[ExI] What are among the world's most important problems to solve, why?

Anders anders at aleph.se
Wed Jul 13 21:11:16 UTC 2016


On 2016-07-13 13:02, Dave Sill wrote:
> I asked "Are the most intelligent people the best problem solvers?", I 
> didn't say intelligence isn't a factor in problem solving. I was 
> challenging the assertion that achieving superintelligence is a very 
> important goal and the apparent belief that it enables solving 
> otherwise unsolvable problems.

Yes, that is the underlying thought for much of the superintelligence 
debate. Typically the starting point is an intelligence definition along 
the lines of Shane Legg's "intelligence is the ability to efficiently 
figure out how to achieve desired goals in general environment" (see his 
thesis for the mathematical statement, which is a bit stricter). So an 
entity better able to achieve its goals than humans in general will in 
general achieve its aims more successfully.

[ There are of course plenty of particular imaginable cases where the 
superintelligence doesn't want anything or wants to solve unsolvable 
problems, is in a situation where it cannot do anything useful (like 
falling down an elevator shaft), or it is simply unlucky despite its 
brilliant plan. There are also worlds where intelligence is not useful 
at all, and the no free lunch theorems, but those considerations rarely 
matter in practice in this world. ]

The "superintelligence is a superpower" thesis is based on the 
observation that human problem-solving has produced pretty 
impressive/world transforming results. In many competitive domains an 
entity that is better at achieving its goals than another entity will 
tend to win. Economically we know there is a premium for smart, skilled 
problem solvers.

Note that we do not have to assume godlike superintelligence for it to 
be extremely useful/good business/able to mess up the world. 
Intelligence explosions are entirely optional. Assuming it to be better 
in any domain is also a simplifying trick; clearly superintelligence 
specialized to smaller domains may be worth a lot too.

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

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20160713/342ab1ff/attachment.html>


More information about the extropy-chat mailing list