Kevin Escoda sets out to examine the nature of AI and to contrast it with human intelligence. He tells us that he makes three claims ('not one more'): thinking is a physical act, which is paid for in energy, and that price governs everything else; it is a claim of thermodynamics, not a metaphor. What we call artificial intelligence is a compression of the written memory of humanity, which makes it dependent, forever, on those who own that memory. And this faculty is becoming an infrastructure that is owned and rented out, and the decisive question in poses is therefore not technical but political.
It's useful that Escoda gives us this summary in his introduction as the book itself meanders all over the place en route to fulfilling this summary. If I'm honest, if you strip away the meandering I'm not sure there's much more than a long article here - though I'm sure the author would argue all the diversions make it more interesting.
The book's layout has occasional moments of showiness - sudden bursts of colour or different font size in the text. It also gets indents the wrong way round (indented on the first paragraph in a section and not the rest, where it should be not indented on the first paragraph and indented on the rest). I found the writing style over-declaratory and unnecessarily obscure. As an example: 'The machine is not a thought freed from matter. It is a thought that weighs infinitely more than ours, an ogre where nature made a hummingbird... If the hummingbird proves you can think with 20 watts, then the ogre is not a fate, it is a stage.'
Because the whole drive of the book is from a starting point of energy - a perfectly sensible approach from a physics viewpoint - there is a dip into entropy and the second law of thermodynamics, where we get a wildly overstated comment (there are more elsewhere): 'this law is the most important and the least known in all of science'. You can obviously argue whether or not it's most important, though it's definitely up there in the top ten. But least known? I suspect it's one of the better known ones - I can list plenty more laws that are far more obscure. How about Kirchoff's third law of spectroscopy or Brewster's law to name a couple at random?
I do like the intent that Escoda has in writing this book, but I'm not sure it's done in the most approachable fashion.
N.B. I think the full title of the book is From the Machine that Calculates to The Machine that Thinks, but I truncated it here to the part emphasised on the cover.
Review by Brian Clegg - See all our online articles or subscribe to a weekly email free here



Comments
Post a Comment