Dr. Peter Norvig is Director of Research at Google. He spoke about the difficulty and inaccuracy of prediction, as well as his thoughts on how AGI will be developed.
Prediction by experts has been found by some researchers to be less effective if the expert knew a lot about particular subjects or were driven by underlying grand-unifying ideas. In effect, they become overconfident, thus clouding their ability to make predictions. The lesson learned from these observations is that any individual can have an opinion as good as an expert. Regarding the Technological Singularity and AGI, Norvig advised audience members to read, compare alternative views, keep up with the topics, and come to their own conclusions.
Norvig provided a list of AGI prequisites:
- Probabilistic First-Order Logic – deal with uncertainties
- Hierarchical representation and problem solving
- Learning over the above
- With lots of data
Rodney Brooks asked Norvig if Google had observed any unexpected emergent behaviors that might suggest an emerging intelligence as sometimes depicted in science fiction movies. While there were unexpected problems and lessons learned, no unexpected emergent behaviors have been observed to date.