Quote:
Originally Posted by traptor
I don't think that's true necessarily.
AlphaGo showed that AI can figure out new things, but typically in very specific areas. It learned new strategies in the game of Go by playing millions of games against itself.
So while some AI mainly helps manage lots of data, others like AlphaGo can discover new ideas within a set of clear rules.
And yes, “AI” is often overhyped, but there are different types, from simple data tools to focused but strategic systems like AlphaGo. It depends on how the AI is built and trained.
|
Not to derail too much I read Nexus and Yuval Harari spoke about how AI is great at a few things but not at others and actually used the go example.
He calls it
Known Knowns - things we know we know, we understand the problem and the solution or how to build it. Examples are playing Go, optimizing delivery routes etc.
it'll be interesting how something like this plays out for hockey, because it partially falls into what Harari calls
Known Unknowns, or things we know we dont know. His example is consciousness, emotions and general intelligence. We know they exist but no clue how to replicate it.
I bring that up just because we hear the term Hockey IQ so much. But do we really know how to define that or make a goal set? to be able to have AI help we need to work to understand what exactly it means and how to track it. But back to the go example, maybe we need the supercomputer to connect data points that we know are there, but our brains cant connect themseveles (similar to new techniques to win go). Either way, its super interesting imo.