For any task in ARC, it's possible for a human to write a reasonably short computer program that can handle new pairs. Everything is straightforwardly computable -- unlike, say, classifying MNIST digits. There's no human advantage in any single task.https://github.com/fchollet/ARC
-
-
Starting with few priors and adapt very well to narrow context is not trivial. Is that intelligence? It depends how it would survive in some other environments, like some life forms can strive in wide varieties of context. I don’t think it’s useful to draw a precise line.
-
Intelligence is an information processing mechanism for learning the regularities of the environment that uses the knowledge to attain goals. This applies to narrow or general scenarios depending on magnitude of priors, dynamics and capacities of the agent
End of conversation
New conversation -
-
-
So in what context does that 5 year old, without personal prior experience, predictably fail? What is essential to few-shot learning
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
As you know, I think 'intelligence' is a property of behavior rather than of agents. Successful behavior in the face of novelty is certainly intelligent, but so is, for example, rapid and correct speech recognition.
-
That may be so,
@tdietterich , yet „generality“ or „abstract reasoning“ as proposed by@fchollet still seems to be a useful criterion to define a future research program, is it not? Metaphors rarely convey the full argument so why not let them be?
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.