That's back to Turing test, right? We've kind of been trying to do that with "real" NLP tasks like QA all this time. The problem is, if the action depends on a complex chain of cognitive phenomena, the system could cheat, producing the right action with some shallow heuristic.
-
-
-
We already know that this is what we ended up getting for many of the costly large deep learning systems: right decisions (~actions) for the wrong reasons. It quacks, but it isn't a duck.
Kraj razgovora
Novi razgovor -
-
-
This perspective is consistent with the definition of intelligence that I find most useful - the ability to accomplish complex goals. This requires perception, decision-making and action.
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.