“Confabulation, or hallucination, is ubiquitous in AI/GPT output already. Efforts to correct this by self-learning algos and back propagation are unlikely to solve the problem because they add to the complexity of the system as a whole, which increases the likelihood of emergent ghosts. The difficulty is that duplicity is hard to detect unless you're a subject matter expert in the topic or you conduct your own research to test its accuracy. This begs the question — if you have to be a subject matter expert to spot the flaws in AI/GPT output, what good is the system in the first place?”