Are computers too smart?
It is already way beyond the programmers, and going to get way more so.
Torch touched on some key issues with Asimov’s 3 laws.
The IBM system “Watson” that beat the best human jeopardy players learned how to do so, by reading all of wikipedia, and a lot else besides. The guys who programmed it’s many different learning systems, and the judgement systems that compared the results of different outputs from the different subsystems, didn’t know what it would do, they only knew the general sorts of probability algorithms in use, not what they would do with that mass of data and any particular question.
AI systems are becoming much more life like, and that will continue.
On current trends, Ray Kurzweil estimates that computers will exceed human level intelligence by 2029 (13 years).
They will use a lot more power to do so, and that will change.
We are going to live in a mixed world, of human, non-human, and part human intelligences.
We need to start seriously respecting intelligent life and its liberty.
Attempts to control will likely end up in unpleasant outcomes (just think how you feel when someone tries to control you, and you find out about it).
What works is being friendly, showing respect.
We need to apply that respect to all humans first, and soon; and then to all intelligence, human and non-human, biological and non-biological.
The biggest single barrier I see to that is our cultural attachment to the idea of money as a useful measure of value.
I understand that it was very useful in our past, and everyone else needs to get that it is very dangerous if taken very far into our future. We need to have individual life, and individual liberty as our highest values, and those need to reflect through all aspects of our social, political, legal and technical systems.
It all comes down to the choices we make – each and every one of us.