A vector isn’t a symbol unless your algorithm treats it that way http://tinyurl.com/ycqmowzl and, psst, functionally symbolic representations can be learned from unstructured, unlabelled data in an unsupervised way (lots of un) http://arxiv.org/abs/1810.01127
-
I agree here with the hunch that analogy making (as described by Douglas Hofstadter) is crucial. However, I'm still waiting for direct computational evidence of how this capability can be generated,
Have you done a Google Scholar search @IntuitMachine? Certainly @mmitchell_ai and Ken Forbus have done lots of work in this area.
-
-
The advantage over structured mapping and even LISA (from which DORA is descended - defining work by John Hummel) is that DORA / predicate learning can discover invariants a dataset and functionally symbolise them.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Yes I am familiar with the copycat and structured mapping work. I am of course focused on achieving this as an emergent capability if neural networks. Bottom up vs top down.
-
What is extra cool about DORA is the use of endogenous signals (neural oscillations) that arise from the processing of the dataset (during the comparison of patterns) to learn predicates
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.