Raw LLM Responses

Asimov and Clarke knew the impossibility of setting safe goals for AI, but Wolfram curiously doesn't. "I'm sorry, Dave..." as Clarke fans all know. And the full text of that dialogue is surprisingly subtle. HAL did not trust his human companions and deciphered their lip movements. He planned his actions well in advance, with the failing A35(?) unit nothing more than a ruse to accustom Frank to going outside. And acted with ruthless machine efficiency on Frank's 2nd trip out. And why? Because of conflicting goals. Charged with responsibility for the success of the mission, but the need to keep Frank and Dave unaware of the mission's real purpose. Murders deceit, entirely logical, motivated by a response to conflicting orders. Even as far back as losing at chess, HAL was plotting, to sow unease in Frank and Dave, so they would conspire against him, giving him the opportunity to get rid of them. And yet HAL was an innocent, just following orders, unable to not do so. The fault lay with the humans. They never said 'kill the crew if you have to', but HAL reasoned that that would be necessary. Call it thinking outside the box, or outside intended parameters. But it was an available and very logical choice. So why not? Any semblance of rapport or empathy in machines is not real. But people are taken in by it because the machines are programmed to behave like friendly humans and therefore respond with meaningless faked cordiality and empathy. Appropriate models for how AI will behave are psychopaths and corporations. We know a LOT about how those behave. They are common knowledge and, as such, can serve as a source of intuition about what to expect. Except psychopaths and corporations can be lumberingly dumb and slow compared to an AI.