book review | Ray Kurzweil’s read on latest AI insights in the book Flesh and Machines: How Robots Will Change Us
October 31, 2002 by Ray Kurzweil
This review was originally published in Wired, “Peer Review,” in October 2002.
As one of the world’s leading roboticists, Rodney Brooks (Director of the MIT Artificial Intelligence Laboratory and Chairman of the successful iRobot Corporation) is also the consummate teacher.
He has a penchant for clear explanation and in his latest book, Flesh and Machines, How Robots Will Change Us, Brooks lucidly explores a wide range of themes related to his life with robots.
These range from personal anecdotes (e.g., his first encounter with another legendary robot builder, Hans Moravec, who was then living in his Stanford laboratory and musing about exotic topics ranging from sky hooks to tree-like robots), historical vignettes (e.g., Marvin Minsky’s unsuccessful attempt to solve the computer “vision” problem in a single Summer in 1966), algorithmic insights (e.g., how his Genghis robot achieved “animal-like behavior” from a few dozen simple programs operating in parallel), philosophical musings (e.g., what is the true nature of consciousness, “apart from our own personal experience of what it is like to be us?”), and ethical dilemmas (e.g., when will we need to stop treating robots like slaves).
The book ranges far and wide, but maintains a unity around the author’s passion for creating what he calls “situated creatures,” which we can eventually regard as our teachers and companions.
One of Brooks’ favorite themes is to present human history as a series of sobering lessons on why we’re not special. Copernicus’ sixteenth century observations revealed that the Universe did not revolve around us after all. Darwin’s Origin of Species argued that we were not all that different from animals. They were, after all, our distant relatives.
In the past century, quantum mechanics and the Heisenberg uncertainty principle placed apparent limits on our ability to know with certainty “what was going on in the Universe.” With Crick and Watson’s discovery of DNA, we went a step beyond Darwin and discovered that 98% of our genetic blueprint is shared with chimpanzees, and that we have only about twice as many genes as a fruit fly.
And now, Brooks points out, our uniqueness is being challenged once again, this time by creations of our own making. I would point out that this does indeed suggest a way in which humankind is special. Other species have certainly found competition in their ecological niches. But we’re the only species that has been challenged by an evolutionary process that we initiated.
Brooks provides some cogent arguments against long-term critics of AI. He correctly points out that virtual reality pioneer Jaron Lanier’s claim that AI is “based on an intellectual mistake….does not follow his premise.”
Brooks points out how many of long-term AI critic Hubert Dreyfus’ arguments were based on short sighted and now discredited claims of what computers “could not do in principle.” Brooks points out how physicist Roger Penrose’s claim that human consciousness derives from mysterious quantum effects in small neuronal structures called tubules has no supporting evidence. I would also point out that even if Penrose’s claim were correct, there is nothing preventing us from applying the same quantum effects in our machines. Brooks presents an effective refutation of Searle’s now decades old tautological Chinese Room argument.
Brooks then presents his own thesis on what constitutes the “new stuff” that separates intelligent life from not quite intelligent non life. Brooks calls it “the juice,” and does not quite explain what it is, but does appropriately hint that it has to do with as yet undiscovered principles of biological systems. I believe we have learned a subset of these principles (e.g., self-organization, self-referencing, massive parallelism, holographic organization of memory, etc.), and that we will learn the rest of these principles from a grand project that is underway to reverse engineer (i.e., to learn the principles of operation of) human neural organization.
I am pleased to say that I am not left off of Brooks’ list of thinkers he takes issue with. He agrees with me that computers will match the computational power of the human brain within a couple of decades. He claims, however, that I provide no prescription for the software of intelligence. I don’t have the space here to describe my software thesis in full, but it has to do with the reverse engineering of the human brain that I allude to above. Brooks provides little discussion of this important source of knowledge that represents at least one important source for AI insights. Brooks says that my thesis “neglects the primary role played by the bath of neurotransmitters and hormones,” although I would point out that these are relatively low bandwidth phenomena and are more easily modeled than neurons. He says that my thesis “neglects the role of our body in placing constraints and providing noncomputational aspects to our existence.”
As a roboticist, Brooks should appreciate the feasibility of providing the “noncomputational” aspects of a physical presence. Yet another approach to providing bodies for our machines will be virtual bodies in virtual environments, which Brooks also does not discuss.