2020-02-07
Link
Two methods:
- The methods that leverage computation .
- The other methods that leverage human knowledge .
These two need not run counter to each other, but in practice they tend to. There are two reasons:
- The researchers wanted methods based on human knowledge input to win and were disappointed when they did not.
- They always tried to make systems that worked the way the researcher thought their own minds worked.
The Truth:
- the actual contents of minds are tremendously, irredeemably complex.
- the outside world are arbitrary, intrinsically complex.
- breakthrough progress eventually arrives by an approach based on scaling computation by search and learning.
Search and learning are the two most important classes of techniques for utilizing massive amounts of computation in AI research. Essential to theses methods is that they can find good approximations, but the search for them should be by our methods, not by us. We want AI agents that can discover the unknown domain like our human can, not which contain what we have discovered.
Examples:
Deep learning methods rely even less on human knowledge, and use even more computation, together with learning on huge training sets, to produce dramatically better the systems.
