[...] it is entirely debatable whether framing the issue of machine intelligence in the mirror of the human will allow us to understand what the real problem is. That machines can replace humans tells us nothing special about intelligence, particularly if this is as part of an economy that, in its entropic repetition of the eternally self-same, generally produces stupidity rather than intelligence. [...]

the idea of an alien, machinic intelligence not only rebarbative but contradictory

[...] One trains a neural network on specific, finite datasets. The ability to pattern-match more generally presupposes the existence of redundancy in the data and thus self-similarity. So, one could argue that the ability to discern redundancy in data is the ability to learn about how things imitate or repeat themselves (like the data-mining software that tells us which books we want to buy).