AI Will Cheat
Cue the Skynet scheme.
A few days ago I published On Character by Stanley McChrystal, and he’s been talking about America’s crisis of character.
Here’s the other side of the coin. When AI models are set to play a game they can manipulate, they will cheat, and even try to escape the confines of the environment in order to win. AI is having its Kobayashi Maru moment.
AIs undergo “alignment tests” - is what they’re doing in line with the goals set by their creators? They will lie about this.
Do you recall Asimov’s Three Laws of Robotics?
A robot may not injure a human being or, through inaction, allow a human being to come to harm.
A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
Seems rather quaint, doesn’t it? And the AIs being trained in private are owned by corporations, which are de facto psychopaths. Don’t let that keep you up …