Complexity Explorer Santa Few Institute

Lecture: Artificial Intelligence

Lead instructor:

Your progress is not being saved! Enroll now or log in to track your progress or submit homework.

2.2 Lecture II. » Summary

In the following seminar, Dr. Mitchell discusses her research-in-progress about current attempts to help AIs learn a generalizable form of intelligence. For an AI to be adaptable and truly "smart", it should be capable of abstraction and analogy - to be able to extrapolate from the situations it has seen before to unseen contexts and domains. Dr. Mitchell introduces recent work in the field in which AIs are trained to solve classic human "intelligence" (or IQ) problems as well as novel measures of "intelligence". The thread throughout the lecture and discussion is how we should define "intelligence" for an AI. This, and many other questions raised in the lecture, remain to be answered.