The Turing Test as an assertion of AI-completeness:pic.twitter.com/UTtIwDL5cf
Show this thread
A few rough musings on AI-completeness and related ideas: "What if we had oracles for for common machine learning problems?" http://cognitivemedium.com/what-if-oracles-ml …
-
-
-
I explain my skepticism that _any_ narrowly scoped problem can be AI-complete. Something like machine translation or reinforcement learning may demand fantastic understanding. But it doesn't necessarily demand that the understanding be surfaced in all the ways we expectpic.twitter.com/k9qpboO3He
Show this thread -
New conversation -
-
-
This Tweet is unavailable.
-
This Tweet is unavailable.
-
-
-
I wonder which problems are narrowly scoped. Does editing math papers count? Explaining jokes?
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
We may not have oracles but we can have the statistics of oracles!https://github.com/andrescorrada/ground-truth-problems-in-business/blob/master/classification/IndependentBinaryClassifiers.pdf …
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Said another way - how would you know that you had a perfect classifier in most data situations? There are gold standards of labeling but they, themselves, are error prone. You need ground truth inference algorithms in the real world.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
And another interesting result from polynomial methods for ground truth (oracle) inference - imperfect classifiers can be used to detect perfect classifiers!
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
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.