'Robot: Mere Machine to Transcendent Mind' (the title says it all)
No future study is complete without graphs of exponential curves, but in Hans Moravec’s ‘Robot: Mere Machine to Transcendent Mind’ the accelerating curve from mundane to profound is an emergent form identifiable in the structure of the book itself. The book opened on page one is not the same one closed on page 211; readers looking for the hard stuff they associate with Moravec might be confused or turned off if they lack the trust to wait patiently—or the impulsiveness to skip ahead.
Moravec uses the first two chapters (50 pages) to relate the history of today’s semi-autonomous robots—a tale he is uniquely positioned to tell largely in the first person. In direct, semi-technical prose, he talks in big blocky paragraphs about maps, and MIPS, and the general trade-offs designers make today when creating robots suited to particular tasks, whether they be high-speed driving, low-speed roving, or anything in between. Most of this information will be easily forgotten, but by the end of the discussion the reader has at least a chance of relating to the scale of the numbers tossed around in succeeding chapters.
Chapter three is a very different take, looking past the misleading veneer of robot limbs and sensory organs and into the mind of the machine. This is a chapter about the possibility of true artificial intelligence, and the nine most common objections to it – attacks Moravec parries with the serene precision of a judo instructor. I have no doubt that he has had many, many opportunities to hone these arguments in discussions with people in and out of academia.
The midpoint of the book details Moravec’s categorization of robot abilities by “generation”. As computing power climbs the range of applications to which new generations of robots can be applied increases accordingly. Physical dexterity grows as well, as do the frequency of opportunities for disaster—robots making mistakes that injure themselves or others. Robot minds may feel analogues to pain and pleasure as part of their designs, learning from traumatic mistakes and great successes. By the fourth generation, robots are roughly as smart as human beings.
Here, I part company with some of Moravec’s conclusions. His world of tomorrow, filled with human-equivalent robots shaped by superintelligent AI, is neither inexplicable nor unstable. Robots, as agents of corporations in a capitalist environment, seek only to profit within the bounds of laws created by and for humans; the machines evolve to become ever more proficient in this domain. Not all robots will play along, he argues, but cooperation is such a beneficial long-term strategy that an overwhelming majority of minds will block the ill intents of the few bad eggs, who in any case wouldn’t survive into the next round of corporate mergers and breakups. Moravec does not seem to share my concerns about runaway seed AI—a mind miles above all others and becoming more intelligent every second, unstoppably bending the universe to a will that may or may not include our own.
But, perhaps to any true humans remaining in such an era, the expansion of the new minds will indeed look like a fiery nova of cooperative capitalism blossoming out from the inner solar system at the speed of thought. Moravec does suggest that traditional human minds will be hopelessly uncompetitive pensioners in the new economy, confined to heavily metaphorical interpretations of activity increasingly beyond their ken.
The final quarter of ‘Robot’ is definitely not your father’s futurism. Moravec shows a solid understanding of cutting-edge physics, and their implications for the future of thought: compact customized matter; the pros and cons of existing on the surface of a neutron star; the “Bekenstein bound” on computation in a single atom. This last one suggests that a single human body’s worth of atoms “could contain the efficiently encoded biospheres of a thousand galaxies—or a quadrillion individuals each with a quadrillion times the capacity of a human mind.” The future is in simulation.
Or is simulation the very cause of the future? Moravec patiently discusses the physics behind time travel with several thought experiments for sending messages or beings back in time. He also explains the closely related curiosities of quantum computation, where even the non-paradoxes can seem, well, paradoxical. Our universe is already saturated with particles that seem to have no regard for time’s arrow.
‘Robot’s’ final ten pages seem almost as dense as the sum of all that came before. Cosmological, philosophical questions on the nature of reality and consciousness boil over as Moravec uses Everett’s “many worlds interpretation” of quantum wave functions to argue the uselessness of differentiating between simulation and reality. Perhaps, he speculates, the universe only appears to have order because our minds make it so, and that our minds will never ceases to be even after everyone else sees us as long dead. We may simply find ourselves in branches of reality where our continued existence is likeliest.
So ends ‘Robot’, at a place very different from the seemingly inconsequential dawn of robotics described in the beginning. Like the strange future we approach, Moravec’s ‘Robot’ is an accelerating trip into the unexpected.
Moravec uses the first two chapters (50 pages) to relate the history of today’s semi-autonomous robots—a tale he is uniquely positioned to tell largely in the first person. In direct, semi-technical prose, he talks in big blocky paragraphs about maps, and MIPS, and the general trade-offs designers make today when creating robots suited to particular tasks, whether they be high-speed driving, low-speed roving, or anything in between. Most of this information will be easily forgotten, but by the end of the discussion the reader has at least a chance of relating to the scale of the numbers tossed around in succeeding chapters.
Chapter three is a very different take, looking past the misleading veneer of robot limbs and sensory organs and into the mind of the machine. This is a chapter about the possibility of true artificial intelligence, and the nine most common objections to it – attacks Moravec parries with the serene precision of a judo instructor. I have no doubt that he has had many, many opportunities to hone these arguments in discussions with people in and out of academia.
The midpoint of the book details Moravec’s categorization of robot abilities by “generation”. As computing power climbs the range of applications to which new generations of robots can be applied increases accordingly. Physical dexterity grows as well, as do the frequency of opportunities for disaster—robots making mistakes that injure themselves or others. Robot minds may feel analogues to pain and pleasure as part of their designs, learning from traumatic mistakes and great successes. By the fourth generation, robots are roughly as smart as human beings.
Here, I part company with some of Moravec’s conclusions. His world of tomorrow, filled with human-equivalent robots shaped by superintelligent AI, is neither inexplicable nor unstable. Robots, as agents of corporations in a capitalist environment, seek only to profit within the bounds of laws created by and for humans; the machines evolve to become ever more proficient in this domain. Not all robots will play along, he argues, but cooperation is such a beneficial long-term strategy that an overwhelming majority of minds will block the ill intents of the few bad eggs, who in any case wouldn’t survive into the next round of corporate mergers and breakups. Moravec does not seem to share my concerns about runaway seed AI—a mind miles above all others and becoming more intelligent every second, unstoppably bending the universe to a will that may or may not include our own.
But, perhaps to any true humans remaining in such an era, the expansion of the new minds will indeed look like a fiery nova of cooperative capitalism blossoming out from the inner solar system at the speed of thought. Moravec does suggest that traditional human minds will be hopelessly uncompetitive pensioners in the new economy, confined to heavily metaphorical interpretations of activity increasingly beyond their ken.
The final quarter of ‘Robot’ is definitely not your father’s futurism. Moravec shows a solid understanding of cutting-edge physics, and their implications for the future of thought: compact customized matter; the pros and cons of existing on the surface of a neutron star; the “Bekenstein bound” on computation in a single atom. This last one suggests that a single human body’s worth of atoms “could contain the efficiently encoded biospheres of a thousand galaxies—or a quadrillion individuals each with a quadrillion times the capacity of a human mind.” The future is in simulation.
Or is simulation the very cause of the future? Moravec patiently discusses the physics behind time travel with several thought experiments for sending messages or beings back in time. He also explains the closely related curiosities of quantum computation, where even the non-paradoxes can seem, well, paradoxical. Our universe is already saturated with particles that seem to have no regard for time’s arrow.
‘Robot’s’ final ten pages seem almost as dense as the sum of all that came before. Cosmological, philosophical questions on the nature of reality and consciousness boil over as Moravec uses Everett’s “many worlds interpretation” of quantum wave functions to argue the uselessness of differentiating between simulation and reality. Perhaps, he speculates, the universe only appears to have order because our minds make it so, and that our minds will never ceases to be even after everyone else sees us as long dead. We may simply find ourselves in branches of reality where our continued existence is likeliest.
So ends ‘Robot’, at a place very different from the seemingly inconsequential dawn of robotics described in the beginning. Like the strange future we approach, Moravec’s ‘Robot’ is an accelerating trip into the unexpected.
posted by Mitch Howe at 4:42 PM
0 Comments:
Post a Comment
<< Home