3 ms·
>Several years ago I devised a little test to find out what style of thinking people use: Access your memory on church steeples. Most people will see a picture
by yters 7y ago
>Several years ago I devised a little test to find out what style of thinking people use: Access your memory on church steeples. Most people will see a picture in their mind of a generic "generalized" steeple. I only see specific steeples; there is no generalized one.
Yet, she identifies all these pictures of steeples as the same thing, which is an abstraction of the particulars. That is different than only associations between particulars in animal thinking.
Also, take care you are not making the genetic fallacy by referencing Dr. Egnor's background:
https://en.wikipedia.org/wiki/Genetic_fallacy https://en.wikipedia.org/wiki/Genetic_fallacy
- nabla9 7y ago> Identifies all these pictures of steeples as the same thing Associatively with no verbal toughs. Artificial neural networks can do the same. Understanding that two things belong to the the same category does not require language.
- yters 7y agoIt's essentially the same thing. A word is a label on a category. I'd also argue that what neural networks do is not the same thing as humans inferring categories from instances. 1. We prefeed the NNs with categories and examples. The main work is already done by the human. 2. What the NN learns from cat pictures is not 'cat', but a bunch of trivial visual features that happen to correlate with a bunch of cat pictures in a very brittle manner. Not at all like what happens when humans learn 'cat' from a few instances, and then can recognize cats everywhere in a huge variety of settings and transformations (as I'm learning with my toddler).