Friday, October 26, 2018

Robots and Artificial Intelligence: Five Mid-Year Articles, Loads of Insights - II


A fourth article is yet another behind the curve, but in some respects a useful review.  Ben Casselman’s June 25th New York Times “Robots or Job Training:  Manufacturers Grapple With How to Improve Their Economic Fortunes” reminded us how automata would allow employers to avoid both hiring flawed people and paying more (actually, the article said “help ease their labor crunch”), that training potential workers might allow them to be suitable (what a novel idea), and that “rising productivity” could allow higher wages (only sometimes, as companies have competitors benefitting from it as well).  Casselman also quoted a Federal Reserve economist saying that with 3.8% joblessness “eventually you’re going to run out of easy-to-find workers” (not when we can easily fill over 15 million more positions, two-thirds to those not officially unemployed), named the problem with “unskilled laborers” not showing up and not doing their jobs (paying them more to increase demand for their positions is still legal, and some business propositions dependent on low wages won’t succeed anyway), and touched on both the significance of accumulated efficiency in reducing the number of positions and higher labor costs encouraging “ways to economize.”  He said that “raising rates too quickly could be a costly mistake for the Fed” (was Wednesday’s 600-point Dow drop a result of that?), and noted that “The Fed’s most recent projections estimate that the unemployment rate will fall to 3.5 percent next year” (what does that assume about the effects of our escalating trade war?).  All worthy of discussion, if not agreement.

The final piece of the five is a view from Andy Clark, “a professor of logic and metaphysics,” that “We Are Merging With Robots.  That’s a Good Thing.”  This further New York Times article suggested that automata’s amalgamation with artificial intelligence may add the third area of human capability.  He named ten things “true today,” some among them that “sex and companionship robots are already here,” that “the human genome itself is now an object of control and intervention” (we are just getting started, with gigantic ethical issues along with massive opportunities just around the corner), and that “neuro-enhancement, the improvement by drugs, practices, or implants of normal mental functioning, is possible and may soon become the norm” (did that start with Prozac, 32 years ago?).  He may have been carried away in saying that “sharing and group solidarity are now easier than ever before,” when that is woefully untrue with in-person interaction, but noted that “the boundaries between body and machine, between mind and world, between standard, augmented and virtual realities, and between human and post-human” are becoming less clear.  More properly, though, they are being redefined, as such borders were breached over 700 years ago with the first pair of eyeglasses.  Clark continued by writing that what he called these machine “subintelligences” were “not yet intelligences like our own,” correct unless we consider a toaster smart for being able to brown bread, and called the “two most important… questions” how we should “negotiate” all of these possibilities” (of course), “and what costs are we willing to tolerate along the way?” something now being negotiated with driverless vehicles.  He also rose the issue of inequality, certain to happen if these improvements require personal resources, and concluded that “what is up for grabs is what we humans are, and will become.”

Instead of my own conclusions beyond the parenthetical statements above, I will defer to a sixth piece of writing, dated August 11th and once more in the New York Times, from Sherry Turkle, for decades a clear and freethinking voice on the subjects of the program she teaches at M.I.T., “science, technology, and society.”  The title, “There Will Never Be an Age of Artificial Intimacy,” says most of it.  No matter how effective robots or computer applications seem to be, their lack of emotions will invariably “lead to an empathic dead end.”  There is all the difference between these things and living beings, even those of terrestrial or other species with which we cannot verbally communicate, which can feel.  In that sense artificial intelligence is a misnomer, and we can see no work in progress with a chance to change that.  More than anything else about robots and artificial intelligence, that is what we need to understand. 

Friday, October 19, 2018

Robots and Artificial Intelligence: Five Mid-Year Articles, Loads of Insights - I


Five pieces published a few months ago helped us.  Not by giving views we could all agree with, but by adding to our necessary national dialogue and refreshing us on what’s been happening in this field.  Here’s what they had to say on this two-in-one area.

In “Moguls and Killer Robots,” which took up most of page 1 of the June 10th New York Times business section and two-thirds of another one, Cade Metz set out to tell us what happened between business information technology titans Mark Zuckerberg and Elon Musk, and ended up providing information bearing repetition.  That included Musk’s statements that artificial intelligence was “potentially more dangerous than nukes,” and that “we are headed toward either superintelligence or civilization ending,” of which Zuckerberg, perhaps due to his experience being in a less physically hazardous area, Facebook, than Musk’s space tourism at least partially disagreed.  We also saw what can happen when business decisions, in this case “a $9 million A.I. contract (Google) had signed with the Pentagon,” conflict with the views of employees, who threatened a “rebellion,” and the statement apparently not obvious to some from an Oxford research director that “you can now talk about the risks of A.I. without seeming like you are lost in science fiction,” echoed by concerns from a Cornell computer science professor that “the kind of systems we are creating are very powerful… and we cannot understand their impact.”  Metz indirectly touched on the difference between narrow AI, focused and benign, and general AI, which need not be either, and reminded us how they can depart from the usual human ways of solving problems, when such a facility was directed to maximize scores in a boat racing computer game and did that “while spinning in circles, colliding with stone walls and ramming other boats.”  The founder of Google’s Deep Mind, the effort creating the board-game Go program which beat a major champion, Demis Hassabis, summarized our needs well by saying “we need to use the downtime, when things are calm, to prepare for when things get serious in the decades to come.”  We can disagree that all is tranquil now on the AI-danger front, but not that it will at least threaten to get worse later. 

“If the Robots Come for Our Jobs, What Should the Government Do?”  Neil Irwin posed this question in article form in the June 12th New York Times.  He acknowledged that “lots of smart technologists and futurists are convinced that we are on the cusp of a world in which artificial intelligence, robotics and other technologies will make a large portion of today’s jobs obsolete,” and considered a guaranteed income, along with “overhauling intellectual property law so that the companies that develop valuable patents and trademarks don’t have such a lengthy monopoly on their innovations” (will Mickey Mouse be in the 2100 public domain?) and “work-sharing programs” by in effect removing 20% of each of 500 positions instead of fully laying off 100 people (worthwhile).  He quoted someone saying it was “increasingly crucial that people continually upgrade their skills to keep up with changing technology” (a clear strategy for individuals, but a cop-out for helping employment in general), and advocating “making job benefits like health insurance and retirement funds more “portable”” (a direction in which we have already been moving).  Now, if we only had similar sets of proposals to deal with globalization and efficiency…

I published many of the points made by Mary Flanagan in “The Rise of the “Automacene”:  How Robots Will Define the Next Epoch in Human History” (Salon, June 16) between book covers five and six years ago, but they are also worth reiterating.  Flanagan cited research-driven estimates that automation, timeframe unspecified, has 98% and 99% chances of spelling the end of positions as loan officers and tax preparers respectively, but less than 1% to do the same for nursing assistants and mental health social workers.  To prepare for a changing “future of work” and “a future in which unprecedented unemployment is the norm,” the author recommends “an educated populace” (see my comment above) and “ability to interconnect disparate ideas” (my personal experience from a lifetime of having this skill, and everything I’ve read about actual workplace trends, say that employers still almost never care about that).  There’s much more to our “need to decouple our jobs from the meaning and identity we expect them to provide us,” on which I wrote an entire chapter and which has been dropping since the 1950s, but there’s no arguing with Flanagan’s conclusion that “we’ll only be able to navigate the upcoming tumultuous changes in society by embracing deep conversations on what it means to be human in the era of machines,” even though precious few Americans want to have such things.

Two more to follow, for next week along with conclusions.


Friday, October 5, 2018

September Employment Report: Latent Demand for Jobs, Per the American Job Shortage Number (AJSN), Dropped, As Did Unemployment – We’re Now Only (?) 15.6 Million Jobs Short


This morning’s was another one-of-a-kind Bureau of Labor Statistics Employment Situation Summary.

I didn’t see a projected number for net new nonfarm payroll positions, and it didn’t matter much, as they took a breather, up 134,000 or essentially the same as our growing population absorbed.  Other results, though, were surprising.  The marquee adjusted unemployment rate found another 0.2% to decline for 3.7%, and is its smallest since December 1969.  The adjusted rate, now lower as September is an above-average working month, fell 0.3% to 3.6%.  We are now at an adjusted total of 6.0 million jobless, down 200,000, and average private nonfarm payroll wages again exceeded inflation with an 8 cent per hour gain to $27.24.  While the labor force participation rate stayed at 62.7%, the other indicator of how common it is for Americans to be on the job, the employment-population ratio, improved 0.1% to 60.4%.

There was a downside, though.  Although unemployment was off, those officially jobless for 12 months or longer now count 1.4 million, up 100,000.  The set of those working part-time for economic reasons, or holding on to shorter-hour opportunities while looking thus far unsuccessfully for full-time ones, gained back the 200,000 it lost in August and is again at 4.6 million. 

The American Job Shortage Number or AJSN, the measure of how many additional positions could be filled if everyone knew they were truly easy to get, plunged 943,000, maybe half of that seasonal (the AJSN is not adjusted), as follows:



The large drop was from the one-two punch of latent demand from those officially unemployed, down 544,000, and that from the greatly reduced count of those wanting work but not looking for it for 12 months or longer, off 445,000.  Changes in the other components were relatively small, with the largest other loss from those discouraged and the highest gain, once again, from people claiming no interest in working.  The share of latent demand coming from the unemployed is now 33.2%, which is, along with the AJSN itself, another post-Great-Recession low. 

Once again, the metric is down over one million from a year ago.  In September 2017 there were 800,000 fewer unemployed and 364,000 fewer wanting work but not searching for at least a year.  Other six-digit year-over-year AJSN changes came from those non-civilian, institutionalized, or off the grid, 1.2 million higher then, and the 2.1 million gain in people saying they do not want a job. 

What to make of this month?  I think it’s OK that net new jobs only matched national growth this time – that has been very strong for years, and cannot be expected to always exceed population gain.  It is healthy that fewer who want jobs aren’t looking, though that number has been oscillating more than improving.  I am concerned about two things.  Two groups which seem to slip through the cracks of prosperity analyses, the long-term unemployed and those working part-time for economic reasons, are within another month like this one of being disturbingly large given the floor-low jobless rates.  And second, the steady growth of those on the shelf, or saying they do not want to work, means that we are in long-term danger of employment being less usual.  Overall, though, the good things dominated – the turtle took another step in the right direction.