Amusing summary <https://www.computerworld.com/article/3836728/you-thought-genai-hallucinations-were-bad-things-just-got-so-much-worse.html> of research into how bad AIs can get in some of the rubbish results they produce. We knew they could lose their hold on reality and start making things up, but did you know they could also cheat, and tell lies about that cheating when confronted with it? For example, when given charge of doing share trades, and told not to (illegally) take advantage of insider information, they might do so anyway, and then simply deny it when challenged. Oh, and the phrase “vegetative electron microscopy” is complete nonsense; it originated when an AI was processing multicolumn text, and saw words next to each other which were in separate columns, not part of the same sentence at all, and ran them together anyway. From there, the phrase somehow spread to other publications ...