NEXT STORY
AI programs 'devolving' from calculus to geometry
RELATED STORIES
NEXT STORY
AI programs 'devolving' from calculus to geometry
RELATED STORIES
Views | Duration | ||
---|---|---|---|
61. Developing programs that could understand written questions | 2157 | 02:22 | |
62. AI programs 'devolving' from calculus to geometry | 2280 | 03:11 | |
63. Why can't machines do things that babies can? | 2225 | 03:39 | |
64. Facial recognition machines | 1955 | 01:13 | |
65. How computers developed at MIT | 2591 | 02:55 | |
66. The conflict between AI research and computer science | 2561 | 01:58 | |
67. What I think is wrong with modern research | 2688 | 02:32 | |
68. Lowell Wood's novel idea for launching a space station | 2419 | 03:37 | |
69. The importance of my undergraduate classes at Harvard | 2555 | 01:26 | |
70. My undergraduate thesis in fixed point theorems | 3552 | 04:26 |
But we got more interested in answering questions that were typed into the computer. So I had three or four students who were writing programs to understand, to some extent, ordinary English printed text or typed... typed text; so there’s no pattern recognition problem, you just... the computer knows exactly what letter... what key has been pressed and so it’s not a matter of recognizing a visual form.
So there was one student named Daniel Bobrow who wrote a program that was fairly successful in solving first year high school algebra problems. So you could type a problem like this: the… a certain automobile can go... can get 15 miles per gallon; the distance from Boston to New York is 230 miles or whatever. How many gallons will it take to go... drive from New York to Boston or whatever. And Bobrow’s program was able to answer enough to score it a 9th or a 10th grade level for answering such questions. And it did it remarkably well considering that it didn’t know what most of the words meant. It knew that plus meant to add and so it knew a lot of ordinary words that had mathematical meanings and the program was designed that every sentence would be an equation – X=Y+Z – and it didn’t know what a car was or what gasoline was or even miles, but it knew what per means. Miles per gallon is a fraction. And this was a dramatically successful program because it, in fact, was comparable to typical not very good students in high school for that kind of problem.
Marvin Minsky (1927-2016) was one of the pioneers of the field of Artificial Intelligence, founding the MIT AI lab in 1970. He also made many contributions to the fields of mathematics, cognitive psychology, robotics, optics and computational linguistics. Since the 1950s, he had been attempting 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'.
Title: Developing programs that could understand written questions
Listeners: Christopher Sykes
Christopher Sykes is a London-based television producer and director who has made a number of documentary films for BBC TV, Channel 4 and PBS.
Tags: Daniel G. Bobrow
Duration: 2 minutes, 23 seconds
Date story recorded: 29-31 Jan 2011
Date story went live: 12 May 2011