Prediction #1: “The computational ability of a $4,000 computing device (in 1999 dollars) is approximately equal to the computational capability of the human brain (20 million billion calculations per second). (“2019” timeline prediction, page 203.)
Reality:
A $4000 computing device in 1999 dollars is equivalent to about a
$6000 dollar computing device today. There is no $6000 computing
device that can compute anywhere near as fast 20 million billion
calculations per second. A computer such as the Apple iMac Pro (with
a price of 5000 dollars) performs at a clock speed of 3.2 gigahertz, which
results in about 3 billion operations per second. This article claims that when loaded with 18 core processors (which raises the price above $10,000), the Apple iMacPro can do 11 trillion floating point operations per second (11 teraflops). Even if we use that figure rather than the clock speed, we still have computing capability about 1000 times smaller than the computing capability predicted by Kurzweil for such a device in 2019.
Prediction #2: “Computers are now largely invisible and are embedded everywhere – in walls, tables, chairs, desks, clothing, jewelry, and bodies.” (“2019” timeline prediction, page 278.)
Reality: Nothing like this has happened, and while computers are smaller and thinner, they are not at all "largely invisible."
Prediction #3: “Three-dimensional virtual reality displays, embedded in glasses and contact lenses, as well as auditory 'lenses,' are used routinely as primary interfaces for communication with other persons, computers, the Web, and virtual reality.” (“2019” timeline prediction, page 278.)
Reality:
Such things are not at all used “routinely,” and I am aware
of no cases in which they are ever used. There is very little
communication through virtual reality displays, and when it is done
it involves bulky apparatus like the Occulus Rift device, which
resembles a scuba mask.
Prediction #4: “Most interaction with computing is through gestures and two-way natural language spoken communication.” (“2019” timeline prediction, page 278.)
Reality: Partially correct, if we consider smartphone swipes as a gesture.
Prediction
#5: “High-resolution, three dimensional visual and auditory
virtual reality and realistic all-encompassing tactile environments
enable people to do virtually anything with anybody, regardless of
physical proximity." (“2019” timeline prediction, page 278.)
Reality: This sounds like a prediction of some reality similar to the Holodeck first depicted in the TV series Star Trek: The New Generation, or a prediction that realistic virtual sex will be available by 2019. Alas, we have no such things.
Prediction
#6: “Paper books or documents are rarely used and most learning
is conducted through intelligent, simulated software-based teachers.”
(“2019” timeline prediction, page 278.)
Reality:
Paper books and documents
are used much less commonly than in 1999, but it is not at all true
that we have “intelligent, simulated software-based teachers.”
For example, you cannot download a “history teacher” app who you
can talk to about history like you would talk to a flesh-and-blood
history teacher.
Prediction
#7: “The vast majority of transactions include a simulated
person.” (“2019” timeline prediction, page 279.)
Reality:
A large percentage of
transactions are electronic, but very few of them involve a simulated
person.
Prediction
#8: “Automated driving systems are now installed on most
roads.” (“2019” timeline prediction, page 279.)
Reality:
Although there are a few
self-driving cars on the road, 99% of traffic is old-fashioned
traffic with human drivers.
Prediction
#9: “People are beginning to have relationships with automated
personalities and use them as companions, teachers, caretakers and
lovers.” (“2019” timeline prediction, page 279.)
Reality:
This isn't happening in
2019. You can have a little interaction with an on-screen figure in a video game, but it's a very limited type of thing (such as choosing one of 5 text responses you can make to the person).
Reality: The public has not yet even heard of tiny flying weapons.
Prediction #11: "The expected lifespan...has now substantially increased again, to over one hundred." (“2019” timeline prediction, page 208.)
Reality: The most recent lifespan figure is 78.9 years for US residents. A December 2018 article says the US lifespan has dropped for three years in a row.
So
Kurzweil's predictions for 2019 were very far off the mark. Are there
any reasons to think that his predictions for 2029 and 2099 are
unlikely to be correct? There certainly are.
One reason is that Kurzweil never did much to prove his claim that there is a Law of Accelerating Returns causing the time interval between major events to grow shorter and shorter. On page 27 he tries to derive this law from evolution, claiming that natural evolution follows such a law. But we don't see such a law being observed in the history of life. Not counting the appearance of humans, by far the biggest leap in biological order occurred not fairly recently, but about 540 million years ago, when almost all of the existing animal phyla appeared rather suddenly during the Cambrian Explosion. No animal phylum has appeared in the past 480 million years. So we do not at all see such a Law of Accelerating Returns in the history of life. There has, in fact, been no major leap in biological innovation during the past 30,000 years.
Kurzweil's logic on page 27 contains an obvious flaw. He states this:
The advance of technology is inherently an evolutionary process. Indeed, it is a continuation of the same evolutionary process that gave rise to the technology-creating species. Therefore, in accordance with the Law of Accelerating Returns, the time interval between salient advances grows exponentially shorter as time passes.
This is completely fallacious reasoning, both because the natural history of life has not actually followed a Law of Accelerating Returns, and also because the advance of technology is not a process like the evolutionary process postulated by Darwin. The evolutionary process imagined by Darwin is blind, unguided, and natural, but the growth of technology is purposeful, guided and artificial.
On the same page, Kurzweil cites Moore's Law as justification for the Law of Accelerating Returns. For a long time, this rule-of-thumb held true, that the speed of a transistor doubled every two years. But in 2015 Moore himself said, "I see Moore's law dying here in the next decade or so." Machines smarter than humans would require stratospheric leaps forward in computer software, but computer software has never grown at anything like an exponential pace or an accelerating pace. Nothing like Moore's Law has ever existed in the world of software development. Kurzweil has occasionally attempted to suggest that evolutionary algorithms will produce some great leap that will speed up the rate of software development. But a 2018 review of evolutionary algorithms concludes that they have been of little use, and states: "Our analysis of relevant literature shows that no one has succeeded at evolving non-trivial software from scratch, in other words the Darwinian algorithm works in theory, but does not work in practice, when applied in the domain of software production."
On the same page, Kurzweil cites Moore's Law as justification for the Law of Accelerating Returns. For a long time, this rule-of-thumb held true, that the speed of a transistor doubled every two years. But in 2015 Moore himself said, "I see Moore's law dying here in the next decade or so." Machines smarter than humans would require stratospheric leaps forward in computer software, but computer software has never grown at anything like an exponential pace or an accelerating pace. Nothing like Moore's Law has ever existed in the world of software development. Kurzweil has occasionally attempted to suggest that evolutionary algorithms will produce some great leap that will speed up the rate of software development. But a 2018 review of evolutionary algorithms concludes that they have been of little use, and states: "Our analysis of relevant literature shows that no one has succeeded at evolving non-trivial software from scratch, in other words the Darwinian algorithm works in theory, but does not work in practice, when applied in the domain of software production."
The biggest reason for doubting Kurzweil's predictions beyond 2019 is that they are based on assumptions about the brain and mind that are incorrect. Kurzweil
is an uncritical consumer of neuroscientist dogmas about the brain
and mind. He assumes that the mind must be a product of the brain,
and that memories must be stored in the brain, because that is what
neuroscientists typically claim. If he had made an adequate study of
the topic, he would have found that the low-level facts collected by
neuroscientists do not support the high-level claims that
neuroscientists make about the brain, and frequently contradict such
claims. To give a few examples:
- There is no place in the brain suitable for storing memories that last for decades, and things like synapses and dendritic spines (alleged to be involved in memory storage) are unstable, "shifting sands" kind of things which do not last for years, and which consist of proteins that only last for weeks.
- The synapses that transmit signals in the brain are very noisy and unreliable, in contrast to humans who can recall very large amounts of memorized information without error.
- Signal transmission in the brain must mainly be a snail's pace affair, because of very serious slowing factors such as synaptic delays and synaptic fatigue (wrongly ignored by those who write about the speed of brain signals), meaning brains are too slow to explain instantaneous human memory recall.
- The brain seems to have no mechanism for reading memories.
- The brain seems to have no mechanism for writing memories, nothing like the read-write heads found in computers.
- The brain has nothing that might explain the instantaneous recall of long-ago-learned information that humans routinely display, and has nothing like the things that allow instant data retrieval in computers.
- Brain tissue has been studied at the most minute resolution, and it shows no sign of storing any encoded information (such as memory information) other than the genetic information that is in almost every cell of the body.
- There is no sign that the brain or the human genome has any of the vast genomic apparatus it would need to have to accomplish the gigantic task of converting learned conceptual knowledge and episodic memories into neural states or synapse states (the task would presumably required hundreds of specialized proteins, and there's no real sign that such memory-encoding proteins exist)
- No neuroscientist has ever given a detailed explanation of how such a gigantic translation task of memory encoding could be accomplished (one that included precise, detailed examples).
- Contrary to the claim that brains store memories and produce our thinking, case histories show that humans can lose half or more of their brains (due to disease or hemispherectomy operations), and suffer little damage to memory or intelligence (as discussed here).
Had he made a study of paranomal phenomena, something he shows no signs of having studied, Kurzweil might have come to the same idea suggested by the neuroscience facts above: that the brain cannot be an explanation for the human mind and human memory, and that these things must be largely effects or aspects of some reality that is not neural, probably a spiritual facet of humanity.
Since
he believes that our minds are merely the products of our brains, Kurzweil thinks that we will be able to make machines as intelligent as we are, and eventually far more intelligent than we are, by somehow leveraging some "mind from matter" principle used by the human brain. But no one has any credible account of what such a principle could be, and certainly Kurzweil does not (although he tried to create an impression of knowledge about this topic with his book How to Create a Mind). We already know the details of the structure and physiology of the brain, and what is going on in the brain in terms of matter and energy movement. Such details do nothing to clarify any "mind from matter" principle that might explain how a brain could generate a mind, or be leveraged to make super-intelligent machines.
Holding mistaken beliefs about minds being made by brains, and memories being stored in brains, Kurzweil derives from these beliefs various bizarre predictions for 2099 that will not prove accurate, such as the belief that in that year "there is no longer any clear distinction between humans and computers," and "most conscious entities do not have a permanent physical presence," and "the number of software-based humans vastly exceeds those still using native neuron-cell-based computation," and "life expectancy is no longer a viable term in relation to intelligent beings" (page 280). There is a good reason to think that humans will indeed one day no longer be concerned by the prospect of death, but that will probably come from understanding that our minds never arose from our brains, but from a spirit or soul that is imperishable.
Since he predicts so many fantastic technology advances coming from a Law of Accelerating Returns, there is a certain cosmic problem faced by Kurzweil: the fact that we do not observe any sign of extraterrestrial activity anywhere else in outer space. We would think that if a single civilization got the type of super-powers Kurzweil imagines, that they would have transformed our galaxy; but we can see no signs of such a thing. Kurzweil suggests a solution to this, one that is quite wacky. What he suggests is that extraterrestrials are too tiny to be noticed. He says on page 257, "A computational-based superintelligence of the late twenty-first century here on Earth will be microscopic in size." From this most dubious assumption, he makes the batty conclusion on page 258 that extraterrestrial spaceships are "thus likely to be smaller than a grain of sand...Perhaps this is one reason we have not noticed them."
It's like a new religion
Holding mistaken beliefs about minds being made by brains, and memories being stored in brains, Kurzweil derives from these beliefs various bizarre predictions for 2099 that will not prove accurate, such as the belief that in that year "there is no longer any clear distinction between humans and computers," and "most conscious entities do not have a permanent physical presence," and "the number of software-based humans vastly exceeds those still using native neuron-cell-based computation," and "life expectancy is no longer a viable term in relation to intelligent beings" (page 280). There is a good reason to think that humans will indeed one day no longer be concerned by the prospect of death, but that will probably come from understanding that our minds never arose from our brains, but from a spirit or soul that is imperishable.
Since he predicts so many fantastic technology advances coming from a Law of Accelerating Returns, there is a certain cosmic problem faced by Kurzweil: the fact that we do not observe any sign of extraterrestrial activity anywhere else in outer space. We would think that if a single civilization got the type of super-powers Kurzweil imagines, that they would have transformed our galaxy; but we can see no signs of such a thing. Kurzweil suggests a solution to this, one that is quite wacky. What he suggests is that extraterrestrials are too tiny to be noticed. He says on page 257, "A computational-based superintelligence of the late twenty-first century here on Earth will be microscopic in size." From this most dubious assumption, he makes the batty conclusion on page 258 that extraterrestrial spaceships are "thus likely to be smaller than a grain of sand...Perhaps this is one reason we have not noticed them."
No comments:
Post a Comment