News for hackers. Stuff that matters.

throwawaysymbAI

My 2c:

I think the key term here is "concept formation" as well as "knowledge representation". How do we form concepts, and how are they represented internally to make them tractable?

Symbols are one way to represent concepts (or rather, point to them). But with symbols we are limited to surface-level transformations according to a syntax (I'm pretty sure Chomsky said something similar?). What do the concepts actually point to, though, and can we represent that underlying structure programmatically?

As I wrote in another comment, I'm very inspired by the conceptual spaces model:

https://mitpress.mit.edu/contributors/peter-gardenfors

https://www.youtube.com/watch?v=Y3_zlm9DrYk

Could someone please steelman my thinking here a bit? Would love to advance my own thinking on this matter.

What do you guys think of the conceptual spaces approach of Peter Gärdenfors?

See eg here:

https://mitpress.mit.edu/contributors/peter-gardenfors

https://www.youtube.com/watch?v=Y3_zlm9DrYk

From reading some papers, it seems his approach is a third way beyond symbolic and connectionist.

Indeed the title of that lecture is "The Geometry of Thinking: Comparing Conceptual Spaces to Symbolic and Connectionist Representations of Information"

Would you say conceptual spaces is a third way and how does it apply to the topic discussed in this thread?