A new criticism of transhumanism recently made the rounds. It’s not particularly insightful, but many transhumanists shared it on the social networks.
It turns out that the article is more of a criticism of mind uploading in particular rather than transhumanism in general, though it does attack Ray Kurzweil near the beginning.
The context: NaturalNews.com, where the article was published, is ground zero for health crankery on the Internet. Their Facebook page has 378K likes. Colloidal silver, fluoride paranoia, ‘cancer cures’, you name it. Visit the website to see what I mean.
Writers at NaturalNews see Kurzweil’s vision as directly threatening to their worldview and business. Kurzweil argues we will all become nearly-immortal cyborgs by the late 2040s, through the progress of medicine and bionics. NaturalNews, on the other hand, advocates for taking natural substances to extend life and cure ills. The approaches aren’t mutually exclusive, though NaturalNews seems to think they are.
Both sides are mistaken. We will not become immortal cyborgs by the 2040s. Sometime this century it will be possible to build extremely durable bodies that do not age, but 2045 is premature. The line of thinking that foresaw major changes by that time, which was developed on the Extropy mailing list in the late 90s and was directly borrowed by Kurzweil for his books, has since been discredited. Simply put, progress towards molecular assemblers was much slower than we thought it would be. Other than that, most of Kurzweil’s predictions for 2009 were clear misses, and it will seems unlikely they will be fulfilled before the early 2020s. His entire timetable is a decade or more early, with predictions for 2029 even more over-optimistic than his predictions for 2019 or 2009.
However, this doesn’t rule out that the more extreme futures envisioned by transhumanists won’t come to pass. It’s just extremely evident that these changes won’t occur 30 years from now, exponential change or not. This is probably difficult for Kurzweil and other boomer futurists to accept, because it strongly implies they will not personally live up to the Singularity. That only leaves cryonics, which remarkably few boomer transhumanists are signed up for. Many are having psychological difficulty accepting what becomes more obvious every day; for them, it’s cryonics or nothing.
After introducing Kurzweil and transhumanism and dissing them, the article moves on to the core of its argument, which is directed against mind uploading rather than other features of transhumanism:
Let’s examine the claims of the transhumanism cult leaders like Kurzweil. They are saying that by 2045, all the following technology will exist:
Technology #1) A way to “scan” your entire brain and record every neuron and holographic patterning that exists in your brain.
Technology #2) A way to build an equally complex computing system that has equivalent computational capabilities as your brain.
Technology #3) A way to COPY your brain scan into the computing system. This is called “uploading” your brain to the machine.Once these three technologies exist, we are promised, we can all transfer our minds to computer systems and experience “digital immortality!” But wait a second. Something’s already missing here, do you see it? In this plan there is no mechanism to transfer your consciousness to the machine. So even if all three of these technologies are adequately developed (which is possible, by the way), they still don’t provide a way to merge your mind with a machine.
If you have #1 and #2, then #3 is trivial. If you can scan the brain in totality and record it as a computer file, it’s already copied. Copying it to a computer and running it as a program would be the same as transferring any program of similar size.
The writer is right to be skeptical that Technology #1 will exist before 2045. In all probability, it won’t. But it might. This would have to be a highly invasive form of scanning. Basically, a small swarm of microbots in the brain. (When futurists say ‘nanobots,’ what they actually mean is microbots, meaning robots on the micron scale. The nanoscale is too small to construct very useful robots except for molecular assemblers.) A collection of microbots with the volume of a pill would be enough. They could be flushed from the body after scanning were complete, if desired.
There is no conceivable non-invasive scanning technology that could evaluate the exact positions of single neurons and synapses deep in the brain. Synapses are too small, and the brain is too opaque. Unless there is some huge, unforeseen breakthrough that bends our current understanding of physical law, scanning the brain will require robots to go in there and do the surveying from the inside. These will take a number of decades to develop. Not easy, but the laws of physics clearly don’t forbid it.
Compared to Technology #1, Technology #2 is relatively trivial. We have already built supercomputers that exceed the lowball estimates of human brain processing power, and will have computers that exceed the higher realistic estimates within a couple years.
When considering “computers as powerful as the human brain,” TV show writers and other armchair sci-fi visionaries feel the need to invent fantastical new forms of computing to “bridge the gap,” but this is not necessary. There is nothing magical about computation in biological neurons; in fact, it is well understood. This surprises much of the public, which is about 40 years (or more) behind the state of the art of cognitive science. “The public” includes most sci-fi authors, as well.
Estimates of the computing power of the human brain rely on neuron count, firing speed, and guesses at the information-processing-per-neuron. There are about a hundred billion neurons in the brain, and they fire at most 200 times a second. At any particular time, most neurons in the brain aren’t firing at all. Neurons are not just mere logic gates; they perform operations such as multiplication. Even being generous, however, it’s difficult to assign each neuron more than a few operations per spiking event, and the average is probably far less. While it’s widely understood that smaller neural components like dendrites modulate neural processing, no one seriously thinks they perform independent computations.
Say that each neuron fires 200 times every second, and each spiking event represents a single logic operation. Assuming every neuron in the brain is constantly firing and each individual operation is computationally relevant to the overall picture (highly doubtful), the computational capacity of the human brain is about 100 billion times 200, or roughly 20 trillion operations per second. It can’t be much more than that, but it may be much less. In comparison, the world’s fastest computer operates at 33,860 trillion calculations per second (33.8 petaflops), which is 1,693 times greater.
IBM researcher Dharmenda Modha pegs the human brain’s computational capacity at 36.8 petaflops. Unfortunately, Dr. Modha has no scientific credibility whatsoever, but even assuming he is correct, that puts today’s fastest supercomputer at just a few percent short of the magic number.
The point is that we’re already basically there. We have computers “as complex” as the human brain. Sure, they consume many orders of magnitude more power than the human brain does, and take up a whole room rather than a skull, but we’re getting there. If we had a full readout of a human brain, there are computers that could run it today.
Do our cold, soulless supercomputers have the magic juju needed to simulate the ultra-special organic features of our glorious human brains? Yes. A serial computer, which relies on quickly performing computations in a sequence, can simulate a parallel computer, like the brain, which relies on a massive bank of slower nodes. Conversely, a parallel computer would never be able to simulate a serial computer of equivalent processing power, because its computations are spread out across so many disconnected nodes.
To visualize why, imagine a room full of men with calculators trying to cooperate to keep up with a supercomputer. In serial computing, the operations happen in a sequence; computation A gives result B, which goes to computation C, leading to result D, which goes to computation E, and so on. This happens billions of times a second. The concept is data dependency. The brain, however, has a billion nodes which only conduct a couple hundred operations a second. The brain is like the room full of men with calculators. They can perform many computations, but they can’t pass them between one another fast enough to keep up with the serial supercomputer. The supercomputer can simulate the men with their calculators, but they cannot simulate it. It’s too fast for them, and is much better at handling data dependencies.
Imagine the room full of men with calculators trying to perform computations that involve data dependencies. They might depend on a calculation happening across the room, and need to walk over to get the result, before they can begin their own computation. As a result, most of them would spend their time waiting. This is similar to the way in which most of the neurons in the human brain aren’t firing all the time. In fact, when too many neurons fire at once, that’s called a seizure. A serial computer, in contrast, can go full bore all the time, because it’s designed to.
The fundamental chemistry and principles of operation of human neurons is indistinguishable from flatworm neurons. We just have a lot more of them, in more complex arrangements. So if a mystical new form of computing would be necessary to create computers “as complex as” the human brain, the same mystical computing would be necessary to create computers as complex as a flatworm. Speaking of which, there is an international effort to exhaustively simulate every aspect of the flatworm C. elegans. If this effort is successful (and there’s every indication it will be), it will be a proof of concept that the human brain can be simulated in a computer. The basic computation is the same, it’s just that the human brain is far, far larger.
The standard view of consciousness today is similar to the view of life before the 19th century, or fire in the 17th century: it runs on magic. People don’t get that every phenomenon starts off with a mystical explanation, until it is eventually reduced to being explained in terms of reductive principles. Every single mysterious aspect of nature has gone through the same cycle.
The modern scientific view of consciousness is based on causal functionalism: that consciousness emerges as a sum of the interactions in the brain. That means a machine with the same computations as your brain would have the same consciousness you do. Alternate explanations do not hold up to Occam’s razor; they have to postulate that consciousness is something unique about the proteins or neurons the brain runs on, rather than the computations themselves. These alternate explanations would have to predict that someone with a computer implant to replace a brain module would partially lose their consciousness. Anyone who works with neural prostheses intuitively knows that this idea would be absurd.
If someone with 10% of their brain replaced with a computer that performed the same functions claimed to be conscious, then it’s extremely likely that someone with 20%, 30%, 40%, or 50% would claim the same. That goes all the way up to 100%. Philosopher David Chalmers calls this the “fading qualia” thought experiment.
So, there doesn’t need to be a magical process to port a human consciousness into a computer. The port could be completely natural, because the mind is a series of computations, not a piece of meat. It just happens to run on a piece of meat. Dualists throughout history actually had a point; although the natural world is a unified reality, the mind is better described as a computation running on the brain than as the brain itself.
I understand the incredulity of the Natural News crowd regarding the technology of mind uploading. It does sound fantastic. Its plausibility does need to be corroborated by the exhaustive simulation of simpler organisms. It isn’t right around the corner. However, it is a proposed technology that rests on plausible philosophical and technological foundations. There are even groups, like Global Future 2045, who are putting money and scientific expertise towards developing it as quickly as possible.
It’s easy to ignore developments that seem more than five years away. But, day by day, the evidence accumulates that the long-term vision of mind uploading is possible, and its benefits — virtuality, immortality, complete control over self and environment, cognitive upgrades — are ours to lose.