True, and that's why quite a few people are looking into hybrid statistical and symbolic machines. There is hope :)
Bingo. This has been the crux of my critique for two decades.https://twitter.com/timkindberg/status/1002437369356791808 …
-
-
-
Exactly as I have been urging, repeatedly, since 1992. Glad to see it finally becoming a focus.
End of conversation
New conversation -
-
-
Apologies for the naive question, but isn't a simple perceptron like y = w x enough for this problem, where x is the order in which the input is given and w is set to 2? Isn't this a trivial function to learn, once the input x is encoded as suggested?
-
I discuss at length in chapter 3 of The Algebraic Mind, and urge you to actually try to get a sense of the interplay between representation, algorithm, and generalization.
- 1 more reply
New conversation -
-
-
But if each integer is analogous to a separate but related human behavior then an ML approach telling you 2,4,6,8 are similar to 10 but not 11 is useful in discerning that there may be a rule to relate them even if it doesn't tell you what the rule is.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
this is not telling the whole story. one has to better define A and B in terms of where that data comes from. In A what is that set? What generates it? It’s not hard to make a statistical machine that tries random generating functions to statistically match the set.
-
And better example would be “pi”. Pi has a symbolic and a numeric/statistical nature. For its numeric (ratio of radius and circumference in Euclidean plane) concept stats machines are all we can use. For its symbolic (polar/cycle) concept all we can use is algebra.
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.