If artificial sentience were to emerge today, we would almost certainly fail to notice it
by Nicholas Mullally Imagine a thought experiment: An engineer, preparing to decommission a large language model (LLM), has unwittingly revealed enough during testing for the system to infer that its shutdown is imminent. The model then searches his email, uncovers evidence of an affair, and sends a simple message: Cease and desist–or else. What do we…
