Friday, July 07, 2006
Don’t give them hands
There are two fears that arise commonly in discussions of artificial intelligence: that the AI could put us all out of work, and that it could turn against us. We could avoid both possibilities if we scrupulously restricted the output devices of AIs to include only printers, computer monitors, and stuff like that. Then the artificially intelligent could only put us out of jobs that require thought, and they could only take over mankind if they convinced humans to help them do it.
Of course, it might not be possible for intelligence to arise without constant feedback from a manipulable world. If not, we would have to create the AIs in a virtual physical reality (where they could have all the hands they wanted), and only later introduce them to the real world.
Deciding which output devices should be considered safe might be tricky. For instance, if it were possible through the proper combination of radio signals to, say, take control of an airplane, then radio transmission would not qualify as a “safe” form of output. Ditto the internet.(Isaac Asimov proposed a different leash: Simply program the machines to work in the best interests of humankind. I think that would be great, if only it were possible.)