At the end of the talk I stood up and made the comment that it was obvious that the picture with the tanks was made on a sunny day while the other picture (of the same field without the tanks) was made on a cloudy day. I suggested that the “neural net” had merely trained itself to recognize the difference between a bright picture and a dim picture.
This is still not a source because it’s a recollection 50 years later and so highly unreliable, and even at face value, all Fredkin did was suggest that the NN might have picked up on a lighting difference; this is not proof that it did, much less all the extraneous details of how they had 50 photos in this set and 50 in that and then the Pentagon deployed it and it failed in the field (and what happened to it being set in the 1980s?). Classic urban legend/myth behavior: accreting plausible entertaining details in the retelling.
Another version is provided by Ed Fredkin via Eliezer Yudkowsky in http://lesswrong.com/lw/7qz/machine_learning_and_unintended_consequences/
This is still not a source because it’s a recollection 50 years later and so highly unreliable, and even at face value, all Fredkin did was suggest that the NN might have picked up on a lighting difference; this is not proof that it did, much less all the extraneous details of how they had 50 photos in this set and 50 in that and then the Pentagon deployed it and it failed in the field (and what happened to it being set in the 1980s?). Classic urban legend/myth behavior: accreting plausible entertaining details in the retelling.
I’ve compiled and expanded all the examples at https://www.gwern.net/Tanks