Self-driving cars can be fooled by fake signals

[bananas]

93. Longship, you complained when I posted "The Revolution Will Not Be Formalized"

earlier this year, which was about a related subject - automated theorem proving in mathematics.

And I think the point you're trying to make about autonomous driving is similar to the point that article was making about automated theorem proving: they might work well in many circumstances, but there are also many circumstances where they just won't work at all.

You stumbled on the first paragraph because of the jargon, so I'll just quote these other three paragraphs:

http://www.democraticunderground.com/1016123486

The Revolution Will Not Be Formalized
https://golem.ph.utexas.edu/category/2015/05/the_revolution_will_not_be_for.html

May 18, 2015
The Revolution Will Not Be Formalized
Posted by Mike Shulman

<snip>

What is the future of computer-verified proof? Is it the future of mathematics? Should we be happy or worried about that prospect? Does it mean that computers will take over mathematics and leave no room for the humans? My personal opinion is that (1) computer-verified proof is only going to get more common and important, but (2) it will be a long time before all mathematics is computer-verified, if indeed that ever happens, and (3) if and when it does happen, it won’t be anything to worry about.

The reason I believe (2) is that my personal experience with computer proof assistants leads me to the conclusion that they are still very far from usable by the average mathematician on a daily basis. Despite all the fancy tools that exist now, verifying a proof with a computer is usually still a lot more work than writing that proof on paper. And that’s after you spend the necessary time and effort learning to use the proof assistant tool, which generally comes with quite a passel of idiosyncracies.

Moreover, in most cases the benefits to verifying a proof with a computer are doubtful. For big theorems that are very long or complicated or automated, so that their authors have a hard time convincing other mathematicians of their correctness by hand, there’s a clear win. (That’s one of the reasons I believe (1), because I believe that proofs of this sort are also going to get more common.) Moreover, a certain kind of mathematician finds proof verification fun and rewarding for its own sake. But for the everyday proof by your average mathematician, which can be read and understood by any other average mathematician, the benefit from sweating long hours to convince a computer of its truth is just not there (yet). That’s why, despite periodic messianic claims from various quarters, you don’t see mathematicians jumping on any bandwagon of proof verification.

<snip>