Jeff Hawkins' On Intelligence lays out a fascinating theory of how the brain works. He presents a new synthesis of elements of congnition that he argues have been under-estimated to date: time-dependence, hierarchy, memory and prediction.
The main drawback is that this is very much a first-person narrative; on page 87, for example, the word "I" appears no fewer than ten times. An exception, perhaps? Let's try page 169: "I", "me" and "my" appear eight times.
The ideas are laid out very well, and the reader is eased into neuroscience with clever analogies and metaphors. There are thought-provoking passages on why AI and neural nets failed, on the likely nature of intelligent machines, and on the uses to which they could be put. He led me to stimulating speculations -- what greater compliment for a thought-provoking book? -- like possible explanations for the diversity of personality traits among individuals all endowed with essentially the same cognitive apparatus.
The prose is very readable, thanks largely I assume to the efforts of the co-author, science writer Sandra Blakeslee. The price paid for the collaboration, though, is that the writing doesn't have much character. In contrast, Malcolm Gladwell writes just as well, but one feels that you end up knowing him a little and would like to spend more time with him discussing his ideas.
As a scientist manqué I found Hawkins' lack of rigor faintly annoying. While he makes a good case, the phrases "it's true that" and "I firmly believe that" seem to be equivalent to him. This is a popular book written before the papers are published, rather than after; as far as I could tell (eg from the publications listed at the Redwood Neuroscience Institute, f(o)unded by Hawkins) he has no peer-reviewed publications. This book seems to be more interested in establishing a claim in the public mind than advancing science, though its popularity among the digerati is likely to spur the professional scientists on, in spite of themselves.
No comments:
Post a Comment