Recent Posts

Sunday, January 24, 2016

Book Review: Rise Of The Robots


Martin Ford's Rise of the Robots: Technology and the Threat of a Jobless Future was published last year. It describes the technological forces that lead to job destruction, including white collar jobs that previously were viewed as being technology-proof. Although his views would be described as pro-free market (although he steps on some libertarian orthodoxy), he argues in favour of a Basic Income guarantee.


Book Description


The book was published in 2015, by Basic Books. Martin Ford was a founder of a Silicon Valley software development firm, and previously wrote The Lights in the Tunnel: Automation, Accelerating Technology, and the Economy of the Future.

A Silicon Valley View Of The World

The book is filled with the glorification of technology that is the trademark of the Silicon Valley culture. He even steps lightly on the ground of the some of the wackier cult beliefs that float around there, such as The Singularity.

I would note that although my background is in control systems engineering, which included robotics as a field of application, I am much less convinced about the importance of recent leaps in information technology. For example, he spends considerable space discussing the relentless shrinkage of microprocessors ("Moore's Law") , and implies that this tells us something about their usefulness. For most users, the bloat of the operating system and office applications has consumed whatever gains we got from increased chip speed. Obviously, there are application areas that did not exist in previous decades, but it is unclear whether the ability to take photos of your supper and share them on the internet represents tangible progress.

How Many Jobs At Risk?

The book is written in a journalistic style, and does not delve too deeply into the number of jobs that are at risk. He cites some academic studies on the matter, but the reader is not given details. Instead, he lists many examples of how technology can replace jobs that were previously seen as safe. In my view, we cannot look at just the job title, we also need to look at the size of the firm. Large firms may have many employees doing similar tasks of a limited scope; small firms seem to be less likely to have such specialisation. They can outsource some tasks which are outside their core competency, which allows for economies of scale at the outsourcing firm, but what can be outsourced is limited.

For example, he accepts the marketing hype of a manufacturer of a robot that cooks "gourmet hamburgers" (page 12), and then implies that millions of jobs in the fast food industry are at risk. However, the scope of applicability is much smaller than the robot manufacturer suggests. Making a burger, gourmet or not, is a trivial task for a short order cook (based on my limited experience of working as one in high school). A new gadget that prepares burgers for you does very little; preparation work (which takes the most time) is done outside of peak order times when cooks are otherwise unoccupied. Furthermore, you have a minimum crew size; you cannot shut your restaurant operations for the day because one employee called in sick. The only place where jobs could be cut are franchise operations which have a limited menu and a large enough burger order flow that an incremental pickup in efficiency would allow them to trim a cook or two per shift. However, a fast food chain restaurant is already hyper-efficient at producing its limited menu choices. Most of the job losses that will result from economies of scale will already been incurred when the franchise started operation.

Of the examples he cites, the ones that seem the most worrisome in terms of job losses revolved around document and image recognition software. Medical personnel whose speciality is image analysis, and para-legals who search documents for records that are pertinent for a legal case can have their tasks replaced by software. The job market for law graduates has been notoriously weak, and this could be one of the factors behind that.

The need to go beyond the estimates provided by start up firms to see the impact of technology changes means that it is hard to have much confidence about any job loss forecasts.

Business Shifts Or Technology Shifts?

In my view, technology stories are red herrings distracting us from shifts in industry structure.
  • The rise of algorithmic trading (page 113) is just a replacement of market-making operations by brokers by hedge funds. Since their role is to skim profits from the trading of traditional investors, they cannot grow indefinitely at the expense of those other funds. Since market making was not a mass employer to begin with, the net effect on employment was probably negligible. Elsewhere in investment finance, whatever gains improved back office software provided were offset by the increased need to analyse complex new financial products, leading to limited attrition of employment.
  • There are a number of "sharing economy" companies that are superficially technology companies. However, the secret sauce behind their business plans tend to revolve around regulatory arbitrage (to be polite). For example, some of the "taskification" (as described in this op-ed by Mary L. Gray) of work relies upon contractors not being classified as workers.
  • The rise of internet based booksellers (one of which I have an affiliate relationship with) is not the sole reason of the demise of mom-and-pop bookstores; most of those stores were run out of business by the early 1990s by chain bookstores using pre-World Wide Web technology.

Inequality Or Robots?

Ford gives a non-specialist overview of economic thinking on jobs and technology. He notes that economists generally dismissed these worries, arguing that increased productivity makes everyone better off. Instead, the most fashionable worry at present is that we will not have enough workers due to an ageing population. However, his argument is that disappointing growth and job creation now is a sign that we could be moving towards a "jobless future."

Although he is in the free market economic camp, his prescription why this jobless future is possible would not pass ideological tests. He presents what are now standard arguments about inequality slowing growth. In his view, inequality is driven by technology. For example, he complains that the word "robot" only appears once in Piketty's almost 700 page work on inequality.

However, I do not see a cast iron linkage between inequality and technology; that is, if inequality is the problem, we do not need to reference robots (like Piketty). Economic outcomes reflect policy preferences, and we have seen inequality rise and fall based on changing economic frameworks. I see no reason to fear "techno feudalism" in particular; societies managed to be feudal without the aid of microprocessors.

Basic Income As The Solution

From the perspective of watching intellectual trends, the most interesting part of the book is the suggestion that a guaranteed income should be implemented. The advantage of a basic income from the perspective of free market proponents is that it allows them to dismantle the "bureaucratic" welfare state. However, Ford's worry about economic incentives leaves him to suggest that conditions or incentives be built into the plan (for example, extra payments for those taking entering educational programmes) would bring back bureaucratic meddling through the back door.

The obvious problem with basic income schemes is that non-trivial payments require "high" marginal income tax rates in order to avoid a massive over-stimulation of demand.  (Yes, aggregate demand is weak enough at present to support some fiscal stimulus without posing inflationary dangers, but a $10,000 payment to all adults would blow the existing demand deficiency out of the water.) As Ford observes, about the only reason such a proposal might be feasible in the Unites States is that it lacks a Value-Added Tax (VAT). The idea being that it might be easier to bring in the new VAT than it would be to crank up income taxes to "high" rates. Other developed countries have a VAT already, and so rates on existing taxes would have to be raised.

The alternative of government programmes of direct job creation was ignored within the book. This could be viewed as an example of the general lack of traction behind the Job Guarantee concept.

[Update] The Robots Are Coming For Your Interest Rates!

In a similar vein, see this Bloomberg article on how robots are allegedly going to keep interest rates down. It's based on discussions at the latest Davos summit, and provides yet another example of automation can replace "high level" jobs. With current technology, it should be straightforward to develop an AI routine to write the self-interested flimsy analysis that is the hallmark of the Davos summit.

The whole thesis revolves around:
The argument goes like this: As machines become more and more advanced, many workers will lose their jobs and others will see their wages fall. 
That is, jobs will be destroyed by the "machines," and not the CEO class that is congregated at Davos. We had a major boost in productivity in the early post-war era, and jobs were created, not lost. The reason is that we had a macro framework that supported job creation then; whereas we now have a framework that favours the destruction of jobs (in the developed countries). Even if technological progress stops dead, we are stuck with an economic structure that favours slow growth.

Concluding Remarks

This book is journalistic, and not a work of economic theory. The book provides a lot of background on how businesses can drive down their wage costs, and the wealth of anecdotes indicates the breadth of the issue. My feeling is that these opportunities are mainly available to relatively large firms. It is also interesting in that it shows how interest in a basic income guarantee is spreading across different groups. I have deep misgivings about such plans, but they represent one of the few policies on the radar screen that present a chance of changing the structure of the economy.

Finally, the book is available at Amazon: Rise of the Robots: Technology and the Threat of a Jobless Future (affiliate link).


(c) Brian Romanchuk 2015

15 comments:

  1. "It is also interesting in that it shows how interest in a basic income guarantee is spreading across different groups."

    It falls into the neat, plausible and wrong category. Like Debt Jubilees and all the other quick fixes that make good stump speeches.

    It reminds me of endless "We'll just outsource it to India" discussions I've had with over-promoted executives.

    Nobody wants to hear the downsides when they are in the grip of religious fervour. But there is plenty of money to be made picking up the pieces afterwards.

    Does the book mention automated car technology? That seems to me to be the one disruptive tech on the horizon - since it will impact all forms of public transit, particularly taxi drivers.

    ReplyDelete
    Replies
    1. I view the income guarantee as a trap by conservatives. It allows for the rest of the welfare state to be dismantled, and then the income guarantee is tied to an unpopular tax. This will allow for the gradual elimination of the income guarantee via attrition. Since Ford was offering this idea in good faith, I did not want to bring this up.

      Yes, he had a big section on driverless cars. The point about taxi drivers is plausible. However, he had a long discussion of the potential elimination of private cars, replaced by fleets of cars that are supplied on demand. That transition seemed too unlikely on any reasonable time frame. There is also the problem that autonomous cars are unlikely to be able to drive on snow and ice and time soon, and that it may take considerable time for a vehicle to be operable without a driver being able to step in. I dropped my discussion of that topic, since I found it was too speculative.

      Delete
    2. MCity is designed to help the cars deal with the weather. Ford seem to be further on top of the weather issue than the chaps in Mountain View.

      http://www.wired.com/2016/01/the-clever-way-fords-self-driving-cars-navigate-in-snow/

      Delete
    3. I looked at it, and the key is that they have a high fidelity map of the terrain. I doubt those maps will have the location of all the potholes on our local roads. Driving on sketchy roads in a snow storm with the potential for black ice is inherently tough. However, since the algorithms are biased towards defensive driving, they might do better than people.

      I see the problem as eventually feasible, but I not sure about the delivery timeline. On long forecast horizons, I am more concerned about what exactly those cars are powered by than who is driving them.

      Delete
    4. One of the problems with predictions based around IT is the inherent difficulty humans have thinking in exponential terms. This idea is best illustrated by an example Kurzeil likes to recount about the Human genome project. When they were at 1% completion after like 5 years of work people were concerned thinking about the process on a linear scale 5 x 1 = 500 years!!!@!!!!!!!!!

      When in reality it turned out that they were doubling the pace of genetic translating every year. So while it may have taken awhile to get to 1%, thats only 7 doublings (7 years compared to 500) away from 100%.

      Being a layman wrt to computer coding, I imagine there might be something similar going on with the accuracy of driverless car software as computational power and sensor sensitivity increase even though I have no idea what that might be.

      So I guess this comment was just a roundabout way of asking.....is the capability of driverless vehicles growing on an exponential or linear scale? And the answer to that question has dramatic consequences for the progress of this technology

      Delete
    5. My feeling is that these things depend upon the problem. Progress in an area can explode if there is a new technique, or else you hit a theoretical wall and stagnate. The advantage of these sorts of applications is that you can approach the problem from a number of ways; if you hit a wall in one, you might be able to crack it in alternative fashion.

      I am not up to speed with the progress in this area. Maybe they are on the verge of a breakthrough. But at the same time, the electric car doors in my old car stopped working properly after a year or two. I am not going to be an early adopter of the technology, if you catch my drift.

      I think the rollout process would be staged, and that the initial employment impact would necessarily be limited. For example, even if we have autonomous trucks, you would probably still have a "driver" for cargo security reasons. Something like a taxi service would probably have the biggest immediate employment impact. However, the taxis would have to be totally autonomous (existing cars typically require drivers to be ready to take over), since the taxi could be hailed by a non-driver. If the car requires intervention, the only thing that would work is a ride-sharing operation, where only registered drivers can use the vehicle.

      Delete
    6. Regarding technologies for driven and self-driving cars here is a 44 page pdf outline of basic technologies:

      http://ritzel.siu.edu/courses/302s/vehicle/VehicleSafetyFeatures.pdf

      It appears that most of the problems for self-driving cars are being solved independently in parallel. In heavy industry (to control large loads in trains, crains, etc.) a diesel engine drives a generator, the generator makes electricity, the electricity drives a motor, and the motor controls the load, so I assume traction control for electric vehicles will be better than traditional vehicles.

      This article on the future of computing power expresses broad vision for machine to machine (M2M) otherwise known at Internet of Things:

      http://bigthink.com/dr-kakus-universe/the-future-of-computing-power-fast-cheap-and-invisible

      Personally I expect to see fast encryption enabling security for the smart grid integrating distributed energy resources (DER) at the core of M2M. Models for everything complex should improve such as weather and financial systems, however, M2M will also drive complexity so human systems become more non-linear and more like weather patterns too.

      I like the idea of a Job Guarantee with local business and government helping determine job placement and federal government paying the bills. I wonder if the internet of things could help solve the complex distributed problem of inflation via smart spending and taxation?

      Delete
  2. " I am not going to be an early adopter of the technology, if you catch my drift."

    Totally with you on that one.

    Whats possible with the 100X more computing power we'll have in 10 years? Very difficult to say.

    ReplyDelete
  3. Auburn, computing power is not holding up progress, is actual human resources: aka (quality) programmer time. Scalability of software and increasing complexity are always the issue. Why do you think banks still use COBOL (apart of the efficiency on fixed arithmetic calculus)?

    The "technical debt" the world is building up on software is HUGE, with billions of lines of code hard to replace. Increasingly complex software (which is what automation is) further adds to the difficulty of management of the "global code base".

    My take on 'robots' and automation is not actually robotics per se. Dexterity is much harder to achieve than people realizes, industrial production chains automation has required over decades of standards and procedures, robots solving issues in 'chaotic' environments like we humans do on a daily basis is a super complex issue.

    But there is a lot of space to automate away a lot of mid-tier white collar jobs: those which depend on data analysis and decision making. Increasing ways for pattern analysis and matching, data analysis (in this sense, your typical desktop computer will be able to solve most middle corporation needs, hence computation power not being the bottleneck right now) and decision making based on parameters is something that inherently can be done better by machines. There will be a need for programmers and engineers to adjust the algorithms depending on the requirements of research and management and for maintenance, but it will replace many 'analysts' jobs. Same for many administrative tasks which are on the line for automation.


    Re. cars: the biggest issues are actually legal and ethical (decision making by the driving software, legal implications in case of accident). In relatively temperate climates technical implementation and production should be already possible, the technology is already there (although yes, early adopter problems will happen ofc).

    ReplyDelete
  4. This comment has been removed by a blog administrator.

    ReplyDelete
  5. This comment has been removed by a blog administrator.

    ReplyDelete
  6. This comment has been removed by a blog administrator.

    ReplyDelete
  7. This comment has been removed by a blog administrator.

    ReplyDelete
  8. This comment has been removed by a blog administrator.

    ReplyDelete
  9. This comment has been removed by a blog administrator.

    ReplyDelete

Note: Posts are manually moderated, with a varying delay. Some disappear.

The comment section here is largely dead. My Substack or Twitter are better places to have a conversation.

Given that this is largely a backup way to reach me, I am going to reject posts that annoy me. Please post lengthy essays elsewhere.