Communications of the ACM
Finding Trust and Understanding in Autonomous Technologies
In 2017, we shall have to decide whether we can trust autonomous technologies.
Credit: Eric Risberg/AP
Autonomous technologies are spreading beyond the transportation sector and into health care, advanced cyber defense, and autonomous weapons, writes Carnegie Mellon University professor David Danks. However, he believes that whether autonomous technologies become truly ubiquitous will depend on how much they can be trusted.
Because of their autonomy, society will have to consider the importance, the value, and the challenge of learning to trust these technologies in the same way people trust other human beings.
Danks says as a general rule, society should want autonomous technologies, including self-driving cars, to behave in ways that are predictable and expected. However, he notes true autonomous systems can make their own decisions and plans, and therefore can act differently than expected. For example, DeepMind's AlphaGo won the second game of its recent Go series against Lee Sedol in part because of a move that no human player would ever make, but was nonetheless the right move.
Danks notes one consistent theme of machine-learning research over the past 20 years has been the significant gains made by not requiring artificial intelligence systems to operate in human-like ways. He says machine-learning algorithms often have been able to outperform human experts by focusing on specific, localized problems, and then solving them differently than humans do.
From The Conversation
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA
No entries found