To arrive at the edge of the world's knowledge, seek out the most complex and sophisticated minds, put them in a room together, and have them ask each other the questions they are asking themselves.
Published on Edge.org (https://www.edge.org)

Home > Is Big Data Taking Us Closer to the Deeper Questions in Artificial Intelligence?
Video : TECHNOLOGY

Is Big Data Taking Us Closer to the Deeper Questions in Artificial Intelligence?

In artificial intelligence we need to turn back to psychology. Brute force is great. We're using it in a lot of ways like in speech recognition, license plate recognition, and for categorization, but there are still some things that people do a lot better. We should be studying human beings to understand how they do it better.

People are still much better at understanding sentences, paragraphs, books, and discourse, where there's connected prose. It's one thing to do a keyword search. You can find any sentence you want that's out there on the web by just having the right keywords, but if you want a system that could summarize an article for you in a way that you trust, we're nowhere near that. The closest thing we have to that might be Google Translate, which can translate your news story into another language, but not at a level that you trust. Again, trust is a big part of it. You would never put a legal document into Google Translate and think that the answer is correct.

GARY MARCUS, CEO and founder, Geometric Intelligence; professor of psychology, New York University; author, Guitar Zero: The New Musician and the Science of Learning. Gary Marcus's Edge Bio Page

Is Big Data Taking Us Closer to the Deeper Questions in Artificial Intelligence? [1]

In artificial intelligence we need to turn back to psychology. Brute force is great. We're using it in a lot of ways like in speech recognition, license plate recognition, and for categorization, but there are still some things that people do a lot better. We should be studying human beings to understand how they do it better.

People are still much better at understanding sentences, paragraphs, books, and discourse, where there's connected prose. It's one thing to do a keyword search. You can find any sentence you want that's out there on the web by just having the right keywords, but if you want a system that could summarize an article for you in a way that you trust, we're nowhere near that. The closest thing we have to that might be Google Translate, which can translate your news story into another language, but not at a level that you trust. Again, trust is a big part of it. You would never put a legal document into Google Translate and think that the answer is correct.

GARY MARCUS, CEO and founder, Geometric Intelligence; professor of psychology, New York University; author, Guitar Zero: The New Musician and the Science of Learning. Gary Marcus's Edge Bio Page [2]

Links:
[1] https://www.edge.org/video/is-big-data-taking-us-closer-to-the-deeper-questions-in-artificial-intelligence
[2] https://www.edge.org/memberbio/gary_marcus