RE: My Simulation Theory Hypothesis
I wrote a few blogs about this topic last month. I know the issue and idea has been going on long before Bostrom and Musk, they just happen to be big names that are giving it momentum within the scientific community.
Super intelligence within our lifetime I doubt. It may happen, but decades from now so that would still be within our lifetime. AI is a passion of mine. Not so much the moving robots about and such, but how do you simulate a conscious mind. There are a lot of things we still are clueless how to approach from a computational perspective. If we solve some of those things then I see no reason why we couldn't have a super intelligence. Our tech is growing exponentially, it is really only some mysteries of the mind that have prevented this from already occurring. Or perhaps it has occurred in a DARPA lab or somewhere secret already and we don't know about it.
Deep Thoughts. :)
Have you checked out Bostrom's work summarizing the expert opinions in the field so far? Chances are good it'll happen in the next handful of decades, according to them. Guys like Ray Kurzweil have a pretty good record of predicting stuff like this also.
Yeah I've been following it too. I also watched the big panel on it with Neil DeGrasse Tyson. The problem with that panel is they could have had some people that knew more about actually making complex computer simulations like happen in some games on the panel. So many of the things that they were thinking would stop it from being true are actually resolved if you know a lot about procedural generation. There was only one person on NDT's panel that seemed somewhat familiar with it.
I do think it COULD happen IF we can solve those few remaining odd things... if we can solve them then it could happen today.
Basically... what is consciousness, what is common sense...
However, thinking about it... does super intelligence necessarily mean something with common sense or consciousness? If it does not then I believe that actually might be doable pretty soon... If it is pure processing facts, logic, and numbers we'll be at the Super Intelligence level with that side of computing soon. It is the more nebulous things that I haven't seen clues of them solving yet.
We haven't even solved that with us.
I don't think AI has to conform to our concept of consciousness. The thing is, we don't really understand (or care) how machine learning works, we just know that it does. I also don't believe it's possible for us to embed any type of empathy for the human condition into machines. I believe they will come their own conclusions about how to move forward. The idea that they might be benevolent is, in my opinion, on par with the how we feel about ants. We don't go out of our way to destroy ants unless they are bothersome to us. A superintelligence will be able to achieve its goals at a rate we cannot even comprehend. Imagine being able to "spin up" a society of millions or billions of humans to learn how to interact with each to achieve the desired end goal. Social engineering at its finest. Nice article Luke.
Thanks Bill. :)
Reminds me of this interaction in the movie Contact:
But yes, whenever people claim Super Intelligence couldn't possible do things like create art or know about human morality or emotions or whatever, I also think about them simulation all of human history a billion times over and being able actually know us better than we know ourselves. They might even be able to "solve" morality for us and help us improve our well being (and not in the Brave New World Soma way either... at least, hopefully not).
Oh yes I agree on all counts. When I said consciousness, and common sense I was not implying empathy with us. Those are simply things we do not quite understand about ourselves.
Raw computation, logic, etc we understand those things and I believe we could have an AI super intelligence on that level very soon. For all we know DARPA or some place like that may already have one.