AI FUTURES (Qualcomm VP, Technology): May 23, 2018

AI can presently do an individual thing better than humans. Examples:

• Speech recognition
• Image analysis: identify objects; state what people are doing; analyze medical images and detect melanoma
• Beat human champions in GO and chess

AI will likely be good at "Multi-agent systems". Examples:

• self driving cars: understand other drivers on the road and intersecting roads

AI deep learning will be applied to reasoning. Examples:

• what happens next?
• what caused this to happen?

FUTURE:

• HOME: Ubiquitous sensors talking to each other and anticipating your needs.
• FACTORIES: Machines optimally cooperating, and detecting failure before it happens.

CEILING: energy consumption: exponential increases in speed and hardware will exponentially increase power

RESEARCH on 3 types of machine learning:

• unsupervised - with no labels, a machine tries to find repetitive patterns
• supervised: - provided labels, have the machine predict whats in the next image
• Reinforcement: a machine acts within the world. It makes decisions/performs actions, picks things up and learns; plays a game or drives a car.