[katsura]

I've been thinking about building a robot that can use a camera to look around, use motors to go in different directions, and when it sees a human, it could also ask if they've seen John Connor, and if the person is being "difficult" then press a button to terminate them.

The interesting thing is that the three laws of robotics says that robots shouldn't harm humans, but I don't really see a way for an AI agent to understand that by "pressing a button" they actually hurt the human.

[voidUpdate]

You have stumbled upon the point of the three laws of robotics, which is that they are part of a series of stories showing why they don't necessarily work

[gryfft]

To wit, the three laws are actually a formulation of three laws of tool design: a tool must not harm its user; a tool must be fit for purpose and do what the user wishes, as long as that doesn't harm the user; and a tool should be sturdy and reusable so long as that doesn't interfere with the tool's safety or usability.

These design principles make sense when you are talking about a non-sentient object, but intelligent, adaptable beings cannot be so easily constrained.

[rvnx]

At some point (~50 years from now ?) they could even form their own type of life. If they can mine for resources, think, do actions and reproduce. "synthetic life"

[diggan]

> If they can mine for resources, think, do actions and reproduce. "synthetic life"

Essentially the story of the Horizon series of video games: https://en.wikipedia.org/wiki/Horizon_(video_game_series), and I'm sure many other sci-fi novels.

[rvnx]

Or like in Futurama, the apparition of "Robosexuals"