tziteras

I've been following some of my friend Matthew's "Reality Report" where he is currently doing a series on the dangers of runaway AI with Eric Brulyant, who was on his way to the machine intelligence research institute the San Francisco Bay Area in California to a workshop on super intelligence and value alignment: Making sure that when you create an all purpose "general intelligence" it has ethical values that are compatible with human values. They imagine what would happen if a general or crackpot inventor somewhere, came up with something resembling a higher intelligence and quickly commercialised it and got it running. If programmed with simple rules like "Make everyone happy", a few code cycles down the line, it will be injecting everyone with happy pills, and it will somehow see around things. Even something as simple as "calculate further digits of PI" would result in it using up more resources in order to make its calculations an...