Born in 1927, Marvin Minsky is one of the pioneers of the field of Artificial Intelligence, founding the MIT AI lab in 1970. He has also made many contributions to the fields of mathematics, cognitive psychology, robotics, optics and computational linguistics. Since the 1950s, he has attempted to define and explain human cognition, the ideas of which can be found in his two books, "The Emotion Machine" and "The Society of Mind". His many inventions include the first confocal scanning microscope, the first neural network simulator (SNARC) and the first LOGO 'turtle'.
One problem with my field is that people publish experiments where the machine does succeed in learn… apparently learning something – maybe really learning, perhaps – but they don’t publish failures when they say: 'Well, we tried to get this machine to distinguish between cats and dogs and we had 4,000 pictures and it didn’t do much better than chance.' We never see papers published, or hardly ever. And so, there’s something a little bit wrong. In physics, you win a Nobel Prize if you can show that some old theory isn’t quite right. In artificial intelligence, you might lose the company that’s paying for your research if you publish a failure. And so that’s another reason why we need more institutions that are not based on quick payoffs.
Title: Why we should publish failures in AI research