"We conclude that the LLM has developed an analog form of humanlike cognitive selfhood."
Slack.
I was just using one (the mini at DDG) that declared one very small value for a mathematical probability of an event, then (in the next reply) declared a 1-in-1 probability for the same event.