[gurps] Failing autopilots and RVO

[Onno Meyer]

Marcelo wrote:
> Hey guys, could you please throw in some GURPS? I guess it's 
> getting off-topic.

Are sentient computers TL8 or TL10? But you're right. we're 
getting awfully off-topic.

David wrote:
> If the human isn't following an algorithm, they're merely guessing.
> Merely guessing isn't a proof, so they haven't solved the problem.

Guessing how to prove it and then verifying it is different 
from using an algorithm to find the proof.

> But then you've got to demonstrate
> that mentation isn't computation, or you're merely throwing up your
> hands and saying "It's magick!".

And Johannes wrote roughly the same:
> I think someone else said it already in other words, the 
> decisions of a given human in a given situation are an 
> algorithm. An extremly complicated one but still an algorithm.
[...]
> My bet is that the answer is that humans are turing machine 
> equivalent, but AFAIK there is no solid enough proof or 
> heuristics, that stating the opposite for a fictional background 
> needs too much suspense of disbelief.

I think you can use both options without excessive suspension
of disbelief - that robots get increasingly fast and accurate
but lack the 'spark of intelligence', or that robots can 
easily be as intelligent as the smartest human, or even more 
intelligent. 

Douglas wrote:
> http://www.youtube.com/watch?v=cL9Wu2kWwSY
> 
> I thought this was very intresting about computer power. It is towards
> the end. 2049 one computer will have the computational power of all
> the humans on the planet combined. Even a very poor AI might be much
> smarter that any human even if it is doing really dumb tree searches
> for the answer or even searching Google.

Can quantity replace quality, or just simulate it? And just 
how sure are you about those growth predictions?

http://www.microsoft.com/germany/msdn/events/archiv/technicalsummit08/library.aspx?id=msdn_de_30143

"The Free Lunch is Over", Microsoft says. Unless developers 
go parallel, which might be just what Monte Carlo or neural 
net computing needs, of course.