Comment by AndrewBissell | Hacker News Reader

Comment by AndrewBissell | Hacker News Reader

Parent Full thread

> His argument is that to solve self-driving you need a deep understanding of your surrounding which you can only achieve with visible light spectrum video. That's the real hard problem to solve

Musk's argument is more that cameras should be sufficient because humans can drive using only two eyes to perceive the driving environment. He always neglects to mention that humans do this with a combination of sight and a brain capable of general intelligence. I'm sure it's true that if Tesla invents AGI, self-driving with just cameras will become tractable. But "real hard problem to solve" doesn't begin to capture the difficulty.

In reality, since no one has yet invented a self-driving computer, it's impossible to say what components are necessary or even whether there may be more than one way to skin the cat. But one source we should probably take with a grain of salt on this issue is those (like Musk) with an intense commercial interest in one perspective.

You could have made the same argument about any hard problem before it was solved.

"So this guy says a machine can outplay chess with just a CPU, some memory and a bit of code. He neglects to mention that humans have a brain capable of general intelligence".

Year later, computer beats human in chess.

"So this guy says you can teach a computer to play Go just by unsupervised training of neural networks. He neglects to mention that humans have a brain capable of general intelligence".

Year later, computer beats human in Go.

"So this guys says you can program neural network to play computer games competitively using vision and deep learning. He neglects to mention that humans have a brain capable of general intelligence".

We don't need AGI for self driving.

The difficulty of self driving is probably less than a dog walking on the street.

No, a dog doesn't steer a car, because he doesn't have hands, but he's performing the same vision and planning tasks like a human (or AI) driving a car.

He knows where he is, he knows where he wants to go and he uses vision and his non-AGI brain to plan a path to get there while also avoiding dynamic, unexpected obstacles.

Do you have any counterexamples which aren't finite games where all the potential moves and outcomes can (at least in theory) be exhaustively enumerated at any point in time?

You can approximate most everything in theory almost perfectly by a set of very large but finite moves. Driving a car can be approximated as a set of finite moves (ie: 0.001 degree changes in wheel position) done at finite intervals (say 10000 "frames" per second). That'd give you more precision and a finer time grain than even a human brain is capable of.

And you don't need a 'very large' set of moves in the first place. 100 wheel positions are enough to give you 1-degree accuracy near neutral, and moderate but sufficient accuracy for strong turns. If you're looking at a delta from the previous position, 30 will do the job. Multiply that by 10 to 40 positions for the pedals, run the whole thing at 20Hz, and you're good to go.

There is Google's Starcraft ai ( https://deepmind.com/blog/alphastar-mastering-real-time-stra... ), the real time nature makes the potential search space absurdly large.

kjksf did give an example: Go. The state space there is much larger than could ever be enumerated before the heat death of the universe.

That's why I said "at least in theory." The fact remains that everything theoretically possible in Go is known ahead of time, and furthermore it is trivial to generate training data and run simulations to improve the algorithm without putting human lives at risk.

It seems to me that one of the difficulties we have with making robots is trying to model them too closely after ourselves. We don’t have a humanoid maid, but we do have a roomba; similarly for lots of industrial automation. Home automation doesn’t look like c3po walking around your house flipping light switches.

It seems like maybe not the best reasoning to say “humans, do it this way, so that’s how my robot should do it.”

Except a Roomba is much more limited than a human maid even in the constrained task of vacuuming the floor. And for a self-driving car you do need human levels of performance. Nor can you cheat and redo the infrastructure to accommodate simpler system like with home automation (ie: replace light switches withe relays).

Except that you could most definitely repurpose some infrastructure to better suit AD. Having a somewhat protected AD only lane in cities would allow, today even, autonomous transport to be utilized.

Reply on news.ycombinator.com