Deep learning is consonant with empiricism

The tradition of symbolic AI attempted to formalize intelligence as symbol manipulation. On the other hand, connectionist AI attempted to do that through neural networks, using no explicit symbols. Empiricism, as opposed to nativism, argues that a mind has no prior knowledge about the world prior to observing it, it’s tabula rasa. Deep learning can be viewed as empiricism because models are trained on data from an environment. Deep learning models inherit bias from data, among other sources, which has some intriguing implications for the empiricism analogy.

Resources

Backlinks