It’s kinda funny cause usually isn’t it the AI agent that has a misaligned goal? Like when I say don’t die, and it discovers that pausing Tetris technical means you never die. But now it’s students that have been given the wrong goal: pass the test by whatever means (e.g. use AI).
It’s kinda funny cause usually isn’t it the AI agent that has a misaligned goal? Like when I say don’t die, and it discovers that pausing Tetris technical means you never die. But now it’s students that have been given the wrong goal: pass the test by whatever means (e.g. use AI).