Faculty Associate Arvind Narayanan contends that Moravec's paradox, which holds that it is easy to train AI to do things humans find hard and hard to train AI to do things humans find easy, is more a statement of AI technologists' values than of fact. On his YouTube channel and on AI as Normal Technology, he writes: "Much of my writing is based on the idea that superhuman AI reasoning is a myth. Of course, you don’t have to accept this. But when you hear these predictions of imminent superintelligence, it’s helpful to understand the underlying mental model that many people in the AI field have. And it’s definitely useful to know that we have five or six decades of evidence that whether a skill such as reasoning is easy or hard for AI can depend a lot on whether you’re talking about a closed domain or an open domain."
You might also like
- communityInside the Black Box
