4.1 What artificial intelligence is, and the four ways to define it

Work through these

• Place a definition of AI in the four quadrants: thinking or acting, humanly or rationally

  Definitions of the field split along two questions: is the goal thinking or acting, and is the standard human behaviour or ideal behaviour. Placing any definition into that grid tells you what its author cares about.

• Explain the Turing test, and why it is neither reproducible nor constructive

  The famous imitation test defines intelligence by whether a machine can pass as human in conversation. It is memorable, and it is neither repeatable as an experiment nor useful as a design goal.

• Argue why the field settled on the rational-agent approach

  The field converged on defining an agent as one that acts to achieve the best expected outcome, because that standard can be measured and engineered toward. Human imitation cannot.

• Describe where AI stands today, and the problem that once something works it stops being called AI

  Where the field stands now, and the recurring pattern that a capability stops being called intelligence as soon as it works reliably. That pattern explains a lot of the discourse around this subject.