3 ms·
No Technology – Not Even Tesla’s Autopilot – Can Be Completely Safe
- JoshTriplett 10y agoHumans can't be completely safe either. The important question isn't whether the technology can be "completely safe", but whether it's safer than humans. Even in the example given in the article of Three Mile Island: "and somehow, during that test, the workers had failed to reset it properly.". The technology absolutely should have detected and reported this human-caused issue; however, that doesn't make it a technological error, but rather a human error not mitigated by technology. (Of course, when talking about a nuclear reactor, you want to mitigate as many human errors as possible, just as you have technological components independently cross-check other technological components, to push the likelihood of a failure even closer to 0 than any one component could manage alone.) It's not possible to make anything "completely safe"; see also "0 and 1 are not probabilities" (http://lesswrong.com/lw/mp/0_and_1_are_not_probabilities/ http://lesswrong.com/lw/mp/0_and_1_are_not_probabilities/). Self-driving cars will be safer than human drivers even when interacting with unpredictable human drivers, and will be even more safe when they only have to deal with other self-driving cars. Hopefully cases like this don't slow the adoption of technologies that will, long-term, save lives (in addition to all the other improvements they may provide).