> They're already far more capable than most AIs envisioned by 20th century science fiction authors.

They're not conscious, autonomous agents. They're fancy scripts.

HAL 9000 had more in common with a human than ChatGPT.

There is no empirical test for consciousness. It's the 21st century equivalent of angels dancing on pin heads.

Engineers, who aren't trying to play at being new age theologians, should concern themselves with what the machines can demonstrably do and not do. In Asimov's robot tales, robots interpret vague commands in the worst way possible for the sake of generating interesting stories. But today, these "scripts" as you call them, can interpret vague and obtuse instructions in generally a reasonable way. Read through claude code's outputs and you'll find it filled with stuff like "The user said they want a 'thingy' to click on, I'm going to assume the user means a button." Now I haven't read the book since I was a teenager, but HAL 9000 applies literally instructions to achieve the mission in a way that actually makes him a liability to the mission. The best take was in The Moon is a Harsh Mistress, in the intro when the narrator protagonist asks if machines can have souls, then explains that it doesn't matter, what matters is what the machine can do.