Everything
felt possible at Transhuman Visions 2014, a conference in February
billed as a forum for visionaries to "describe our fast-approaching,
brilliant, and bizarre future." Inside an old waterfront military depot
in San Francisco's Fort Mason Center, young entrepreneurs hawked
experimental smart drugs and coffee made with a special kind of butter
they said provided cognitive enhancements. A woman offered online
therapy sessions, and a middle-aged conventioneer wore an electrode
array that displayed his brain waves on a monitor as multicolor
patterns.On stage, a speaker with a shaved head and a thick,
black beard held forth on DIY sensory augmentation. A group called
Science for the Masses, he said, was developing a pill that would soon
allow humans to perceive the near-infrared spectrum. He personally had
implanted tiny magnets into his outer ears so that he could listen to
music converted into vibrations by a magnetic coil attached to his
phone.
None of this seemed particularly ambitious, however,
compared with the claim soon to follow. In the back of the audience,
carefully reviewing his notes, sat Randal Koene, a bespectacled
neuroscientist wearing black cargo pants, a black T-shirt showing a
brain on a laptop screen, and a pair of black, shiny boots. Koene had
come to explain to the assembled crowd how to live forever. ''As a
species, we really only inhabit a small sliver of time and space,''Koene
said when he took the stage. ''We want a species that can be effective
and influential and creative in a much larger sphere.''
Koene's
solution was straightforward: He planned to upload his brain to a
computer. By mapping the brain, reducing its activity to computations,
and reproducing those computations in code, Koene argued, humans could
live indefinitely, emulated by silicon. ''When I say emulation, you
should think of it, for example, in the same sense as emulating a
Macintosh on a PC,'' he said. ''It's kind of like platform-independent
code.''
Koene's solution was straightforward: He planned to upload
his brain to a computer. By mapping the brain, reducing its activity to
computations, and reproducing those computations in code, Koene argued,
humans could live indefinitely, emulated by silicon. "When I say
emulation, you should think of it, for example, in the same sense as
emulating a Macintosh on a PC," he said. "It's kind of like
platform-independent code."
The audience sat silent, possibly
awed, possibly confused, as Koene led them through a complex tour of
recent advances in neuroscience supplemented with charts and graphs.
Koene has always had a complicated relationship with transhumanists, who
likewise believe in elevating humanity to another plane. A Dutch-born
neuroscientist and neuro-engineer, he has spent decades collecting the
credentials necessary to bring his fringe ideas in line with mainstream
science. Now, that science is coming to him. Researchers around the
globe have made deciphering the brain a central objective. In 2013, both
the U.S. and the EU announced initiatives that promise to accelerate
brain science in much the same way that the Human Genome Project
advanced genomics. The minutiae may have been lost on the crowd, but as
Koene departed the stage, the significance of what they just witnessed
was not: The knowledge necessary to achieve what Koene calls "substrate
independent minds" seems tantalizingly within reach.
The
concept of brain emulation has a long, colorful history in science
fiction, but it’s also deeply rooted in computer science. An entire
subfield known as neural networking is based on the physical
architecture and biological rules that underpin neuroscience.
Roughly
85 billion individual neurons make up the human brain, each one
connected to as many as 10,000 others via branches called axons and
dendrites. Every time a neuron fires, an electrochemical signal jumps
from the axon of one neuron to the dendrite of another, across a synapse
between them. It’s the sum of those signals that encode information and
enable the brain to process input, form associations, and execute
commands. Many neuroscientists believe the essence of who we are—our
memories, emotions, personalities, predilections, even our
consciousness—lies in those patterns.
In the 1940s,
neurophysiologist Warren McCulloch and mathematician Walter Pitts
suggested a simple way to describe brain activity using math. Regardless
of everything happening around it, they noted, a neuron can be in only
one of two possible states: active or at rest. Early computer scientists
quickly grasped that if they wanted to program a brainlike machine,
they could use the basic logic systems of their prototypes—the binary
electric switches symbolized by 1s and 0s—to represent the on/off state
of individual neurons.
A few years later, Canadian psychologist Donald Hebb suggested that
memories are nothing more than associations encoded in a network. In the
brain, those associations are formed by neurons firing simultaneously
or in sequence. For example, if a person sees a face and hears a name at
the same time, neurons in both the visual and auditory areas of the
brain will fire, causing them to connect. The next time that person sees
the face, the neurons encoding the name will also fire, prompting the
person to recollect it.
Using these insights, computer engineers
have created artificial neural networks capable of forming associations,
or learning. Programmers instruct the networks to remember which pieces
of data have been linked in the past, and then to predict the
likelihood that those two pieces will be linked in the future. Today,
such software can perform a variety of complex pattern-recognition
tasks, such as detecting credit card purchases that diverge dramatically
from a consumer’s past behavior, indicating possible fraud.
Of
course, any neuroscientist will tell you that artificial neural networks
don’t begin to incorporate the true complexity of the human brain.
Researchers have yet to characterize the many ways neurons interact and
have yet to grasp how different chemical pathways affect the likelihood
that they will fire. There may be rules they don't yet know exist.
But
such networks remain perhaps the strongest illustration of an
assumption crucial to the hopes and dreams of Randal Koene: that our
identity is nothing more than the behavior of individual neurons and the
relationships between them. And that most of the activities of the
brain, if technology were capable of recording and analyzing them, can
theoretically be reduced to computations.
On
a warm afternoon in late January, I follow Koene up the stairs of the
second-floor walkup he shares with his girlfriend on the edge of San
Francisco’s Portrero Hill. He leads me through a small living room
crammed full of synthesizers and Legos and into a bedroom, where a
standing desk represents his home office. It holds oversize computer
screens and laptops arrayed like the electronics of a star-ship command
center. It’s a modest setting, but Koene is only in the third decade of
his quest—a mere blink of an eye when you consider that his goal is
immortality.
Koene, the son of a particle physicist, first
discovered mind uploading at age 13 when he read the 1956 Arthur C.
Clarke classic The City and the Stars. Clarke’s book describes a city
one billion years in the future. Its residents live multiple lives and
spend the time between them stored in the memory banks of a central
computer capable of generating new bodies. “I began to think about our
limits,” Koene says. “Ultimately, it is our biology, our brain, that is
mortal. But Clarke talks about a future in which people can be
constructed and deconstructed, in which people are information.”
It
was a vision, Koene decided, worth devoting his life to pursuing. He
began by studying physics in college, believing the route to his goal
lay in finding ways to reconstitute patterns of individual atoms. By the
time he graduated, however, he concluded that all he really needed was a
digital brain. So he enrolled in a masters program at Delft University
of Technology in the Netherlands, where he focused on neural networks
and artificial intelligence.
It was while at Delft in 1994 that
Koene made an important discovery: a community of people who shared his
ambition. Exploring the new medium of the Internet, he stumbled upon the
“Mind Uploading Home Page,” owned by Joe Strout, an Ohio-born computer
buff, aspiring neuroscientist, and self-described immortalist. Strout
facilitated a discussion group that Koene quickly joined, and its
members began to debate whether extracting information from the brain
was technologically feasible, and if it was, what they should call it:
downloading, uploading, or mind transfer. They eventually settled on
“whole brain emulation.” And then they outlined career goals that would
help them advance their cause.
Koene chose to pursue a Ph.D. in
computational neuro-science at McGill University, and later landed at a
Boston University neurophysiology lab, where he attempted to replicate
mouse brain activity on a computer. Strout pursued an advanced degree in
neuroscience, then moved on to the lab of a computational
neurobiologist at the Salk Institute. “We were all trying to push
research problems in whatever way we could,” Strout says. “The trouble
was that for the elder neuroscience researchers, this wasn’t a topic
they could discuss publicly. They would talk about it over a beer. But
it was too fringe for people who were trying to get grants for
research.”
By
then, many of the other group members had earned their credentials. And
in 2007, computational neuroscientist Anders Sandberg, who studies the
bioethics of human enhancement at Oxford University, summoned interested
experts to Oxford’s Future of Humanity Institute for a two-day
workshop. Participants laid out a roadmap of capabilities humans would
need to develop in order to successfully emulate a brain: mapping the
structure, learning how that structure matches function, and developing
the software and hardware to run it.
Not long afterward, Koene
left Boston University to become the director of neuroengineering at the
Fatronik-Tecnalia Institute in Spain, one of the largest private
research organizations in Europe. “I didn’t like the job once I figured
out they weren’t into taking any risks and didn’t really care about
futuristic things related to whole brain emulation,” Koene says. So, in
2010, he moved to Silicon Valley to take a job as head of analysis at
Halcyon Molecular, a nanotechnology company that had raised more than
$20 million from PayPal cofounders Peter Thiel and Elon Musk, among
others. Though Halcyon’s goal was to develop low-cost, DNA-sequencing
tools, its leaders assured Koene he would have time to work on brain
emulation, a goal they supported.
By the time Halcyon abruptly
went out of business in 2012, Koene had created Carboncopies.org, which
serves as a hub for mind-uploading advocates. He had also made a lot of
contacts. Within months, he secured financial backing from Dimitry
Itskov, a Russian dot-com mogul who hoped to upload himself to a
“sophisticated artificial carrier” and considered whole brain emulation
an essential step.
“We need to provide a foundation so the new
field of brain emulation is taken seriously,” Koene tells me from his
bedroom command center. He opens a color-coded chart on one of the
screens. It consists of overlapping circles filled with names and
affiliations, divided into wedges representing the roadmap’s objectives.
Koene points to the outermost circle. “These are the people who just
have compatible R&D goals,” he says. Then he indicates the smaller,
inner circle. “And these are the people who are onboard.”
It’s
all of these individuals, mainstream neuroscientists, who will advance
whole brain emulation, Koene says—not trans-humanists, who he observes
“lack rigor.” And they’ll do so even if philosophically their goals are
quite different.
Today,
as it happens, every pillar of the brain-uploading roadmap is a highly
active area in neuroscience, for an entirely unrelated reason:
Understanding the structure and function of the brain could help doctors
treat some of our most debilitating diseases.
At Harvard
University, neurobiologist Jeff Lichtman leads the effort to create a
connectome, or comprehensive map of the brain’s structure: the network
of trillions of axons, dendrites, and synapses that convey
electro-chemical signals. Lichtman is working to understand how
experiences are physically encoded at the most basic level in the brain.
To do so, he uses a device that incorporates innovations made by a
brain-uploading proponent, Kenneth Hayworth, who spent time as a postdoc
in Lichtman’s lab. It slices off razor-thin pieces of mouse brain and
collects them sequentially on a reel of tape. The slices can then be
scanned with an electron microscope and viewed on a computer like the
frames of a movie.
By following the threadlike extensions of
individual nerve cells from frame to frame, Lichtman and his team have
gained some interesting insights. “We noticed, for instance, that when
an axon bumped into a dendrite and made a synapse, if we followed it
along, it made another synapse on the same dendrite,” he says. “Even
though there were 80 or 90 other dendrites in there, it seemed to be
making a choice. Who expected that? Nobody. It means this thing is not
some random mess.”
When he started five years ago, Lichtman says,
the technique was so slow it would have taken several centuries to
generate images for a cubic millimeter of brain—about one thousandth the
size of a mouse brain and a millionth the size of a human one. Now
Lichtman can do a cubic millimeter every couple of years. This summer, a
new microscope will reduce the timeline to a couple of weeks. An army
of such machines, he says, could put an entire human brain within reach.
At
the same time, scientists elsewhere are aggressively mapping neural
function. Last April, President Obama unveiled the BRAIN Initiative (for
Brain Research through Advancing Innovative Neurotechnologies) with an
initial $100 million investment that many hope will grow to rival the
$3.8 billion poured into decoding the human genome.
Columbia
University neuroscientist Rafael Yuste proposed a large-scale brain
activity map that helped inspire the BRAIN Initiative, and he has spent
two decades developing tools aimed at tracking how neurons excite and
inhibit one another. Yuste likens the brain’s connectome to roads and
the firing of its neurons to traffic.
Studying how neurons fire in
circuits and how those circuits interact, he says, could help demystify
diseases such as schizophrenia and autism. It could also reveal far
more. Our very identity, Yuste suspects, lies in the traffic of brain
activity. “Our identity is no more than that,” he says. “There is no
magic inside our skull. It’s just neurons firing.”
To study those
electrical impulses, scientists need to record the activity of
individual neurons, but they’re limited by the micromachining techniques
used to produce today’s technology. In his lab at MIT, neuro-engineer
Ed Boyden is developing electrode arrays a hundred times denser than the
ones currently in use. At the University of California, Berkeley,
meanwhile, a team of scientists has proposed nanoscale particles called
neural dust, which they plan to someday embed in the cortex as a
wireless brain-machine interface.
Whatever discoveries these
researchers make may end up as fodder for another ambitious government
initiative: the European Union’s Human Brain Project. Backed by 1.2
billion euros and 130 research institutions, it aims to create a
super-computer simulation that incorporates everything currently known
about how the human brain works.
Koene
is thrilled with all of these developments. But he’s most excited about
a brain-simulation technology already being tested in animals. In 2011,
a team from the University of Southern California (USC) and Wake Forest
University succeeded in creating the world’s first artificial neural
implant—a device capable of producing electrical activity that causes a
rat to react as if the signal came from the animal’s own brain. “We’ve
been able to uncover the neural code—the actual spatio-temporal firing
patterns—for particular objects in the hippocampus,” says Theodore
Berger, the USC biomedical engineer who led the effort. “It’s a major
breakthrough.”
Scientists believe long-term memory involves
neurons in two areas of the hippocampus that convert electrical signals
to entirely new sequences, which are then transmitted to other parts of
the brain. Berger’s team recorded the incoming and outgoing signals in
rats trained to perform a memory task, and then programmed a computer
chip to emulate the latter on cue. When they destroyed one of the layers
of the rat’s hippocampus, the animals couldn’t perform the task. After
being outfitted with the neural implant, they could.
Berger and
his team have since replicated the activity of other groups of neurons
in the hippocampus and prefrontal cortex of primates. The next step, he
says, will be to repeat the experiment with more complex memories and
behaviors. To that end, the researchers have begun to adapt the implant
for testing in human epilepsy patients who have had surgery to remove
areas of the hippocampus involved in seizures.
“Ted Berger’s
experiment shows in principle you can take an unknown circuit, analyze
it, and make something that can replace what it does,” Koene says. “The
entire brain is nothing more than just many, many different individual
circuits.”
That
afternoon, Koene and I drive to an office park in Petaluma about 30
miles outside of San Francisco. We head into a dimly lit, stucco
building decorated with posters that superimpose words like “focus” and
“imagination” over photographs of Alpine peaks and tropical sunsets.
Guy
Paillet, a snowy-haired former IBM engineer with a thick French accent
and a cheerful Santa Claus–like disposition, soon joins us in a
conference room. Paillet and his partner had invented a new kind of
energy-efficient computer chip based on the physical architecture of the
brain—an achievement that had earned them inclusion in Koene’s chart.
Koene wanted an update on their progress.
Paillet reports that he
is negotiating to take over an economically troubled computer
chip–fabrication foundry in the South of France. Would Koene be willing
to serve as a scientific advisor and possibly a fund-raiser on a related
project, he asks? Koene shifts impatiently in his chair. “I just had an
idea,” he announces. “You are thinking of getting into the foundry
business. At the same time people at UC Berkeley are thinking of
building new types of neural interfaces. When they get their prototype
to work, would you consider . . . .”
“That’s a very good idea!” Paillet interrupts, before Koene can even finish asking whether he might fabricate their device too.
As
we pull out of the parking lot, Koene is ebullient. I had just
witnessed his job at its best. “This is what I do,” he says. “You have
got tons of labs and researchers who are motivated by their own personal
interests.” The trick, he says, is to identify the goals that could
benefit brain uploading and try to push them forward—whether the
researchers have asked for the help or not.
Certainly, it seems,
many scientists have proven willing to consult and even collaborate with
Koene. That was clear last spring, when scientists from institutions as
varied as MIT, Harvard University, Duke University, and the University
of Southern California descended on New York City’s Lincoln Center to
speak at a two-day congress that Koene organized with the Russian mogul
Itskov. Called Global Future 2045, the conference’s objective was to
explore the requirements and implications of transferring minds into
virtual bodies by the year 2045.
Some of those present, however,
later distanced themselves from the event’s stated “spiritual and
sci-tech” vision. “We were trying to get people with a lot of funding
who can do big things to start investing in important questions,” says
Jose Carmena, one of the Berkeley neuroscientists working on neural
dust. “That doesn’t mean we have the same goal. We have similar goals
along the way, like recording from as many neurons as possible. We all
want to understand the brain. It just happens that they need to
understand the brain so they can upload it to a computer.”
Carmena’s
reticence was shared by other researchers, some of whom grew alarmed at
even a faint possibility that their opinions about the technical
plausibility of brain uploading—however qualified and cautious—might
somehow be misinterpreted as an endorsement. “There is a big difference
between understanding and building a brain,” Yuste says. “There are many
things that we more or less understand but we cannot build.” For
example, the brain’s hardware could prove critical, he explained, “or
there could be intrinsic stochastic events, like in quantum physics,
that could make it impossible to replicate."
Harvard’s Lichtman
was more comfortable speculating on the concept. “I am not sure any new
laws of physics have to be invented as they go forward,” he says. “It’s
not completely impossible, like the idea of putting a cow head on a dog.
It’s a science-fiction idea, but making a brain of silicon does not
seem crazy to me.” In fact, he thinks the movement has helped advance
neuroscience and hopes people like his former postdoc Hayworth
succeed—not so they can live forever but to accelerate cures for brain
dysfunction.
Hayworth, for his part, is now a senior scientist at
Howard Hughes Medical Institute’s Janelia Farm Research Campus, a leader
in connectomics, where he is developing techniques to precisely image
much larger sections of brain than currently possible. He also founded
the Brain Preservation Foundation, which has offered a prize for
inventing a method that can preserve the brain until emulation
technology catches up. “I know this is a controversial topic,” he says,
“and there aren’t a heck of a lot of scientific institutes of any type
that relish being dragged into it. Hopefully at some point that will
change.”
In the meantime, many scientists seem to puzzle over a
question more fundamental to the brain uploaders’ goal: What’s the
point? Existing indefinitely in the confines of computer code, Lichtman
points out, would be a pretty boring life.
Earlier in the day, I
had asked Todd Huffman, a member of Strout’s early discussion group,
whether the quest really boiled down to achieving immortality. Koene and
I had dropped by Huffman’s company, which received venture capital to
develop automated brain-slicing and imaging technologies. Huffman was
wearing pink toe nail polish on his shoeless feet and sported a thick
beard and bleached faux-hawk.
“That’s a very egocentric and
individualist way of characterizing it,” he responded. “It’s so that we
can look at the thought structures of humans who are alive today, so
that we can understand human history and what it is to be human. If we
can capture and work with human creativity, drive, and awareness the
same way that we work with, you know, pieces of matter,” he said, “we
can take what it is to be human, move it to another substrate, and go do
things that we can’t do as individual humans. We want as a species to
continue our evolution.”
Brain uploading, Koene agreed, was about
evolving humanity, leaving behind the confines of a polluted planet and
liberating humans to experience things that would be impossible in an
organic body. “What would it be like, for instance, to travel really
close to the sun?” he wondered. “I got into this because I was
interested in exploring not just the world, but eventually the universe.
Our current substrates, our biological bodies, have been selected to
live in a particular slot in space and time. But if we could get beyond
that, we could tackle things we can’t currently even contemplate.”