> Some may say game-playing AIs such as CICERO are benign, but such thinking is short-sighted; capabilities developed for game-playing models can still contribu

> Some may say game-playing AIs such as CICERO are benign, but such thinking is short-sighted; capabilities developed for game-playing models can still contribute to the proliferation of deceptive AI products. What is short sighted is writing articles that manifest hypothetical situations without basis. As most developers know, "software" can do just about anything, including bringing joy, or creating suffering. Is it possible to code up something that "tricks" the user? Absolutely. Humans drive the outcomes, usually, and it will be humans that deceive other humans, not AI doing it on its own (at least at this time). Having intent to deceive is TOTALLY different than deceiving as a by-product of playing a game, or having conflicted requirements and an objective outcome as a measurement for success or failure.