“In 2001 Arthur C. Clarke not only envisioned the future of artificial intelligence, but he also foresaw the ‘uncanny valley’ in which AI becomes creepy–and dangerous–when it is oh so close to human intelligence but not fully there,” said Ray Kurzweil.

“HAL realized he had made a mistake and then imagining that he must be perfect, felt that he would be disconnected for having made a mistake and reasoned that he had better turn off his crew mates first. That was perfectly logical, but demented.

“We will indeed need to go through the uncanny valley before we get to Turing test-capable AI, so we should take Clarke’s warning seriously.

“I had the pleasure and honor of discussing this with him using a ‘teleportec’ virtual reality system from Sri Lanka and I will always treasure my dialogue with him.”