Tourist
That article you link to is very interesting, but I don't think it has relevance to this discussion.
Sorry but that's called "moving the goalposts" - some parts of the article may not be strictly relevant to this discussion but very relevant to your quite strongly stated assertion that "there are millions of optical illusions. Computers don't suffer from them"..
I'd think much more important is the researcher's comments that "what we don’t have, what we still need, is a better understanding of what’s really going on inside these neural networks.” That would suggest to me at least we are nowhere near the level of confidence required if we're planning on neural networks being used in fully autonomous airliner ops.
Ultimately I understand where you're coming from, and one day you'll be right, but I do think you're being unrealistically optimistic if you think that day is less than 30-40 years away.