We live in the so-called dataquake:
‘A few thousands of years ago, you needed to be a god or goddess if you wanted to be painted, be sculpted, or have your story remembered and told. A thousands years ago you needed to be a king or queen, and a few centuries ago you needed to be a rich merchant, or in the household of one. Now anybody, even a soup can, can be painted.”
China rightly prides itself of being at the forefront of AI research. Worldwide.
As a visit in one random Chinese supermarket can remind the casual observer,
they do have the biggest demographic indeed, and they cannot fail to harness that wealth of human data generating processes to train ever more advanced networks. But what if video games can be used to train networks? What if, as AlphaGo Zero has recently accomplished, there is less and less scope for human agency even at the stage of training such a network? Human data generating processes can at best inhabit a vast region of collinearity, a small spanning set of independent components. See also here.