Top critical review
Place in your library after Ray Kurzweil and Nick Bostrom
19 July 2018
Adds to a wonderful discussion about risks of Ai in which the seminal work this far has been by Nick Bostrom. But some extrapolations are drawn out of thin air. Looks like Tegmark is an expert on math/physics, but has only a cursory knowledge of the electronics behind the working of computers. Maybe he has forgotten his computer architecture basics.
Tegmark writes on a topic that's not his area of expertise, he is a physicist and a reductionist. I'm not sure real Ai and consciousness should not be viewed as other emergent fields like biology or more specifically neuroscience. It doesn't lend itself to the sparseness of equations (especially when they have not been clearly dilineated) - this focus had me a bit disappointed with the book
When to read:
This is definitely not the first read for anyone interested in the field of Ai security. There is no alternative to Bostrom there. If you are in general interested in superintelligence you should start with Vernor Vinge 's original paper and then move to Ray Kurzweil. In both of these aspects this could be a supplementary read.
There are some sublime pieces of writing where Tegmark' s clear logic and incisive thought made me go wow. For example when he tries to derive subprinciples and goals from any possible ultimate goals (page 264) or in the section where he defines consciousness through our knowledge of decision making (page 312). These were fresh perspectives for me and were my main takeaways from the book.
i felt much more could be said about neuroscience of our brain (given that he chose to touch on the topic). If the topic of the book was AI security more could be said about why each principle was taken up. Possible there was too much emphasis (without enough mathematics) on going directly from physical substrates (quarks, electrons) to "sentronium" (sentient matter)
The prelude is a masterful imagination of the future and reads like a fast sci-fi piece. Loved it.
The book print quality (i got the hard copy) is excellent for the price.
As an Ai researcher my views may be colored by higher expectations.
PS: Irrespective of whether you buy the book, head over to futureoflife.org - it's a movement started by Tegmark that deserves a read through.