop-de-fuck-do.

The second best solution to any problem!

[DGuller]

Neural network? Whoop-de-fuck-do.

The second best solution to any problem!

Deep learning networks are the #1 solution to any pattern recognition problem.  And a long, long way ahead of #2 solution.

[Baron von Schtinkenbutt]

Deep learning networks are the #1 solution to any pattern recognition problem.  And a long, long way ahead of #2 solution.

Depends on how you rank solutions.  In most practical use cases, MaxEnt and SVM outperform deep learning/neural networks because MaxEnt and SVM models are easier to train, and in some cases faster to classify.

[The Brain]

There's only one way to settle this: a nerd-off.

[DGuller]

Deep learning networks are the #1 solution to any pattern recognition problem.  And a long, long way ahead of #2 solution.

Depends on how you rank solutions.  In most practical use cases, MaxEnt and SVM outperform deep learning/neural networks because MaxEnt and SVM models are easier to train, and in some cases faster to classify.

By accuracy.  I don't recall ever seeing an image recognition predictive modeling competition being won by something other than a variant of deep convolutional neural nets.  That said, I've never taken part in one, nor have I done that for work, so I may be missing some nuance that is important in practice.

[The Brain]

I can recognize basic shapes, like a cat or a tree.

[DGuller]

I don't think anyone doubted your familiarity with various animals. 

[Zanza]

I very much doubt that the autonomous driving software of his competition, be it Google or a car manufacturer, is based on long lists of if/then/else statements. Maybe his software is better, but he is not the only one that tries to have his car learn.

That said the hardest part about self-driven cars is making them road legal and covering liability. I doubt his system is able to do either at the moment. Getting to 99% precision for your self-driven car is not too hard these days. It's the other .9999% that make it hard to achieve.

The Bloomberg article sounds like the journo that wrote it is completely clueless with regards to AI or software development.

[Malthus]

I can recognize basic shapes, like a cat or a tree.

Yes, much to the regret of the cat. 

[Josquius]

Hope it works out for him.  The whole garage project keeping secret how everything works thing....sounds like it has a big potential to go bad.