Friday, June 30, 2017

Driverless Vehicles - Where Are We Now? – Part 2

One problem with writing on autonomous vehicles is that the topic can be so hot, with information pouring out so quickly, that I can write week after week without finishing it.  However, I will continue with my plan for summarizing and interpreting the June articles, though my stack now contains 16 of them. 

The first is Billy Duberstein’s and The Motley Fool’s June 3rd “Everything You Need to Know About Self-Driving Cars (But Were Afraid to Ask),” a good summary mentioning LIDAR, radar, and another car-autonomy-characteristic level scheme.  We stay on the business side with “G.M. Wants to Drive the Future of Cars That Drive Themselves” (The New York Times, June 4th), in which Bill Vlasic shows how General Motors, not previously on the technology’s front lines, is now moving money from other-country efforts to this cause.  Two days later, according to Mike Isaac in the same publication, Lyft, now adding software maker nuTonomy to its driverless partnerships, while using an approach “different from Uber’s,” is still “behind others.”  Its CEO Logan Green was quoted as saying “we don’t think there will be a single player that will win the whole autonomous vehicle game” – that is correct, but there may well be one consortium winner.

While reading John Markoff’s June 7th “A Guide to Challenges Facing Self-Driving Car Technologists,” also in The New York Times, I wondered if it hadn’t aged for some time before publication.  Still, it named six areas of technical concern, with the more original ones being “the ability to respond to spoken commands or hand signals from law enforcement or highway safety employees” and “detecting which small object on the roadway must be avoided,” which is not routine even for experienced drivers and will take a lot of iterative work, unless they simply avoid everything not flat.  The problem of “driving safely despite unclear lane markings,” in the news recently, though, cannot fall on the roads themselves.  The same day, author, and source yielded “Robot Cars Can’t Count on Us in an Emergency,” which told us that Google’s effort there was now focused on vehicles that cannot be driven – in other words, at a higher level than will first be implemented by others – before discussing “over-trust,” the valid concern that people, when needed to take over driving, won’t be ready.  An Audi A8-version luxury car, with self-driving expressway capability including notification that the driver must take over within a sensible 8 to 10 seconds, will be released next month, and will serve as a good trial subject.

Going back to corporate moves, Fox Business published two pieces, credited to Associated Press on June 13th and Reuters on June 15th respectively, “GM raises output of self-driving Bolts, boosts test fleet” and “Uber’s trucking ambitions on lower gear after Otto deal.”  They are almost self-explanatory, with General Motors claiming the title of the first company to “mass-produce self-driving vehicles” with a run of 130, and, with that, the statement from Cox Automotive senior director of content that this company is “the furthest along” in both their partnerships and their technical progress, followed by Uber’s latest bumbling move, plowing the Otto name under in favor of their main, badly tainted one, and their threat to prematurely implement automated 18-wheelers.  If Uber goes out of business, nobody will be able to say they did it quietly.  The same site offered Matthew Rocco’s “Avis scores Waymo deal, Hertz said to work with Apple on self-driving cars” (June 26th), showing that the major rental companies are also getting involved, including Avis soon offering such vehicles to “select customers in Phoenix.” 

On the legal front, we learned that “Texas explicitly allows driverless car tests” (John Fingas, Yahoo Finance, June 17th), if these vehicles, eminently reasonably, “have to obey existing traffic laws and carry insurance,” record video, and take on liability.  In case we didn’t know already, the “driver killed in Tesla self-driving car crash ignored warnings, NTSB reports” (Nathan Bomey, USA Today, June 20th) – in fact 13 of them, both visual and auditory – “and took no braking, steering or other actions.”  Now, I hope, we can retire discussion about this highly nonrepresentative event, with the lesson that driver-error fatalities will be possible for decades to come. 

The Wall Street Journal published two pieces on the effects of autonomous vehicles.  “Your Next Car May Be a Living Room on Wheels” (Chester Dawson, June 19th) opened our eyes to the massive changes and innovations we will see once internal car spaces need not focus on accommodating drivers.  Beyond storing passengers, we will have the potential to do about anything with them, especially if seat belts become unnecessary, and cars will have the capability to change these spaces from, say, TV-watching dens to conference rooms to bedrooms.  Late in the century, driverless safety may be so improved that we won’t even care about windows, leading to, along with the possibilities of vehicle sizes everywhere from smaller than Smart cars to larger than buses, great car-interior diversity.  The other piece, “The End of Car Ownership” (Tim Higgins, June 20th), offered a previously written misleading idea that cheaper taxis and autonomous vehicles forming “the majority of transportation in urban cities with temperate weather” will make all cars communal, and the expected but already publicized idea that auto companies will offer on-demand transportation packages.  Higgins also mentioned the coming different use of car interiors, but did not seem to recognize that having custom ones will increase, not decrease, ownership.

We got caught up, somewhat, with the government side in two June 25th USA Today pieces.  In “Regulators scramble to stay ahead of self-driving cars,” Nathan Bomey and Thomas Zambito told us that “more than 50 bills have been introduced in 20 states over the last several months providing some degree of regulation,” mostly in the South, Southwest, and across the Canadian border.  The need for states to continue regulating “driver behavior,” and the federal government to maintain standards for “testing” and “crash liability” are clear, but not “design requirements,” and all must be aware that what will be happening with autonomous vehicles in five years will not be the same as in 50.  Overall, governments have, thus far, struck a good midpoint.  In the three-page second, “States get ready for the self-driving car revolution,” author Marco della Cava gave a book’s worth of quotations, propositions and assertions, ranging from the intriguing but debatable “this is as challenging a position as it was when we went from horses to cars” to the misguided “the most critical upgrade amounts to making sure the lines on our 4 million miles of roads are solid” (many of which are gravel, of which, as the article itself points out, 60,000 miles are in Michigan alone), and the opinion that we need “a national vision for autonomous vehicles” (way, way too soon for that, and governments will know the least).  While sensors that communicate traffic and weather information are valuable, it is Sisyphusian for the burden of driverlessness to be put on the external environment.  At the same time, “cities vie to become hubs of self-driving technology” (also June 25th in USA Today), with Austin (highly appropriate in an interesting way, since it would not tolerate Uber), Columbus, Detroit, Nashville, Reno, Phoenix, Pittsburgh (watch out for that Pittsburgh left!), and the Silicon Valley area leading the way.

Last is an attempt at a high-level autonomous vehicle summary, June 25th’s “Transport’s coming upheaval,” from Cyrus Radfar in Yahoo Finance.  It paired driverless technology with Hyperloop, Elon Musk’s “new form of terrestrial travel using pod-like vehicles traveling over 700 mph in near-vacuum tubes,” which, while having potential, is not being pursued as intensely and extensively.  Radfar mainly summarized the progress in the field, but, strangely, concluded that it will not cut employment, thinking, for example, that repurposing some parking lots will be (indefinitely?) labor-intensive.  He gets points, though, for saying that more efficient transportation will in effect expand the coastal areas, which have historically had much of the population and prosperity, further. 


Next week’s post will be dedicated to the June federal jobs report.  On July 14th, you can look for more driverless car reporting, a conclusion, and updates of last year’s projected timelines on its implementation and its employment effects. 

Friday, June 23, 2017

Driverless Vehicles - Where Are We Now? – Part 1

This jobs-related subject not only hasn’t quit for years, but is intensifying, in both action and reporting and commentary.  How can we sort it out to determine when we are getting there?  First, we review what’s happened since our last posts.

In “Uber’s self-driving cars managed 20,000 miles last week – with a lot of help” (Transport, March 19), David Curry presented data leaked from this now thoroughly beleaguered ridesharing company.  In the week of March 8, Uber’s driverless vehicles managed 20,300 miles, but needed a “driver intervention” every 0.8 miles, with takeovers to avoid a “harmful event” 196 times, or once on average each 103 miles.  That’s not shabby at this point, and if that company were in better shape otherwise we’d say they were well on track. 

Six days later, though, per Mike Isaac in The New York Times, Uber was in trouble again, with one of their driverless cars involved in an Arizona accident causing minor injuries.  It was not at fault – the other driver did not properly yield – but the Uber car ended up on its side, and the mishap pointed out the difficulty of programming vehicles to account for others’ errors.  The company suspended their testing, but brought it back three days later after their investigation, per The Wall Street Journal.  Uber still drew commentary that same day asking “Is Uber’s self-driving program veering off track?” (Marco della Cava, USA Today), including concern that “testing in inclement weather is considered the Achilles heel of autonomous vehicles” and compared Uber’s “headlong rush into the self-driving car race” to the legendary Icarus flying too close to the sun.  Here, I think the author was overly influenced by Uber’s legal, ethical, and management problems, as everyone in the industry, almost perforce given the stakes involved, is moving quickly.  That company, though, was ranked 16th of 18 in “Why Ford and GM Are Actually Way Ahead of Tesla and Uber in Developing Self-Driving Cars” (Natalie Walters, The Street, April 6th), a piece which seemed not to recognize that, with consortia forming, Ford, for example, will soon not mean only Ford.

In “Smart cities need smart cars:  In the future, your car will be able to “see” through buildings” (Salon, March 26th), Angelo Young described a demonstration at January’s Consumer Electronics Show of late-development-stage autonomous car features, including electronically determining from stoplights when they will change, detecting pedestrians through their mobile phone’s signalling, and receiving information from nearby cars even out of their line of sight.  These improvements constituting “V2X,” or “vehicle-to-everything,” communications, are superb ideas, but the first two here depend on signals from the environment, which, with consistency problems, is not the way to best implement driverlessness.  That opinion of mine was underscored by another Angelo Young piece in the same publication, “Self-driving cars vs. American roads:  Will infrastructure speed bumps slow down the future of transportation?” (April 20th), which featured a Volvo executive from Scandinavia complaining about the lack of consistent lane markers in this country.  Determining the respective responsibilities of infrastructure and the vehicles themselves is a good conflict to work on now, even if my view that almost everything should fall on the latter does not carry.

They seem to seek involvement in everything else, so should it surprise us that “Amazon Wants to Use Self-Driving Vehicles” (Matthew Rocco, Fox Business, April 24th)?  That company is quietly, by their standards, moving into the field, with its Alexa virtual assistant going into Volkswagen, Ford, and Mercedes-Benz cars, and a special interest in “autonomous forklifts and big rigs.”  Whether it materializes as a producer or only as a massive user, it is as foolish to disregard Amazon here as elsewhere.

We learned from April 25th pieces in Fox Business and The Wall Street Journal that Waymo, part of Google’s Alphabet, will be using 500 Chrysler minivans to test a driverless vehicle program in Phoenix.  Along with putting their current technology to an extensive trial, the effort’s goal is, once again, to gauge customer acceptance of autonomous cars.  Although it is too early to expect any number of people to give up their vehicles, there will be a few, and it is an excellent idea for them to try out that concept as well.  Three weeks later, we saw “Lyft and Waymo Reach Deal to Collaborate on Self-Driving Cars (The New York Times, May 14th), in which Mike Isaac confirmed that “the deal calls for the companies to work together to bring autonomous vehicle technology into the mainstream through pilot projects and product development efforts.”  Isaac points out, presciently, that these agreements have a “fluid nature” – although no large joint venture in this field has yet collapsed, it is almost certain that some will – and that “many believe” driverless technology “will ultimately be a multibillion-dollar industry” (I hope he meant multitrillion).  He added the unintentional comic relief that Uber’s then-CEO Travis Kalanick called autonomous know-how “existential” to Uber’s future – more than he himself proved to be, anyway.


Next week, we look at what has happened with driverless vehicles in June.  Two weeks after that, you can expect the latest projected implementation timelines.  

Friday, June 16, 2017

Robots and Jobs: Three More Months in the Books – Part 2

As we move into April’s stories, the views on automata and employment seem to shift.  “Robotics revolution:  To really help American workers, we should invest in robots,” written by Nikolaus Correll and published in Salon on April 2nd, gave us something on the need for these devices to be integrated with humans’ jobs.  Correll’s main idea is that China, although once taking over American jobs with their own workers, is now leading at automating them, with 2015 purchases of industrial robots over twice the American total.  He also pointed out, correctly, that robots are not only getting cheaper but easier to program, with the latter likely to be increasingly done by their owners, as “building and programming robots is very similar both physically and intellectually to doing your own plumbing, electrical wiring and car maintenance.”

Kriti Sharma’s “You’ll Be Working With Robots Sooner Than You Think” (Fortune, April 10th) dealt more with artificial intelligence (AI) and how it is expected to be used not only to get automata to learn to do their jobs more efficiently but to behave ethically.  Sharma mentioned the possibility, perhaps less fanciful than it seems, “to create a rewards-based learning system that motivates robots and AI to achieve high levels of productivity.”  Indeed, throughout the process of AI getting more robust, we will need to continuously hold off the problem of goal-seeking software lacking humane underpinnings. 
“No, robots won’t put us all on the unemployment line,” Betsy McCaughey’s New York Post story headline said on April 11th.  She fell into the common trap of assuming higher-technology jobs created by the likes of automata, in this case “monitoring or repairing a fleet of delivery robots,” will be as plentiful as those made obsolete, and called those advocating robot taxes Luddites.  Better were her statements that “college is no cure-all,” that “the economy thrives when businesses, not politicians, call the shots on technology,” and that “embracing robots will create more goods and services, a bigger pie for all to share.”  None of these three ideas, however, are inconsistent with a large net loss in employment opportunities.

After the startling but defensible view in the May 3rd Atlantic Daily that 65% of Las Vegas jobs could be automated away within eight years, we move on to Greg Ip’s May 10th Wall Street Journal effort “Robots Aren’t Destroying Enough Jobs.”  The author offers several thought-provoking points, starting with his declaration that “”Robot” is shorthand for any device or algorithm that does what humans once did, from mechanical combines and thermostats to dishwashers and airfare search sites.”  He asserted that “American consumption is gravitating toward goods and services whose production is not easily automated” (not sure at all), and, although he wrote that the number of child-care workers doubled from 1990 to 2010, that could change even if “working parents won’t leave their children in the care of a robot,” if, for example, fewer workers can care for more children through remote or semi-remote monitoring.

“Will Robots Fire Us All?”  Bill Samuelson, this time in the May 10th Investor’s Business Daily, is at it again, and still hanging his I-don’t-think-so attitude on the simply incorrect idea that new technologies consistently create as many positions as they eliminate.  He will be disappointed.

Simon Parkin, in the May 12th 1843 Magazine, made a valiant effort to show how we can succeed at “Teaching Robots Right from Wrong,” but the doubts he brought up demonstrated that it won’t be easy.  First, he mentioned that if hard-coding of core values was implemented a few centuries ago, it might have directed machines to agree with slavery and women’s unequal rights.  Second, as I have written about driverless vehicles, in some situations they face, robots must be told which lives are, in effect, more valuable than others.  Third, while artificial intelligence may allow automata to learn from mistakes, “a killer robot is more likely to be disassembled than offered the chance.”  Fourth, although Parkin discussed the possibilities of learning, for example, the details of how to behave properly before, during, and after a restaurant dinner, those things are remarkably variable and excruciatingly difficult to precisely define, as anyone who has studied ethnomethodology can attest.  I predict that those issues, even after they have been overcome enough with driverless cars for them to be well established on our roads, will still be severe enough to bar many less dedicated machines.

We end with an unusual but exaggerated viewpoint in “Robots Will Save The Economy” (Bret Swanson and Michael Mandel, The Wall Street Journal, May 14th). The authors made the case that our country can use more technology, and they are right, but that won’t help jobs in the ways they mentioned, such as trucks, which are getting ever hardier, requiring more mechanics once they are being self-driven for more miles than with drivers.  We have great things coming up, but such as 3D-printed body organs won’t permanently help employment even if they arrive this century, and the comparison they cited, of 397,000 new e-commerce jobs in ten years with 76,000 in-person retailing ones lost, cannot be a valid one. 


What does this all mean?  As the last article shows, it is tempting to confuse technology’s net value, which is extremely high, with its overall effect on jobs, which is sharply negative.  We need, though, to keep those two ideas separate while maintaining awareness of both.  The first thing will help us tremendously, and we can deal effectively with the second.  That is the real significance of robots.      

Friday, June 9, 2017

Robots and Jobs: Three More Months in the Books – Part 1

There has been no shortage of recommended countermeasures, if you can call them that, to artificial workers put forward since February. 

In “Automated labor apocalypse:  Why a French socialist’s case for taxing robots is better than Bill Gates’ idea” (Salon, February 26th), Kate Aronoff put the ideas of robot levies and a guaranteed income together, and proposed the former pay for the latter.  That’s not a bad rough idea, though, as Lawrence Summers pointed out soon thereafter (“Picking on robots won’t deal with job destruction,” The Washington Post, March 5), we don’t have much agreement on what constitutes those automata, which are, for work-task-replacement purposes, little or no different from “kiosks that dispense airplane boarding passes,” “word-processing programs that accelerate the production of documents,” “mobile banking technologies,” “autonomous vehicles,” or even “vaccines that, by preventing disease, destroy jobs in medicine.”  Although images of almost and increasingly human machines with humanoid appearance can trigger our emotions more than thoughts of NCR cash dispensers, their effects on anthropoid workers are no scarier.

If we see them as opponents, can we use a primer titled “How to Beat the Robots”?  Yes, even if we disagree with that premise.  In this piece, published March 7th in The New York Times, author Claire Cain Miller presented a variety of ideas on how we can, even if we cannot truly overcome the things, keep prospering ourselves.  They include “more education and different kinds” – while Miller’s first suggestions of raising enrollment in universities, stepping up vocational study and “government retraining programs” are all somewhere between misguided and overrated, she did ask for more coursework on “skills that still give humans an edge over machines, like creativity and collaboration” and to find ways for people “to learn flexibility and how to learn new things,” the last being what we could call meta-learning.  She also said that governments could “subsidize private employment,” perhaps, as has been done before, by partially paying compensation, and touched on infrastructure efforts, “strengthening labor unions” (not responsive to the problem of too few jobs), bringing in “advanced manufacturing” (we’re doing that when we can as it is is), subsidies for relocation (thinkable), and a grouping of ways to “bolster the safety net,” ranging from wage insurance to guaranteed income.  Miller’s thorough survey of possibilities also included a section on ways we could “change the way work is done,” including making benefits portable (as we should be and are doing now), “building co-working spaces,” facilitating more small businesses, “reducing licensing requirements” (would help specific people but not create jobs), and the most appropriate idea for quick implementation, bringing workweeks below 40 hours for the first time in, incredibly, 77 years.  The final section, “Give Workers More of the Profits,” provided a brief but even-handed look at higher minimum wages, along with changes to corporate levies, instituting “a minimum pension” as an adjunct to Social Security, and the aforementioned robot taxes.  In all, Miller provided an excellent comprehensive look at what we might do, not only about automation but, by inference, about foreign competition, efficiency, and other factors making our jobs crisis permanent. 

Are we heading for a “looming economic collapse”?  In Fox Business on March 28, one prominent figure said so in “Robots Will Steal U.S. Jobs Warns Billionaire Investor Jeff Greene.”  As gloomy as this view may be, it is sort of refreshing, considering the nonreversing nature of technical progress, to see someone claiming unequivocally that we really do have a problem.  The piece cited PricewaterhouseCoopers research concluding that the next 15 years may see 38% of American work opportunities lost to artificial intelligence alone (high, but believable), and quoted Greene as saying that, during “the next five to seven years,” office positions would be savaged by artificial intelligence and “big data” as well as robots.  He also wrote as I did, that reduced employment is no excuse for our need to “embrace technology,” but ended with the thought that “we have to love it.”  No, we don’t. 

Perhaps, though, the tide in the press is turning.  On March 28th, Claire Cain Miller returned to the Times with “Evidence That Robots Are Winning the Race for American Jobs.”  She cited a paper by Boston University and Massachusetts Institute of Technology economists Pascual Restrepo and Daron Acemoglu, who concluded that the pervasive idea “that increased automation would create new, better jobs, so employment and wages would eventually return to their previous levels” was empirically unfounded, and that many of the displaced workers successfully returning take a long time to do that.  Miller also touched on the idea, put forth by another M.I.T. economist, “that machines will complement instead of replace humans, and cannot replicate human traits like common sense and empathy,” but did not make a key point, that there is no need for empathy in painting, welding, or even analyzing X-rays.   On the same day, Matthew Rozsa reached a similar conclusion in Salon (“Steve Mnuchin is wrong:  Robots really could take your job.”), naming a National Bureau of Economic Research estimate that each automaton replaces an average of 6.2 human workers, and that implementation of each 1,000 such devices was associated with a 0.18-0.34% employment rate drop and an even higher average wage reduction, in contrast to Treasury Secretary Mnuchin’s rather opposite conclusion that it would take “50-100 more years” for the robotics threat to employment to materialize. 

We end with a March 30th Harvard Business Review look, from that publication’s usual big-business vantage point, “How to Win with Automation (Hint:  It’s Not Chasing Efficiency).”  Author Greg Satell called the 20th century “an era of unprecedented prosperity” despite jobs lost to mechanization, but did not mention that being due to the massive widening of the third phase of economic activity, or services.  Although, as he wrote, task automation can lead to task commoditization, many strictly in-person work details are no less so.  Satell named, as high-level strategies, “innovate business models” by setting workers on nonautomatable tasks, “redesign jobs” accordingly, and recognize that “humanity is becoming the scarce resource,” a view which would resonate better if we could quickly absorb fewer than 16.75 million new positions, including almost two-thirds by people not counted as officially jobless.  Still, as with the other articles here, it is healthy for robots to be taken seriously as anti-employment forces. 


More on this topic next week.        

Friday, June 2, 2017

Higher Latent Demand for Work in May: American Job Shortage Number Up 150,000 to 16.75 Million on Soft Employment Report

The federal jobs data took a breather last month.

Although seasonally adjusted employment bettered yet again, from 4.4% to 4.3%, this morning’s Bureau of Labor Statistics report was a weak one. 

The headline number will probably be 138,000, the count of net new nonfarm payroll positions we achieved last month, on pace with population growth but no better, and almost 50,000 less than publicized projections.  Unadjusted joblessness held its 0.5% April improvement at 4.1%, and the number of Americans working part-time for economic reasons, or holding on to less than full-time propositions while seeking, thus far unsuccessfully, full-time ones, added to its last-time 300,000 improvement by dropping another 100,000, to 5.2 million.  There were 6.9 million unemployed, down 200,000 from April.

The other numbers, though, were discouraging.  The two indicators best showing how common it is for our countrymen to actually be working, the employment-population ratio and the labor force participation rate, each fell 0.2%, a lot for one month, to 60.0% and 62.7%.  The number of long-term unemployed, or those officially jobless for 27 weeks or longer, bucked its improvement trend with a 100,000 rise to 1.7 million.  Average hourly private nonfarm payroll earnings were up only 3 cents per hour, less than inflation, to $26.22. 

On the marginal attachment front, the big news was in the “did not search for work in the previous year” and “not available to work now” categories.  These were up 1.53 million and 322,000 respectively, massive increases for one month, and added over 200,000 to latent work demand.  While the number of discouraged workers moved sharply in the other direction from 455,000 to 355,000, that was not enough to stop the American Job Shortage Number or AJSN from increasing 140,000, as follows:



Compared with a year before, though, the AJSN still substantially improved.  The May 2016 number was almost 850,000 higher at 17.6 million, with almost all that difference in official unemployment and the did-not-search categories.  One cause for concern, though, is the almost one million gain in people non-civilian, institutionalized, and unaccounted for, or “off the grid.”  The share of those who would absorb new readily available jobs who are officially unemployed reached another long-time low, at 35.3%. 

Although this month was hardly terrible, it was not good.  We need to continue looking at the numbers above, especially the participation rates.  If June’s data returns to the standards of earlier this year, we can still take credit for doing well, if hardly superbly.  In the meantime, the turtle stood still last month.