Friday, March 25, 2016

Musings on Artificial Intelligence and Superintelligence, Earth-Generated and Otherwise

I’m going off jobs this week, though this material is strongly if peripherally related.  Here are some thoughts on Nick Bostrom’s Superintelligence:  Paths, Dangers, Strategies, released last year, and the follow-on Raffi Khatchadourian November 23rd New Yorker article, “The Doomsday Invention”:

Point 1:  What we have been calling “artificial intelligence” really isn’t – it’s only algorithmic.  If A happens, do B.  Although capabilities of what can be done automatically have gone up exponentially since the BASIC computer language was developed in the 1960s, almost anything that is electronically generated could still be done, if need be, by using that form of code.  We can’t seriously use that phrase in the present tense until we break that boundary. 

Point 2:  Once we do that, and get, as in the first Terminator movie, “autonomous goal-seeking” from machines, what is to prevent them from the likes of what Arnold Schwarzenegger’s character did, determining that the best way to achieve its preset objective is to kill all the people?  Can we find a way of assuring that all post-algorithmic automata avoid that?  Could terrorists program and release them without that restriction? 

Point 3:  None of what Bostrom writes about is a new fear.  The classic in the field is still Bill Joy’s now 16-year-old article, “Why the Future Doesn’t Need Us.”  It explains how close we may be to letting nanotechnology, robotics, and genetic technology, which unlike nuclear, chemical, and biological threats can be furthered by forces much smaller than governments and large universities, get away from us and possibly even kill off our species.  It’s still available from its original source, Wired magazine, at http://www.wired.com/2000/04/joy-2.

Point 4:  We still don’t know about Ray Kurzweil’s Singularity, in which human and electronic intelligence merge and such nuisances as mortality go away, but between the slowing of Moore’s Law and Robert J. Gordon’s lifestyle observations in his new The Rise and Fall of American Growth, it doesn’t look good. 

Point 5:  Bostrom, according to the article, plans to be a corpsicle, in other words will have his body frozen after death for future revival.  Per science fiction author Larry Niven, we have real doubts about whether such semi-dead beings will even be welcome decades or centuries from now – they may be seen as selfish liabilities who had their lives already and whose bodies may be harvested for other purposes.  With our fear of dying thoroughly justified, I can’t blame him, but hoping to come back to life in that way may have the same disadvantages of religion… and be more expensive.

Point 6:  We have got about nowhere on knowing the source of consciousness.  Accordingly, we can never assume that machines of any kind, even if they talk and act a good game, have anything behind that.  The same goes for people after being teleported.  As we know from the Turing Test (the ability of computers to convincingly imitate people), it is possible, even easy these days, for things to act human with all the consciousness of an off-and-on air conditioner.

---
According to Bostrom, space probes travelling at 1% of the speed of light, or 1,860 miles per second, could canvass the entire Milky Way from a spot inside it within 20 million years.  Given that the galaxy contains, per Khatchadourian, ten billion “Earth-like planets,” and per Bostrom they have precipitated “a sum total of zero alien civilizations that developed technologically to the point where they become manifest to us earthly observers (italics his),” why have we not been visited or even contacted by life forms originating elsewhere?

Point 7:  As Bostrom suggested, they may just plain not exist, having been stopped from spacefaringness by any of many “Great Filters,” such as not experiencing the life-starting spark we still don’t understand, not progressing beyond single-celled organisms, or succumbing to asteroid strikes or stellar disturbances.

Point 8:  It is possible that intelligent life, in effect, carries the seeds of its own destruction, that certain types of technology sure to be developed at some point will cause all life on a planet to become extinct?  Could we ever know?  Bostrom philosophized on this, as have others in the past few decades.

Point 9:  Sentient creatures elsewhere could consistently be more like whales than humans, with little or no use of tools and, though intelligent, bound to a water or other environment hard to leave. 

Point 10:  Between these limitations and the restriction of the speed of light, we simply won’t be contacted all that often.  It could be tens of thousands of years, or more, between aliens’ appearances in our space. 

Point 11:  We have no reason to assume that extremely advanced creatures from elsewhere would even be visible to us.  Even if they chose to arrive in person, instead of watching us through the equivalent of super-powerful telescopes (and if they were half of the galaxy away, they would now be watching what we were doing in 48,000 BC), they could be cloaking themselves by staying out of our light range.  We have no idea what beings could do after thousands of years of post-Industrial Revolution progress, let alone millions, so we can’t rule anything out. 

---
That last sentence applies to artificial intelligence and superintelligence as well.  Don’t say I didn’t warn you. 


Friday, March 18, 2016

This Year’s Book – How Can We Get More Big Changes? – Part 2

Last week’s post was about Robert J. Gordon’s large, comprehensive, and stunning Rise and Fall of American Growth, which made the case that our country is economically and technologically leveling off, with almost all of the truly fundamental life improvements we have seen since 1870 already in place 45 years ago. 

There is no doubt that Gordon is correct.  If you look at one of the most important books by Herman Kahn, likely the leading futurist of all time, 1967’s The Year 2000 co-authored by Anthony J. Wiener, you will find, among many others, two related lists.  Most of the items in the first, “One Hundred Technical Innovations Very Likely in the Last Third of the Twentieth Century,” have come to pass at least partially, but almost none of the 25 in the second, “Some Less Likely but Important Possibilities,” described by the authors as “even money bets, give or take a factor of five,” have happened, even with 16 additional years.  The only two which have seen real achievement are “effective chemical or biological treatment for most mental illnesses” (not most, but some), and “practical laboratory conception and nurturing of animal (human?) foetuses” (not full-term development, but in vitro fertilization anyway).  There is no doubt that these most sober, thoughtful, and eminent prognosticators of almost 50 years ago, while having no concept of the Internet, would be shockingly disappointed at 2016 technology, and dumbfounded that, as Gordon points out, current apartments would be so functionally similar to typical 1940s living quarters.

Gordon wrote a section, the Postscript, with recommendations on how to implement major innovations.  Unfortunately I found that the weakest part of the book.  It read like a catalogue of a general, mostly liberal, public policy agenda, with connections to life-changing inventions almost all ranging from weak (more immigration) to nonexistent (a higher minimum wage, less prison time, a carbon tax).  If we keep the only idea there favorable to large life-changing technological improvements, a lifting of “regressive regulations,” what can we add to it?

The first move is to agree upon and realize the problem.  Per Peter Thiel, 140-character messaging is not as profound as flying cars.  Even if it were, there is evidence that Moore’s Law, the doubling of computing capability every 18 months, along with its commensurate price decreases, is ending at least temporarily, so it will help us even more to look beyond electronics. 

Second, we can combine that with an understanding of why predictions often fail.  As Joel Garreau wrote in 2005’s Radical Evolution, there are five major reasons for that.  They are more complications than originally expected, prohibitive costs, replacement by other new and unexpected technologies, negative experiences or perceptions related to the idea, and conflicts with cultural and other human behavior.  These have prevented, in the United States and elsewhere, a cancer cure, magnetic levitation trains, home mainframe computers, a nationwide identification database, and more extensive public transportation respectively.  Possible innovations should be checked against these standards, both to anticipate and possibly overcome future problems and to identify those with no real chance of becoming widespread reality. 

Third, we can agree, at least to some extent, on which inventions would be fundamental improvements and which would be only incremental.  For example, though I take issue with Gordon’s view that self-driving cars are not fundamentally different – they will have far-reaching effects on everything from alcohol and drug-consuming patterns to much of American manufacturing and on to our philosophical sense of what freedom actually is – we can develop a set of standards to determine who is right and differentiate the likes of electric power from such as the latest iPhone release. 

Fourth, we can then offer incentives for people and companies to create and develop large life-changing improvements with reasonable chances of widespread acceptance.  One way is through taxes, where research and development on such technologies should gather large breaks.  Another is by appealing to the wealthiest individuals and corporations, those accumulating amounts of cash so vast they explain why the money supply is growing much faster than inflation.  Even those with far less than Bill Gates’s $76 billion and Apple’s $178 billion will see the merit, once the above issues are reasonably settled, of offering some large prizes ($1 billion and up?) for widespread implementation of fundamental advances.

Fifth, we need to move space exploration out of the pure-science-at-taxpayer’s-expense stage and into industrialization.  That is the single most important area-specific change we can make.  Space travel, especially when involving humans, has long been not only a symbol for but a source of innovations, and more of it would help us move forward more than might make logical sense.  The National Aeronautics and Space Administration has served us well over its 58 years, but it is time for it to get out of the mission-originating business and become exclusively an advisor and technology resource for SpaceX, Virgin Galactic, and literally thousands of other and future companies.  There are at least three large justifications for this change.  First, government is inherently too large and slow to excel at innovating with leading-edge technology.  Second, as we have learned since the Apollo moon landings, when space exploration becomes only another federal expense that makes it susceptible to budget cuts.  Third, there are so many possibilities for making large profits in space without governmental competition that companies will have great incentive to be involved there, especially if special tax reductions, as in general above, help them along.  One massive area almost certain to be technically viable is the harvesting of solar power in space, which, with the potential to end almost all of our energy needs cheaply, would certainly qualify as fundamental change. 


In the effort to resume large life improvements, there are two traps we must avoid.  First is cutting off or greatly inhibiting outer-space industry for environmental-protection reasons.  Space is incomprehensibly gigantic, and if manufacturing can be done freely there not only is it better in many functional ways but it can be much less damaging than on Earth, even if regulations are strictly limited.  The second is rewarding ideas instead of their implementation.  Ideas are not what we need.  From technical journals to science fiction to just-plain common knowledge, we have long had plenty, and there will be more.  It is time to focus on action.  We can get back on track, if we realize what we need to do and do it.   

Friday, March 11, 2016

This Year’s Book: Are the Big Changes Over? – Part 1

More than four-fifths of 2016 still remains, but we already have a strong candidate for the book of the year.   Northwestern professor Robert J. Gordon, cited by Bloomberg Markets as one of its most influential 2013 thinkers, has issued a massive but remarkably welcoming-looking hardback on a topic affecting almost everything in public or private business policy.  Titled “The Rise and Fall of American Growth:  The U.S. Standard of Living Since the Civil War,” it develops the thesis that, from a variety of standpoints, we have not continued the improvements to our lives which were greatest in the 100 years ending 1970, and cannot expect to.  It is a three-pound counterweight to the Moore’s Law-fueled projections of rapidly improving computing capabilities driving enormous lifestyle improvements, in which Gordon maintained that the changes we have seen since 1969 have almost all been incremental instead of fundamental, and have not approached the value of “electric lighting, indoor plumbing, home appliances, motor vehicles, air travel, air conditioning, and television,” all of which were in place over 50 years ago. 

For something not mentioned much in print (though I did pass along David Bodanis’s observation that the home lives in the movie E.T. seemed hardly changed 25 years later, and expressed doubt that computer speeds doubling every 18 months meant similar improvements elsewhere, in 2013’s Choosing a Lasting Career), Gordon’s theory has real merit.   What more can we say about it?

First, the end of fundamental innovations is not from lack of possibilities.  On only one area of those, not covered by Gordon, consider the following quotation, from R. Buckminster Fuller’s 1970 I Seem to Be a Verb:  “By 1988, says a federal report, the biggest businesses in the U.S. will be:  (1) The manufacture and service of cars, and (2) the manufacture, insertion, and service of artificial hearts.”  We were on that pace through the early 1980s, until the most advanced such device, the Jarvik-7, proved inadequate, and successors fared little better.  Since then, the leader in artificial hearts has become the American company SynCardia, which in 2013 said proudly that its devices had been installed, as temporary pre-transplant stopgaps, in 161 patients.  As for those transplants, only 2,332 took place in the United States in 2011, which would be a fundamental gain if it were closer to the 610,000 annual American heart disease deaths.  Our general lack of fundamental progress is summarized well in another quotation, passed along by Gordon from author Peter Thiel:  “We wanted flying cars, instead we got 140 characters.”      

Second, as I have observed publicly for five years, new technology never employs more than a tiny fraction of what the likes of cars, modern houses, and air travel did and continue to do.  As of 2010, Twitter, used by tens of millions of Americans, employed a total of 300 people, half the number working in single AT&T buildings I worked in decades ago.

Third, although Gordon emphasizes slower gains (not a drop) in per-worker productivity, one explanation I did not see in the book.  A high proportion of jobs added since the Great Recession have been low-level – at one point, fully one-third of them were with temporary help agencies – and have inherently small productivity.  Cleaning and retail counter positions are necessary, but are never going to contribute the dollar values once added by now obsolete manufacturing or even office jobs, not to mention their management. 

Fourth, I have three quarrels with what Gordon called “headwinds,” or factors preventing larger economic growth.  Education levels flattening out, which they are doing in this country, is not a problem, as, consistent with what Gordon himself wrote, most of the truly life-altering improvements going into widespread service between 1870 and 1970 were invented early in that time, when the high school graduation rate was less than 5%.  Baby boomers leaving the workforce as they reach retirement age is not a demographic obstacle, it is a permanent jobs crisis effect, since, as my 2006 doctoral research showed, even before the Great Recession the massive majority wanted to work into their 70s and beyond.  And inequality is not a cause of issues with technology or anything else, it is an effect, a natural result of higher and higher scalability in which products are reproduced almost without cost and so must benefit ever more limited sets of people; in contrast, cars provided money for millions, as they required not only designers and investors but factories full of people to be paid to actually make them.

Fifth, is it possible that, with the amount of business innovation always in progress, that we are in nothing more than a 50-year lull?  That, which I have not assessed in detail, is a real potential explanation for what this author has wrought. 

If what Robert J. Gordon is saying is true, how can we best deal with it?  Be here next week. 


Friday, March 4, 2016

February: Another Fine Jobs Month, but AJSN Says America’s Still 18.2 Million Jobs Short

According to at least one story issued late last night, our stock market was hoping for a “Goldilocks” jobs report.  That means not too bad (giving bad tidings of the economy in general) or too good (to egg on the Federal Reserve to raise interest rates again), but “just right.” 

This morning’s data, I’m afraid, may have been too good.  Net new nonfarm jobs came in at 242,000, way over the 200,000 consensus projection.  Official seasonally adjusted unemployment held at 4.9%, but more people are working, as shown by labor force participation and the employment-population ratio each up 0.2%, a lot for one month, to 62.9% and 59.8%.  On the bad side, those officially jobless for 27 weeks or longer ticked up 100,000 to 2.2 million, those wanting work but not looking for it during the previous year rose more than that to 3.626 million, and average hourly earnings gave up 4 cents of January’s 12-cent gain to reach $25.35.  Unadjusted unemployment was down 0.1% to 5.2%.  Those working part-time for economic reasons, unsuccessfully seeking a full-time opportunity, stayed at 6.0 million.  The number of people in every single other one of the unattached and marginally attached labor categories, even those claiming no interest whatever in working, dropped from January to February.  Overall, the American Job Shortage Number, which indicates the total latent demand for work opportunities, fell 188,000, as follows:




Given the positive data above that may seem like a weak improvement, but the number of officially unemployed Americans fell only 90,000, which, when translated into the number of jobs they would absorb if work were readily available, was more than offset by the gain in those not searching for it for a year or more. 

Compared with a year before, the AJSN, which is not seasonally adjusted, improved over 1.1 million, 788,000 from the drop in officially jobless.  Most of the categories of marginal attachment were off as well, most notably the count of discouraged workers, which fell from 732,000 to 599,000.


So how much better are we off than in January?  Given not only the strong number of new jobs, but, more importantly, the large improvements in labor force participation and employment to population, we are considerably.  It is a good sign as well when the count of people claiming no interest in work declines, which it did by over half a million, as many reconsider that when perceptions of more opportunities reach them.  We have established a strong pattern of monthly job growth above the 125,000 to 140,000 needed to cover population increases, and the participation ratios are well away from last fall’s recent records.  
Average wages are not really doing anything more than inflation, but there are still too many would-be workers for sustained improvement there.  Overall, while the turtle is still a turtle, he stretched his legs quite a bit, by his standards, last month.