I wonder if that quote is still applicable to systems that are hardwired to learn from decision outcomes and new information.
What (or who) would have been responsible for the Holodomor if it had been caused by an automated system instead of deliberate human action?
LLMs do not learn as they go in the same way people do. People's brains are plastic and immediately adapt to new information but for LLMs:
1. Past decisions and outcomes get into the context window, but that hasn't actually updated any model weights.
2. Your interaction possible eventually gets into the training data for a future LLM. But this is incredibly diluted form of learning.