As I understood more about how Type 1 / Type 2 processing is defined and what the Global Workspace Theory for consciousness is about. I came to realized while Type 2 cognitive processing and consciousness explained by GWT share a striking similarity, namely, they both require the working memory to hold and process elements coming from unconscious T1 processing, they are not the same thing.
If we are only interested in furthering AI development by leveraging knowledge from psychology, consciousness might offer less value than Type 2 because the gap between current SOTA DL and human level AI is probably type 2 processing.
I took back the statement that consciousness might offer less value. Even tho consciousness is only a typical correlate of type 2 processing, it appears that all type 2 processing are conscious (we can’t say the opposite is true tho).
The gap can also be broken down and illustrated through the below three questions:
- How do we solve out-of-distribution generalization and transfer?
- How do we achieve high-level semantic representations which can be used to deal with compositionality and causality?
- How do we build agents that seek knowledge, understand causality and better their world models on their own?
Gang Fang's Blog 
Leave a comment