The main flaw in superintelligence arguments is to think of intelligence as an intrinsic property of an agent, like the height of a tower.
-
Compression is a attractive, even universal, notion for learning. However, it also ties down to a prior (which is harder to formalize).
I take the "cognition == compression" concept seriously, but not literally. My actual views are a lot more nuanced ;-)
-
-
It is not enough to represent available data in a compact form, one must also discard massive amounts of data, & pick wisely what to discard
-
So cognition is compression, but a very lossy form of compression tied to a specific objective function that is not intrinsic to the data
End of conversation
New conversation -
-
-
I take it quite seriously, even literally, actually. :) Atleast in terms of its ability to "in principle" compress.
-
But it is not clear to me how to formalize this "in principle" beyond abstract computational models.
- Show replies
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.