Adventures in AI with Lisik · Cursik · Chatik
From Measurement to Knowledge
An Experimental Investigation into How AI Uses Knowledge
How can we tell whether an AI model has acquired knowledge rather than simply learned to produce convincing answers?
Building on the measurements of Book Three, Book Four investigates this question through controlled evaluations, unfamiliar situations, decisions under uncertainty, and an ambitious space-mission planning experiment.
The research examines how an AI Teacher applies information, identifies missing evidence, explains its choices, and responds to new constraints. It also asks what we can conclude from the model’s behavior, and what remains unproven about how knowledge is stored or used inside the network.