Book Review: Blindsight

Monday, May 13, 2013 § 2

Blindsight is a hard sci-fi novel by Peter Watts [Amazon | Full free version on author’s website].  It takes place in a technologically advanced future, where humans have reached technological heights to the point that they are generating and discovering more data than their brains can handle.  The protagonist, Siri Keeton, is a cybernetically augmented human who specializes in interpreting incredibly high-order data and patterns, thereby acting as a middleman between humankind’s incomprehensible technology and humankind itself.

The book, published in 2006, is a very smart thought-experiment novel that incorporated the cutting edge of neurobiology, cognitive science, information theory, and artificial intelligence theories of its time.  It is also a book that also addresses many of our current anxieties: witness the explosion in the number of articles, blogs, and books that promise to “hack” our “brains” to help us perform optimally.  Blindsight, however, suggests a more frightening answer to our human anxieties than perhaps we would like.

While Blindsight is first and foremost an idea-driven novel, the author has wrapped the idea into a classic first-contact plot.  The story opens as follows: one day in 2082, Earth is visited by a group of projectiles that flash over the globe and then vanish, leaving little clue as to their origins or purpose. Siri and four other augmented humans are placed in a five-man team on a spaceship called Theseus, and sent out to investigate. The crew is a complex and fascinating mix of specialists, and are also manifestations of the author (a professional marine biologist).  Their specialities and augmentations allow Watts to explore a number of neurology, biology, and cognitive science theories such as manufactured multiple personalities, sensory augmentation, and even resurrected vampires. The overall plot is not extremely novel, but is a fitting vehicle for the kind of idea-driven writing the author is interested in.

What are the author’s main ideas?  To summarize, he is interested in distinguishing between intelligence and sentience, and exploring the evolutionary fitness, or lack thereof, imparted by those traits.  The first contact with the aliens—who are named Scramblers by the crew—is a harrowing one, and as incredible discoveries about the aliens come to light, we find ourselves asking along with the protagonists the big question of the book: what is consciousness really, truly good for, in this wide universe?  It’s a risky question to pose to an audience that is only able to process the story through consciousness.  The titular real-life condition of blindsight—in which a person with blindness resulting from brain damage to their visual processing center can nevertheless respond to visual stimuli, such as catching an object that they are unable to see—is used to help us imagine being an entity with intelligence but no sentience.

Uncoupling “intelligent responses” from “sentience” is an interesting move, because it’s a major question that the AI community deals with.  Any number of papers and books are devoted to asking the question of what comprises an intelligent machine.  Much of this work, including the Turing Test and the Chinese Room, are discussed in the book.  Blindsight, for its part, argues that we should use the only measure we have of intelligence, which is whether the entity is capable of learning, pattern recognition, and responding to additional data with improved performance.  This is exactly how the current machine learning field measures their algorithms when creating computer programs that are capable of learning.

Blindsight, however, goes a step further.  Many people believe that the be-all end-all of machine learning must necessarily end in sentience; a “thinking machine” is still viewed as the Holy Grail by many, though not all, AI researchers.  The hypothesis suggested by this book is that sentience may be the wrong goalpost.  Indeed, Blindsight asks its readers to consider that it is entirely possible, and perhaps evolutionarily favorable, for a living organism to become intelligent without developing sentience.  Such organisms would not so much think as calculate, sifting and gathering through patterns and then reacting.  This is how Scramblers live.  They have survived for ages by crunching the pure hard data of natural physics and mathematics, squeezing every bit of data out of their environment in order to improve their survival and fitness.  Compared against that, human data is infinitely messier, often seemingly purposeless, stuffed with things that entertain a sentient mind but could appear useless to a data processor.

What happens when these two “intelligences” meet? I won’t spoil you; I will only say that if this scenario intrigues you, then you must read this book.

The idea of an unthinking but intelligent entity—and even worse, an unthinking but superior entity—is uncomfortable to most.  The story’s strength lies in its unflinching willingness to push human readers into that uncomfortable place: “what if humans are an evolutionary dead end?” Cleverly, Watts minimizes the chances that a reader will angst over this outcome: the story is told with minimal emotions.  Siri, the narrator, has had a brain surgery during his childhood that removed much of his capacity for empathy or emotions—and that is the point of view through which he narrates the story, and thus, the point of view in which we readers perceive the story.  This surgery has made Siri, who operates his job and navigates through life on pure data, in many ways cognitively closer to the unconscious aliens than to conscious humans.

If you are able to make this leap into another way of thinking, you might end up—like me—rooting for these aliens made of pure data, as well as for Siri, who is almost a precursor of them.  Blindsight made me believe in its arguments, and that is the highest praise I can give the book.

Part Two - Thoughts

Generally I do not indulge in the exercise of “if I had been writing this book” for the very good reason that I did not write the book.  However, I took Blindsight to be part novel and part genuine scientific hypothesis, and as a scientist, it is the hypothesis that I wish to respond to.

Spoilers after the jump!
The author’s hypothesis, as I understood it, is that the Scramblers and Humans will come into conflict because humans cannot help putting out “intelligent but meaningless” data, which the Scramblers view as a viral attack on their data processing.  (And that the humans will lose, because consciousness imparts a fitness cost in this environment.)  The catch phrase is, “How do you say ‘We come in peace’ when the very words are an act of war?”

I have two responses to this.

One, while humans are not that great at communicating via pure data, our machines are.  In the world of the book, humans are quite a far ways along already: Siri is able to pull incredible amounts of data by the use of his own augmentations, after all, as are the book’s other post-humans.  Clearly, humans are able to communicate with Scramblers.  That potential might have been enough to capitalize on, had Siri been less fatalistic.  Had the crew been less skeletal, something (such as the ship Theseus, which is later revealed to be a super-powerful AI) could have been set up to collect Scrambler data and to learn how to respond to it in the most peaceable—or otherwise effective—manner.  So ‘we come in peace’ doesn’t work?  Fine, how about opening with ‘there’s a carbon source over there’ followed by ‘and we have a lot of nukes over here’.  Problem solved.

Two, I would counter that humans are creatures of intelligence, who do not put out data for no reason.  Even the silliest piece of gossip at least contains social data, when taken into context.  Siri’s argument is that there is “no fitness payoff” involved in human communication such as art criticism, but literacy and access to art are socioeconomic markers for better status, and with better status does come better human fitness, although I will admit the data could be very noisy.  In addition, Siri says that to the Scramblers, human text would be too self-referential, but Siri, who spends his life analyzing other people based on the tiniest bits of data that they generate, should know better: those little bits of data are the equivalent of the data that he uses to assess his own situation.  The data are not self-referential; they point towards a larger social structure that has a different topology of fitness.  It is possible the Scramblers have a problem with that, but that is not the argument that Siri is making.  Indeed, the phrase “We come in peace” is informative towards an organism’s future fitness; it suggests you will not get blown to bits as soon as you approach.  In other words, while Watts was able to construct his novel in a way that convinced me of his arguments, his narrator might just be selling humans, human communication, and humans’ technological abilities a little bit short.  But in the end, I should say that even though I disagree intensely, it does not surprise me in the least that Siri is making that particular argument.

Imagine you are Siri Keeton: alienated from humans, augmented to live in data, surgically stripped of the emotional ability to appreciate the (few, and ultimately meaningless in the grand scheme of evolution or the universe) benefits that sentience does have to offer—

You’d feel that way, too.

What's this?

You are currently reading Book Review: Blindsight at .:. Another Likely Story.

meta

§ 2 Response to “Book Review: Blindsight”

§ Leave a Reply