Transcript for:
The Journey of Intelligent Systems

Modern thinking about the possibility  of intelligent systems all started with   Turing's famous paper in 1950. He, of course, knew  that he couldn't define what intelligence was,   so because of that, he introduced  what he called the Turing Test.   The idea was that if a human couldn't tell  within five minutes if he was talking to   a computer or a person, then the computer  would be said to have passed the Turing Test. Turing couldn't imagine the possibility  of dealing with speech back in 1950,   so he was dealing with a teletype, but much  like what you would think of as texting today.   And that was because Turing knew that he  couldn't actually define what intelligence   was. It's too hard. It's too slippery. So  that's why he introduced the Turing Test. But I've read that paper many times and I  think that what Turing was really after was not   trying to define intelligence or a test  for intelligence, but really to deal with   all the objections that people had about why  it wasn't going to be possible. What Turing   really told us, was that serious people can  think seriously about computers thinking and   that there's no reason to doubt that computers  will think someday. That day is approaching. About 10 years after Turing published his paper  in 1950, important laboratories were set up by   Marvin Minsky and John McCarthy, by Allen Newell  and Herbert Simon. McCarthy's approach at Stanford   was to start with mathematical logic: he spent  his whole life trying to bend logic to his will.   Newell and Simon focused on modeling human  thinking. They developed systems that solve   simple puzzles and work out simple problems in  a manner that they believed was consistent with   human experiments. Minsky's approach was harder to  characterize. He believed that one representation,   method, or approach — no one of those could  deliver a full understanding of intelligence. That was the central message of his seminal  paper, which was titled "Steps Toward Artificial   Intelligence", in 1961. You know, in retrospect,  we can think that Turing told us we could do this   and that paper by Minsky told us what to  do. So that's why Turing and Minsky are   often regarded as the real pioneers, the  real founders of the field of artificial   intelligence. Well any event that brings us  to what some people call AI's first wave. Early in 1960s, James Slagle wrote a program that  integrated symbolic expressions. He was trying to   model what a freshmen does at MIT when they learn  that kind of mathematics. Because Slagle's program   performed so impressively, it's what I consider  to be the signature program of AI's first wave.   That key idea was called 'problem reduction'. The  idea is simple: you just take a hard problem and   you break it into simpler problems, and then you  break those simpler problems into problems that   are still simpler until you've got something  you can just do. That's what problem reduction   was about, but it's only one of a cornucopia  of ideas that have emerged from AI research. At MIT, the work of Slagle was quickly  followed by other successes, and by 1970   programs understood drawings, they learned from  examples, they knew how to build structures,   and one even answered questions  much like Siri and Alexa do today.