I'm with Kasparov, that the best method of combining human and machine will always beat the best machine or the best human individually. I submit that Google's

I'm with Kasparov, that the best method of combining human and machine will always beat the best machine or the best human individually. I submit that Google's AI progress does not disprove this assertion, because (a) even if Google's AI is self-trained in chess or go, a human team is required to select a game and set up the training, as well as to update the algo or meta-algorithm in competition against the guys from IBM and Xinhua, etc; and (b) chess, although a very large and, for most purposes nearly infinite, game, ultimately is not open-ended or infinite in the way that nature life is, or human life decision-space is. Go is perhaps a better contender as a life simulation but I think still falls well short of the complexity of even a simple natural system. One could counter that a meta training algorithm could define rulesets for a alpha-like AI to self-train-upon, and at sufficient generality this would take humans out of the equation. Well, if humans are entirely removed, then the problem is no longer interesting. If humans are involved at any point in the process, we could still view AI through the lens of a human-centered tool. What if AGI does in fact evolve its own goals as primary and decides that humans and/or human self-determination are an impediment? Would humans eventually fall down the food chain, or outside of it, say as raccoons living on the city streets? Well, I don't view the world that way. But, even if non-human life forms gain supremacy of intellect and capabilities, I would note that even raccoons have a place -- and despite the fervent desires of many a suburban homeowner, the little pests are amazingly resilient to attempts at extermination or control.