[ExI] Fwd: Re: AI risks

[Anders Sandberg]

On 2015-09-06 23:50, BillK wrote:
> On one side you have the techs trying to devise some sort of ethical
> AI that won't harm humanity in general. On the other you have humans
> deliberately (with collateral damage) trying to harm other humans and
> they see AI as a powerful weapon.
The problems are different kinds of problems. Without AI safety the 
systems are dangerous tools for people who see them as weapons (or 
business advantage, or a secret sauce for academic triumph - ascribing 
malign motivations to these people is too limiting). Risk-accepting 
players may race to develop powerful but unsafe AI which may be a 
lose-lose situation.

But solutions to one problem doesn't fix the other. Making AI or people 
using them nice are orthogonal.

-- 
Anders Sandberg
Future of Humanity Institute
Oxford Martin School
Oxford University