3 ms·
An interesting article, but (understandably, if frustratingly) very light on the details. It reminds me of some of the work being done in the Artificial General
by arnia 11y ago
An interesting article, but (understandably, if frustratingly) very light on the details. It reminds me of some of the work being done in the Artificial General Intelligence (AGI) community. In AGI, you are looking to come up with more universal approaches to mimicking intelligence, rather than bake-an-architecture-solve-a-specific-problem.
In particular, this reminds me of some of the operational logic work done in OpenCog (http://opencog.org/ http://opencog.org/) and, especially, Pei Wang's Non-Axiomatic Reasoning (http://cis-linux1.temple.edu/~pwang/papers.html http://cis-linux1.temple.edu/~pwang/papers.html). I've liked Non-Axiomatic Reasoning for a long time, and this sounds in the same broad area.
Pei Wang, incidentally, came up with the best operational definition of intelligence I've seen (and one I apply to other systems and approaches): Intelligence is the ability to act appropriately with limited knowledge and limited resources (including time and space).
For what it's worth, my (published, although very open to change) position is that a layered architecture (akin to old-school subsumption architectures) is probably going to be most effective here. Combining the 'symbol creation' abilities of a deep neural network with the evidential reasoning, generalisation, and planning capabilities of a cognitive layer will allow us to get the best of both worlds.
- mindcrime 11y agoFor what it's worth, my (published, although very open to change) position is that a layered architecture (akin to old-school subsumption architectures) is probably going to be most effective here. I have been going back and spending some time with the old "blackboard architecture" idea lately. I harbor a suspicion that that, or something like that, will turn out to be a useful way to integrate the capabilities of various different elements of cognition.
- arnia 11y agoYes, although it is probably going to have a notion of flow to it too. As someone who subscribes to the cognitive science idea of embodiment, the concepts in our head have meaning entirely because of how they relate senses to actions (however indirectly, and vice versa). I believe AI will be no different.
- elcritch 11y agoPersonally I haven't looked into the AIG fields above, however it appears "obvious" that layered "hybrid" approaches will be fundamentally more useful than either deep learning or expert systems alone. In that context the article doesn't seem light on details to me since it's primarily advocating layered approaches and I think there are only limited set of layered approaches that would be useful. Why would a practicioner in AI not agree with a hybrid approach being necessary is what I'm wondering. Are there arguments supported with theory or evidence to the contrary? The definition from Wang makes sense. How does AIG differ from other methods? For example, I noticed this article on Bayesian Program Learning a few days ago which IMHO takes a layered approach and is based on human learning models. [https://news.ycombinator.com/item?id=10723485 https://news.ycombinator.com/item?id=10723485] Interestingly there were no comments on it. Could you post refs to some the relevant publications your published opinions? I'm curious about your approach!