The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World
Overambitious pop science from a lively and charming narrator. He tries to sketch all of machine learning in a couple hundred pages. The warmth of his teaching voice comes through the page:
but he needs a better editor, more even than Nassim Taleb does. This is often just a stream-of-consciousness analogy-dump, and with precise topics that just doesn't fly. (Both Penguin productions.)
There's more wrong with it than prose, unfortunately: he gives equal time to unpromising approaches (genetic programming, analogical reasoning) and so has to skim over the single most important approach (deep learning), with no real sense of the giant differences in success. Couple this with his terrible argument against AI risk ("unlike humans, computers don’t have a will of their own. They’re products of engineering, not evolution... the evaluation function is determined by us. A more powerful computer will just optimize it better... The same reasoning applies to all AI systems because they all—explicitly or implicitly—have the same three components. They can vary what they do, even come up with surprising plans, but only in service of the goals we set them.") and it becomes actively unhelpful.
(Pedantic aside: he commits linguistic violence every time he uses "algorithm" instead of the unsexy true referent, "program". He obviously knows the distinction much better than I do, but skips this to talk down / excitingly to the audience.)
Read his great dense paper instead.
DNF 50%
As you read the book, feel free to skim or skip any parts you find troublesome; it’s the big picture that matters, and you’ll probably get more out of those parts if you revisit them after the puzzle is assembled.
but he needs a better editor, more even than Nassim Taleb does. This is often just a stream-of-consciousness analogy-dump, and with precise topics that just doesn't fly. (Both Penguin productions.)
There's more wrong with it than prose, unfortunately: he gives equal time to unpromising approaches (genetic programming, analogical reasoning) and so has to skim over the single most important approach (deep learning), with no real sense of the giant differences in success. Couple this with his terrible argument against AI risk ("unlike humans, computers don’t have a will of their own. They’re products of engineering, not evolution... the evaluation function is determined by us. A more powerful computer will just optimize it better... The same reasoning applies to all AI systems because they all—explicitly or implicitly—have the same three components. They can vary what they do, even come up with surprising plans, but only in service of the goals we set them.") and it becomes actively unhelpful.
(Pedantic aside: he commits linguistic violence every time he uses "algorithm" instead of the unsexy true referent, "program". He obviously knows the distinction much better than I do, but skips this to talk down / excitingly to the audience.)
Read his great dense paper instead.
DNF 50%