Re:Geoff Hinton Wants Computers to Think More Like Brains
by nradov on Thursday January 03, 2019 @03:22PM
(#18818926)
How many passengers died in the USA last year on Part 121 flights due to aircraft accidents caused by human pilot error? When do we expect to develop pilot AIs with 1/100 the error rate of humans? How will those pilot AIs cope with unexpected environment conditions or mechanical failures which aren't programmed in advance?
[ Parent
| on HN ]
Re:Geoff Hinton Wants Computers to Think More Like Brains
by modeless on Thursday January 03, 2019 @03:49PM
(#18819132)
Obviously this is a hypothetical example and I cannot predict the future. If you want to talk seriously then take self-driving cars, where humans kill over a million people per year and we are very likely to have AIs that would kill less than that, but still a nonzero amount, within 20 years.
In my view it would be unethical to stop the deployment of such AIs on the grounds that we don't "understand" them well enough or can't predict their behavior in every hypothetical circumstance, because we can still test them well enough to know with statistical certainty that they will save many, many lives.
Re:Geoff Hinton Wants Computers to Think More Like Brains
by nradov on Thursday January 03, 2019 @04:50PM
(#18819655)
Let's first see a prototype of such an AI for level 4+ autonomous driving on the majority of roads worldwide. Then we can quantify the potential benefits and have a meaningful discussion about the ethics of deployment. Until then it's premature and pointless since we don't know what the actual parameters will be.
Estimates about availability have consistently been overly optimistic. This may be a case where the first 99% of the work takes 99% of the effort, and the last 1% takes the other 99% of the effort.