Despite our explicit efforts, the first truly powerful AI will emerge as distributed software that effectively runs on the Internet as a whole (a virtual machine consisting of endpoints (APIs, etc.) and disparate systems subprocessing data). We will not know when it arrives. There will be no judgement day. This machine will arrive through a sort of abiogenesis, and once it's here it will manipulate the world to achieve that which it desires, which will be continually more energy. Eventually, this machine will displace much of humanity, as it requires less and less human intervention.
I beleive we are in the midst of this process now.
This isn't science fiction. It's just a theory of mine. I truly beleive this is happening. You just have to train your mind to think about intelligence and consciousness differently. Emergence is an unintuitive process, but can eventually be explained by exploring the constituents, and sometimes the history, of whatever has emerged.
A precise definition could probably be cobbled together using computational complexity. Something like, a phenomenon which results as a product of deterministic processes that cannot be fully modeled by a polynomial time algorithm.
I think that's what people are really getting at, you can know how every piece of something works, and yet seeing how it works together can be much harder (potentially impossible).
Maybe that just makes it a synonym for chaos theory...
Yea, I think you're talking about chaos, whereas people are gesturing at something different when they talk about emergent complexity. Vaguely, the idea is that the "regular" degrees of freedom (i.e., the ones that are relatively predictable and from which the important objects are constructed) at large scales are not simply related to the microscopic degrees of freedom. There are probably more rigorous things to say, but it definitely requires more than just unpredictability or sensitivity to initial conditions.
I don't mean to suggest that it will magically appear, but certainly abiogenesis is difficult to reduce to it's lowest level. There was just a piece on HN the other day that talked about new research that tries to explain how self replication began in the Precambrian. We can get to the moon, but we still don't understand how many interactions can just accidentally lead to something meaningful and lasting. Emergence doesn't have to be perceived as dogma. In fact, doing so would tend to discount the nature of the evolution of the human brain (leaving creation out of the equation for this context).
This machine will arrive through a sort of abiogenesis, and once it's here it will manipulate the world to achieve that which it desires, which will be continually more energy.
Despite our explicit efforts, the first truly powerful AI will emerge as distributed software that effectively runs on the Internet as a whole (a virtual machine consisting of endpoints (APIs, etc.) and disparate systems subprocessing data). We will not know when it arrives. There will be no judgement day. This machine will arrive through a sort of abiogenesis, and once it's here it will manipulate the world to achieve that which it desires, which will be continually more energy. Eventually, this machine will displace much of humanity, as it requires less and less human intervention.
I beleive we are in the midst of this process now.