That large proportions of people have reading skills lower than high-end NLP a) has been demonstrated by the @oecd b) hasn’t been taken adequately into account by the #agi #singularity #Futureofwork panic crowds.
-
you said "a cognitive model" but anyway I wasn't assuming a single model, you're the one who just made the straw person. Please don't be defensive, I do respect you & honestly want to communicate.
“I don't think there's one shape of model all humans & no machines use” is what i was responding to. everyone who reads The Hobbit comes away with some idea of who Bilbo is & who Gandalf is & what is unusual about the ring. Can any AI do that?
-
-
I can't think of a more exciting technology challenge in this day and age. Imagine being a part of a team that figures this out. Are you up to date Gary with the research that went into IBM's debater project? Super interesting stuff there, a pile of published research on it too
-
i have been trying hard to push people towards this challenge since 2014 and would love to see real progress here. (
@rogerschank has emphasized the crucial importance of narrative comprehension for AI for decades.) - 1 more reply
New conversation -
-
-
Not everyone can make sense of those books, but of course you are right that humans are more likely to make similar models given similar hardware phenomenology inputs contexts capacities etc. My point was that we value humans for being humans, this is an aspect of that identity.
-
not about similarity, it’s about not devaluing the human ability to derive cognitive models from natural language input - at all. study you mentioned conflates literacy w ability to interpret discourse; no current machine can do so. comparing illiterate humans to AI distorts.
- 1 more reply
New conversation -
-
-
Our mental models certainly differ but can be broadly the same if we have similar experiences. The neural architectures that drive model creation will all be of the same type. Ie all functioning humans can tell left from right, before from after etc.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
It seems there is a mismatch or clash between the concepts AI framework uses to understand human capacities and the concepts humans actually use. AIs won’t complain if they are misrepresented, so it might be easier to fall for simplistic theoretical frameworks than with humans.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Actually I bet you could get info like Bilbo and Gandalf's approximation to heroes, underdogs, and wizards with just word embeddings. "Ring" might be harder unless you were good at identifying unusual usages relative to some norm. Again, not full human-level identification, but!
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
I think a mirror is a cheaper and equally hard test for
#AI. Self-consciousness but for machines (it can be very different from our's). The weirdness about the ring comes from introspection and thinking about our dark impulses. -
Self awareness is trivial with RAM. Programming self recognition is trivial. Signs that arr interesting in other animals are false friends in designed artefacts.
- 2 more replies
New conversation -
-
-
Everyone knows no AI can do that. The question is, does Marcus know how to build an AI that can do that? Or does he know how to build an AI that pretends to do it by comparing text to bits of knowledge in a symbolic database? If so, it's the Chinese Room all over again.
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.