Communications of the ACM
Microsoft Research Chief: AI Is Still Too Stupid to Wipe S Out (and Will Be For Decades)
Chris Bishop, Microsoft's director of research at Cambridge, says warnings mankind is on the brink of developing The Terminator's Skynet and other homicidal AIs are 'nonsense' and will be for decades to come.
Credit: Ed Miller/Microsoft
Microsoft Research Cambridge laboratory director Chris Bishop dismisses the fear artificial intelligence (AI) is on the cusp of overtaking human intelligence, and says it will continue to lag human performance for decades to come.
"Yes, deep learning has achieved human-level performance in object recognition, but what does that mean?" Bishop asks. "It means the machine makes about the same number of errors as the human."
Bishop stresses even vaunted examples of machine intelligence, such as Google DeepMind's Go-playing system, have to be understood within the context of the immense time and manpower invested in their development.
He also cites the common misconception that machine-learning systems' ability to perform some of the individual tasks people can do means they are on the brink of matching more general human abilities. Imperial College London professor Maja Pantic says this myth is debunked by the fact that building generalized systems capable of solving any possible problem was proved impossible.
Although Bishop downplays fears of homicidal AIs annihilating humanity, he acknowledges more mundane dangers and risks fundamental to the technology. For example, Bishop notes the opaque nature of deep neural networks raises the possibility the AI's decisions could be shaped by unknown biases stemming from the vast amount of data on which such systems are trained.
From TechRepublic
View Full Article
Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA
No entries found