15.5.3.1 Human level artificial general intelligence will be common.
Assuming that computing power continues the exponential doubling it has historically enjoyed, we can expect the capacity of a 2030 supercomputer to be available “on the desktop” (or whatever the appropriate metaphor will be) by 2050. Under this assumption, I think it reasonable to conclude that software entities that appear to have human level intelligence will be commonplace by the midpoint of this century.
I believe that the “strong AI” verses “weak AI” debate, touched on in Chapter 14, will be a significant social issue by 2050. As you may recall, the difference between strong and weak AI is not based on behavior but on whether humans choose to view an AI as “really” intelligent or just “a good fake”. If we believe our machines to be truly intelligent, then the issue of AI rights becomes very important. However, because the question of whether an AI can ever really be aware, in the same sense that humans are aware, has no objective answer, the issue will probably be approached on moral and religious grounds. Depending on whether the AIs derive from human brain simulations or whether they are engineered from first principles may tilt the debate one way or the other. Regardless, it seems clear that conflict is almost certain to arise.