Well, many researchers understand this and are actively trying to build systems which has these inductive biases.https://twitter.com/GaryMarcus/status/1200646264150323200 …
i actually don’t know any systems working in some of them (eg affordances of objects) and would love references. meanwhile i do think there is overall bias in ML against building in innate priors, with a frequent goal of mastering things “without human knowledge”
-
-
Re: “mastering things without human knowledge” — progress in ML has always been about using human knowledge. These manifest as inductive biases — ConvNets, for instance, are built on the human intuition that objects of interest tend to be spatially clustered.
-
yes - but people sometimes title their papers otherwise. see my arxiv critique of AlphaZero https://arxiv.org/abs/1801.05667
- 5 more replies
New conversation -
-
-
What about this paper by Christopher Manning http://et.al ? We introduce the Neural State Machine, seeking to bridge the gap between the neural and symbolic views of AI and integrate their complementary strengths for the task of visual reasoning. https://arxiv.org/pdf/1907.03950.pdf …
-
looks interesting! thanks!
End of conversation
New conversation -
-
-
@jshristov recent work on building in knowledge of relational representations to explain and repeat plans taught to a robot is a good move towards this. https://arxiv.org/abs/1907.13627Thanks. 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.