I listened to an online version of the "IBM impact 2010" talk about the progress of transhumanist technology in the coming decades by Ray Kurzweil . (Please do us all a favor and listen to this talk before reading further!) I must say that his persuasive power is far beyond that of any politician, news anchor, or radio talk show host I have ever known, and as a result I found myself trying desperately to put into words why most of us know deep down in our hearts that he is nuts, and to do so in a way that he couldn't try to strike down. I think I finally have gotten somewhere.
I was born on January 4, 1987. Before I discuss the specifics, I would like to point out that, if people of my generation buy what he is saying, it is likely that my peers and I will end up not addressing climate issues, saving at all for retirement or the college expenses of our children (not that this isn't already a problem in the U.S. of A.!). If we really think that biological beings as the status quo for humanlike-intelligent beings will end by 2045, what incentive would we have for curbing global warming, saving for retirement, saving for the college expenses of our children, or balancing the budgets of government(s)? This is one of the reasons why I think that it would be, to say the least, a massive travesty for people born after 1985 to take Kurzweil seriously, resulting in a generation of ordinary biological humans born into a world ruined by their forefathers.
Now on to the specifics. First off, as in his 2005 book "The Singularity is Near", Kurzweil repeatedly points out (and subsequently abusively extrapolates) the exponential growth of computing/information processing technology capacity since 1960 (or so). Why are we to expect this to continue? There have been other exponential trends relating to information technology that have not continued forever, so why are we to expect Moore's_law to continue to be valid until beyond 2030 or 2040? Second, as much of the centerpiece of his arguments, Kurzweil, in his "IBM impact 2010" talk, makes many claims about neuroscience and medicine which he never cites any scholarly references for. Any degree of fact checking confirms that every one of them is a joke. "Brain scan resolution has doubled every year for the past decade", "Nanobots have already been shown to be useful in medicine", "We can already simulate a major part of the human brain" etc. If you want the truth, look at actual papers published in the medical journals. Or, easier, Wikipedia. Third, Kurzweil's track record of predictions is not good, except for those relating to "ordinary" computers/information technology that were made less than a decade before the fact. Many of those predictions could easily have been made by other, more mainstream, computer scientists. While I do trust Kurzweil on predictions for the 5-7 year future (at the time of those predictions) in his area of expertise, I trust him neither on longer time scales nor on anything outside his field, such as neuromedicine. After all, we do not have computers beaming virtual reality onto our retinas, we do not have "wings for humans", etc. as he had predicted in 1990 would happen "by 2009". Heck, I even think the automated answering and voice recognition system of the local power company sounds stupider than some that were on the market almost ten years ago! Also, his new "book reading machine for the blind" released this year is really just a combination of text recognition of printed text (Which I can attest was already mature 15 years ago) and crude speech recordings and processing (I remember that elevator in a Miami hotel that said "Going up" and "Going down" already in 1998 that sounded no worse than Kurzweil's machine). So, just putting together two things that existed over a decade ago and calling it revolutionary. Dream on, Ray Kurzweil!
Finally, even if we could achieve what Kurzweil imagines this century, it would be highly dangerous and unethical to say the least. If we really had human-like AI, its development would be as unethical as "experimenting on human subjects" in unthinkable ways such as depriving them of all human interaction from birth. Oh, and yeah, we have no way of making sure that quasi-conscious, self-replicating robots won't take over the world and then "decide" to reprogram themselves so that they consider it a good thing to wipe out humans as quickly as possible. And surely if everyone spent much time in "full immersion virtual reality" the public would become psychotic due to Solipsism syndrome! (Oh yeah, "full immersion virtual reality" has existed for thousands of years, it's called "hallucinogenic drugs"!)
Predictions
i) Robots will not take over 40% of human jobs before 2040 without causing near-complete economic collapse (and thus no one able to build or maintain such robots) (Kurzweil predicts "By 2019")
ii) True self-replicating (from only worldwide raw materials that exist as of July 31, 2010) non-biological robots will not exist before 2100 (Kurzweil by implication predicts "early 2030s")
iii) Machines which beam images onto human retinas will not be used by more than 0.1% of the world's humans any time before 2035 (Kurzweil predicted 2019)
iv) No non-biological machine, before 2050, will be able to have a fully reciprocal, public political policy debate about all issues without prior, specific, human pre-programming, and trick humans into thinking it is human (Kurzweil predicts "By 2029")
v) Computers will not beat human grandmasters at Go (the game) on a 19 by 19 or larger board before 2023.
vi) Nanobots will not cure cancer and circulatory system disease before 2050.
vii) "Mind uploading" will not be possible before 2150 (Kurzweil predicts "By 2029".)
viii) If "virtual partners/spouses/mates" come into being before 2100, either a) Less than 0.1% of the world's population in the age range 15-45 will prefer them to human partners, or b) Most who do will end up with a strange new type of psychiatric disorder requiring clinical treatment. This will continue to hold true until 2115 or 20 years after the introduction of such virtual partners to the public, whichever comes first.
No comments:
Post a Comment