[theoriginaldave]

I share some of the same sentiment.

In Roger Williams's "The Metamorphosis of Prime Intellect", a scientist didn't understand the statement made by his pet AI, so he opened up the debugger to see the decision tree and set of axioms that led to the decision, and was able to debug it, prune the logic and make adjustments.

I wish I could do this with ChatGPT. The way human beings reason is as opaque as ChatGPT.

When creating learning systems, I think it should be required to include the capability to visualize the thought process which leads to a result. As much to debug the system as to glean insights into reasoning in general.