Communications of the ACM
Toward Computers That Teach Themselves
Scientists are exploring approaches that would help computers develop their own sort of common sense.
Credit: Todd St. John
The artificial intelligence industry (AI) is experiencing a supervised learning revolution, as AI researchers teach computers to recognize patterns and learn from them.
However, computers will have to go beyond supervised learning, instead relying on observation or trial and error, in order to reach human-level intelligence.
Currently, supervised learning is constrained to narrow domains defined mostly by the training data.
While these methods have produced many practical applications—such as self-driving cars and language translation—supervised learning still cannot do many things that are simple even for toddlers.
Now, leading AI researchers have focused their attention on less-supervised methods. For example, reinforcement learning involves very limited supervision that does not rely on training data.
Self-supervised learning refers to computer systems that ingest huge amounts of unlabeled data and make sense of it without supervision or reward.
From The New York Times
View Full Article
Abstracts Copyright © 2020 SmithBucklin, Washington, DC, USA
No entries found