@fchollet's 'The Measure of Intelligence' arxiv.org/abs/1911.01547 is correct in that AGI tests should be with respect to anthropocentric capabilities and not some ill-defined universal intelligence. It's simple enough to create non-anthropocentric tests.
@fchollet The assumption that tests should be based on Spelke and Kinzer's 'Core Knowledge' is a reasonable idea. In general, we should design tests that can reveal cognitive skills that humans are expected to have.
@fchollet His definition of intelligence: "a measure of its skill-acquisition efficiency over a scope of tasks, with respect to priors, experience, and generalization difficulty". Said differently, how quickly can you formulate new concepts and procedures given what you already know.
@fchollet His Abstraction and Reasoning Corpus (ARC) is similar to the Raven Progressive Matrix test but shaped by the Core Knowledge priors. It is indeed a good idea.
@fchollet It's important to remember that AI has always made progress based on the cognitive tests that are made available. This is analogous to evolution. Organisms exist based on the challenges demanded by an environment.
@fchollet @threadreaderapp unroll
