The Age of EM: Work, Love, and Life When Robots Rule the Earth
by Robin Hanson (a professor of economics at George Mason
University) is a new book that speculates in the greatest detail about
what kind of world will arise if a weird possibility occurs. The
weird possibility that Hanson imagines is that humans become "not the main inhabitants" of the earth, being overshadowed by what he calls “ems.”
Hanson defines an “em” like this: “An em results from
taking a particular human brain, scanning it to record its particular
cell features and connections, and then building a computer model
that processes signals according to those same features and
connections.”
What Hanson is
imagining is essentially android robots, but not robots who are
controlled by artificial intelligence programming. Hanson is rather
skeptical about artificial intelligence programming, and says on page
382 it will take two to four centuries before it is capable of
achieving “human level” capabilities. But Hanson thinks there's
a shortcut. He thinks we can get robots as smart as people by
scanning human brains and then loading that information into robots.
He thinks this will be feasible within about a hundred years. Hanson
thinks that before our planet sees any robots smarter than man, there
will be one or more centuries in which our planet is largely
inhabited by robots about as smart as humans – robots whose
electronic minds are duplicates of human brains.
Hanson gives us a
cold vision that is morally repulsive. After discussing a destructive
form of brain scanning that is rather like a turkey slicer (page
148), he suggests using eugenics to produce a superior stock of
humans to have their brains copied into robots.
The rest of the
world has pretty much rejected eugenics since the fall of the Nazis,
but apparently not Hanson. On page 162 he says:
It is possible
that sometime in the next half-century or so we we will be able to
create babies of substantially higher genetic quality by embryo
selection in the context of in vitro fertilization. If so, when they
reach the right age these may be very attractive candidates for
scanning into ems.
Does the professor
thinks we should wait until the genetically engineered super-kids are
grown, and then feed their brains into his brain scanner that is like
a turkey slicer?
On page 146 the
professor proposes that the millions of robot “ems” that are
copies of human minds should be given only “wages near subsistence levels,” even though he assures us on page 150 that these “ems”
are “mentally quite human.” He offers this cold justification for
this cruel proposal.
Readers of this
book may find near-subsistence wages to be a strange and perhaps
scary prospect. So it is worth remembering that such wages in effect
applied to almost all animals who ever lived, to almost all humans
before a few hundred years ago, and for a billion humans still today.
And what about all
the humans who would have all their jobs lost by this great army of
“ems” that would appear? Hanson seems to have little interest in
the fate of the old-fashioned creatures known as humans. I looked up
“humans” in the index of his 384-page book, and I see the last reference to
“humans” is on page 14. Hanson does not imagine that humans will become extinct, but he imagines on page 8 that humans will live in retirement cities, with the "em" robots living in separate cities (how is he is able to make so specific a claim I don't know).
Hanson imagines that
the human race will be economically eclipsed by robots that have
copies of human minds, but we may ask: what would be the point of
that? Hanson tells us that it would occur because such robots will be
more efficient, and work much faster. But it is very far from clear
that this would be true. Give me a year 2100 human armed with all
the best year 2100 gadgets (and some pills helping him think faster),
and such a person should do just as well as a robot with a brain
copied from a human.
Hanson creates
endless speculations based on this idea of robot minds being copies
of human minds, but he does nothing to show the feasibility of the
idea. So Hanson is like some person building a 10-story tower before
laying a concrete foundation for such a tower. There are actually
very good reasons for thinking that it is impossible to ever create a
robot mind that is a copy of a human mind. One such reason is that
while Hanson claims that the mind is just the brain, there are very
good reasons for thinking they are not the same, and that you would
not at all capture someone's mind by scanning his brain.
Let me specify some
preliminary requirements that we should demand of any book centered
upon speculations that robot minds can be created from scanning human
minds. After these preliminary requirements are met, we can think to
ourselves that we are not wasting our time by reading speculations
about some possibility that will never be able to occur.
The first
requirement we should demand is a detailed explanation of how it is
that the human brain could possibly store memories stretching back
50 years. For the reasons I give in this post, scientists don't
understand how the human brain can store memories for longer than
three weeks. The most common explanation for memory is that humans
store memories in synapses. But synapses are subject to molecular
turnover which should prevent them from storing any memories for
longer than a few weeks. So how is that humans can store memories for
50 years?
It may be too much
to demand that a brain emulation theorist provide the answer
to how humans store memories for 50 years, but we should demand that
he can at least provide some possible answer – at least
some speculative scenario which, if it were true, would explain how
humans can store memories for 50 years. If no such scenario can be
provided, we should not believe that you will ever be able to have
robot minds that are copies of human minds.
The second
requirement we should demand from our brain emulation theorist is an
explanation of how such a mechanism for remembering 50-year-old
memories could have naturally evolved (seeing that such a “humans
will be replaced” thinker is presumably not someone who thinks some
higher power had a role in human origins). This is a huge challenge,
for while such a theorist may be tempted to imagine some very
complicated scheme by which memories might be stored in the brain for
50 years, such a scheme will not be a plausible natural explanation
unless it is something that might have naturally appeared through
evolution. Here there is the huge difficulty that we cannot think of
any reason why natural selection would give us memories lasting
longer than a year or two. Primates could survive just as well if
they only remembered memories going back a year or two.
The third
requirement we should demand from our brain emulation theorist is
some explanation that dismisses all of the evidence for human psychic
abilities such as ESP and human survival after death. If such
evidence is valid, then “your mind is your brain” assumptions are
dead wrong. We have no way of explaining something like ESP as a
function of the human brain, although it could be a sign of something
like a human soul. If experiences such as near-death experiences are
real, then we presumably have some kind of soul that survives death.
But the brain emulation theorist cannot admit such possibilities.
Since such a
theorist wishes to maintain that your mind is stored entirely in your
brain (and not at all in any type of soul separate from the brain),
such a theorist must somehow clear from the table all of the great
deal of evidence that has been accumulated suggesting that there is a
soul as well as paranormal psychic abilities. So our brain emulation
theorist would have to somehow accomplish the long, difficult task of
explaining why very many things that seem to be good evidence are not
really evidence. I presume that would require at least a long chapter
in a book.
As if all this is
not hard enough, there is still a fourth requirement that our brain
emulation theorist must accomplish before we should take seriously
the claim that human minds could be copied to robot minds. The
theorist must give a plausible explanation for how it could be that
all the information in a human mind might be copied to a robot mind
without destroying most or much of the information in the human mind.
There are good
reasons for believing that even if your mind was entirely stored in
your brain, it would be physically impossible to read the state of a
mind in a way that would allow some robot mind to copy the contents
of that mind. While things such as fMRI scanners can make large-scale
maps of the brain, there is no type of external scanner that would
allow us to know the exact molecular and electrical state of
microscopic neurons. While we can imagine some tiny molecule-sized
probe that could be injected into a brain and then read the state of
one neuron, such a probe could never store more than a tiny bit of
information. It would not work to simply inject millions of
microscopic probes into someone's brain, for such probes would have
no way of figuring out the exact positional coordinates of the
neurons they were analyzing.
You might think that
it is possible to do a destructive scan of the brain to determine its
exact contents. We can imagine some machine that is something like a
turkey slicer. First it would slice off one little slice of the
brain, and then analyze that like a copy machine reading a document.
Then the next little slice would be done, and so forth. But that
would not work, for the very act of making these thin slices would
disrupt any delicate information storage in the brain that was
preserving memories.
To present a
plausible scenario for robots emulating human minds, our brain
emulation theorist would need to explain how this brain scanning
would work, as well as do the three other items I mentioned. Hanson
doesn't do anything like meeting the requirements I have mentioned.
Hanson does nothing
to substantiate his claim that the mind is just the brain. Hanson's
failure in considering the difficulties here are shown by page 148 of
his book, where he says this about brain scans that would give human
minds to robots: “The very first scans might perhaps be performed
on cryonics customers, that is, people who had previously had their
brains frozen and stored in liquid nitrogen, in the hope of being
revived when technology improved, and who had agreed to allow em
scanning later.” Determining the exact state of a brain that had
just died would be a nightmare enough. Add brain freezing (with all
of its resulting cell damage) to the mix, and the problem becomes 100
times worse. Hanson informs us that he is one of only 2000 living people
who have arranged to be cryogenically frozen upon death.
The cryogenics
fantasy (represented by the slogan “freeze, wait, reanimate”) was
implausible enough. Apparently Hanson believes in an even wackier
notion – freeze your brain, stay in cold storage for a few decades,
then wait until someone copies your brain state into a robot. What convenience – after that fatal car crash, you wake up 40 years
later with a nice steel robot body!
There are countless
feasibility problems in the idea of copying human minds into
robots, problems that Hanson has not paid attention to. The central
assumption behind the idea is wrong. Your mind is not your brain, but
something more than your brain. This is indicated both by
well-substantiated psychic phenomena (such as ESP and near-death
experiences), and also by the spectacular inability of neuroscience
to explain basic facets of our mind such as consciousness, instant
memory retrieval of obscure memories, and memories lasting 50 years
despite very rapid molecular turnover in the brain (see here and here for why it is hard to imagine the last two will ever be explained neurologically). So Hanson's very
elaborate speculations about what would happen if the world became
full of “em” robots (with minds copied from human brains) are no
more relevant than some speculations of what would happen if unicorns
were to take charge of the planet.
No comments:
Post a Comment