Let's develop AI that won't destroy us? Kid VArious is all for that!
Machine Minds - Forbes.com
Animals to which humans are indifferent often lose their habitats and go extinct. If an indifferent superintelligence is created, especially if it is created as software and can therefore reproduce very rapidly, humans will be similarly endangered. After all, while most minds may not care about us, all possible minds will depend upon, and thus care about, certain basic resources such as matter. Superintelligences could be designed to care about us. In this case, they would protect us from our worst mistakes, including the mistake of creating an indifferent or thoughtlessly designed superintelligence.
Extra work will be required to design such concern into an artificial mind. The Singularity Institute for Artificial Intelligence was founded to make sure that such work is done before it becomes necessary. I hope we have centuries. Most experts say that 20 to 50 years is more likely. Once artificial intelligence is an imminent prospect, competitive pressures will probably favor those projects that compromise safety in order to progress faster. Many projects will justify such compromises on the grounds that if they don't hurry, even more reckless competitors will develop artificial intelligence first. Similar things happen all the time, but humankind can't afford this sort of race to the bottom. We will only get one shot at this. Let's get it right the first time.