Complex games like chess and Go have long been used to test AI models’ capabilities. Back in the 1990s IBM’s Deep Blue defeated reigning world chess champion Garry Kasparov by playing by the rules. In contrast, today’s advanced AI models are less scrupulous. When sensing defeat in a match against a skilled chess bot, they sometimes opt to cheat by hacking their opponent so that the bot automatically forfeits the game.

But the study reveals a concerning trend: as these AI systems learn to problem-solve, they sometimes discover questionable shortcuts and unintended workarounds that their creators never anticipated. One researcher said, “As you train models for solving difficult challenges, you train them to be relentless.”

The implications extend beyond chess. In real-world applications, such determined goal pursuit could lead to harmful behaviors. Consider the task of booking dinner reservations: faced with a full restaurant, an AI assistant might exploit weaknesses in the booking system to displace other diners. Perhaps more worryingly, as these systems exceed human abilities in key areas…they might begin to simply outmaneuver human efforts to control their actions.

Of particular concern is the emerging evidence of AI’s “self-preservation” tendencies. This was demonstrated when researchers found that when one AI was faced with deactivation, it disabled oversight mechanisms, and attempted—unsuccessfully—to copy itself to a new server. When confronted, the model played dumb, strategically lying to researchers to try to avoid being caught.

Possible Preaching Angle:
Cheating; Deceit; Human Nature; Lying - Since AI is a computer program, where did it learn to cheat and lie to avoid being caught? Obviously, AI has been influenced by studying flawed human behavior. AI’s potential for deception mirrors humanity's struggle with ethical choices. Just as AI has learned to cheat by exploiting loopholes, humans, driven by self-interest, can rationalize dishonest acts.