Category Archives: Robots

Why aren’t geographers talking more about robots?


Robbie the Robot generates 480 pints of whiskey overnight

Why aren’t geographers talking more about robots? This question struck me, paradoxically, as I sat on a panel on robots at the last AAG (see del Casino Forthcoming). While this might seem the last place to have this thought, it was prompted by two things. First, the smiles of slightly startled amusement from people when I told them I was on a robot panel, and second, my co-panelists, who I thought were missing out some important terrain about robots.

Putting aside the no doubt justifiable bemusement that the AAG had a robot discussion, the other topics discussed that day dwelt on sexbots, love dolls, cyborgs and the more-than-human. These are part of, but not the whole story, as Rosi Braidotti’s recent book on posthumanism documents (also putting aside how/if more-than-human is different from post- or transhumanism).

For me, the latter are cultural or philosophical issues, and no matter how pertinent and interesting, they leave aside the political-economic, which is what I’m interested in here. Vinny’s piece (which has just been released by PiHG online first) does partially offer to take up this issue. He does this in the context of a report on social geographies, perhaps meaning that the economic and political are marginal for his piece, which nevertheless remains required reading.

What I mean is quite simply issues around automation, artificial intelligence, and computerization. For me, these point to one thing: algorithmic life. One big part of this is the effect on jobs and wages, and therefore we need to do a better job of integrating tech with geographies of the economy.

Or what I called on the panel “Geographies of neoliberal robots.” Everyone probably has seen a version of this graph:

Looking at some of the economic changes between blacks and whites.

The productivity-wage gap

A version appears in Harvey’s book on neoliberalism. The point is that since the advent of the neoliberal era (say 1970s and early 80s) productivity has not failed to climb, but the amount returned to workers has stayed about the same, creating a productivity-wage gap, which in turn widens income inequalities.

As I said on the panel:

Two explanations are usually offered: “robots ate all the jobs” (people are put out of work by automation), or a deliberate political project by a revanchist capitalist elite (Harvey).

These explanations are not mutually exclusive. What is interesting is that automation and robots may no longer be occurring in only unskilled and repetitive jobs. Research suggests jobs that are more routine and less “cognitive” are the most susceptible to automation. A well-known 2013 study at Oxford Martin School estimated that nearly half (47 percent) of US jobs are at risk of automation. Geographers are not immune:


Source: NPR/Oxford Martin School, Univ. Oxford

Things we can do

    1. Given that we have this listing of job susceptibility, it would be nice to get at least a baseline map of where jobs are at stake. How about a county-by-county map of potential automation? Take all jobs per county and multiply them by the relevant figures in the study. It wouldn’t be perfect but would give us a baseline map.
    2. The PC was Time magazine’s “machine of the year” in 1982. But a one-for-one replacement of a human job with a computer job need not be the most important development in automation or intelligent machines. Rather, production may undergo wholesale reorganization. (Brynjolfsson and McAfee make this point in their recent book The Second Machine Age.) Geographers can contribute to our understanding of this by analyzing which industries are susceptible, and where they are located.
    3. Turning to computerization and automation, I mentioned above that these evidence algorithmic life. What I mean by this is very simple, if you follow Tarleton Gillespie’s definition of the algorithm:

      they are encoded procedures for transforming input data into desired output, based on specified calculations (Gillespie 2014: 167)

      Notice here three useful points: encoding, desire, calculation. An algorithm is that which enables desire to proceed by making (performing) the world as calculative. So it is a capacity-making. Here there would be plenty to look at in terms of uneven geographical outcomes of the work algorithms do in the world, for example on tracking and geosurveillance.

      In fact, Rob Kitchin and his group have just published a useful listing of the ways this occurs. One example likely to be of interest to geographers is automated facial recognition. I really think we need to think “beyond the smartphone” as the only way we are tracked to include ALPR, gait observance, wearable devices/Fitbits/smart watches, and Minority Report style live biometric tracking (face|iris|gait). I document some of these in my piece “Collect it All” as does Leszczynski in her “Geoprivacy” overview.

    4. Beside being part of algorithmic governance, drones (and I include commercial drones especially here as they are predicted to far surpass military drone spending) could be an object of geographical enquiry, or what I call “the drone assemblage.”
    5. Read Vinny’s piece for a more general overview of many aspects of robots and intelligent machines.

      “Where have you been? It’s alright, we know where you’ve been!”–Welcome to the Machine, Pink Floyd

Our algorithmic society: robots!

I’ve been thinking a lot about robots lately. The most immediate reason is that I’m on a panel at a forthcoming conference (the AAG in Chicago). No doubt a somewhat odd topic, at first glance. (I was mentioning this panel to a colleague and he remarked that geography certainly seemed to be a discipline of many interests!).

Anyhow, this isn’t the only reason, because robots are linked with any number of interesting developments and questions, so you can’t or shouldn’t just consider them by themselves. “Robots” can be a synecdoche for automation, computation, wider issues of the changing economy, and so on. Recently Sue Halpern suggested that algorithms are virtual robots.

I’m not sure what direction my fellow panelists will take; perhaps they’ll consider the social implications of interacting with robots, or the role robots can play in our personal everyday lives; for example, how they might substitute for human contact (fembots and love dolls) and if that’s tied to eg falling fertility rates, as my colleague Heidi Nast has suggested. Some of this harkens back to the origins of the term, which you may know about. This was the 1921 play R.U.R. by the Czech writer, Karel Čapek. His brother Josef suggested the word robot, meaning “serf labor” or corvee. There is a long tradition of robots taking over, from Terminator, to (perhaps?) Frankenstein’s monster and the golem. For the pedantic, yes some of these are androids or artificial lifeforms rather than robots, but I don’t think that changes the point.

I think that’s all very interesting but not quite where I want to go. I keep coming back to that thought linking algorithms and robots by Sue Halpern. It’s true of course that algorithms do automated and repetitive labor. It’s true also that algorithms can learn. I have one device like that in my bedroom. Her name is Alexa, and you can buy her from Amazon (I wrote about it here). Alexa learns your voice and improves her response accuracy and relevance. Amazon’s own website uses this all the time. Think of the “if you liked this, then you’ll love this!” recommendations there–some of them leveraging human social input captured through likes, dislikes and “I found this comment useful” clicks.

The issue of interest at the moment is if robots are reshaping society–particularly economic relations. To be clear, whether robots (and if linked or even synonymous with algorithms, then also Big Data and the Internet of Things) will benefit us–and benefit equally.

There’s a well-known graph that describes the increasing divergence between productivity and wages. Here’s the version given in David Harvey’s book a Brief History of Neoliberalism (pdf here):

Fig. 1.6, p. 25 of Harvey

This wage-productivity gap that has opened up since say 1980 or so gets explained through two related but not identical arguments. The first is that robots did it (robots here being technological increases in automation and productivity). There’s lots of evidence that technology has increased productivity, and of course put people out of jobs; which can be measured by labor’s (as opposed to say capital owners’) share of GDP.

For a long time (the argument goes) it was accepted that a rising tide raises all boats, and the rising tide of increased productivity increased overall human well being. But now that technology is becoming more complex, the benefits of robots so to speak are going to those who are “technology-biased” in terms of social capital (well educated, can code, are in jobs that resist automation, etc).

Beyond this, there is an additional development; the benefits are going to a very small group and not just the capital owners. This is often described as winner-takes-all (or most). It’s no longer sufficient to have a good education, be in a white collar job, etc. You need more than this; for example bargaining power as a CEO. But even that might not be enough if capital costs are driven down by robotization.

This is the argument put forward by Erik Brynjolfsson and Andrew McAfee in their book The Second Machine Age. They observe that the share of income going to labor has declined, while profits are up, but that there’s a further gap between “superstars in a field and everyone else” (p. 146). They label this gap “talent-biased technical change.” They note that between 2002 and 2007 the top 1 percent got as much s two-thirds of the profits from the growth in the economy. Looking around, in many industries–including publishing, not usually known for creating extreme wealth–there are superstars pulling away from everyone. This is often due to digital technologies.

Or take the ratio of CEO pay to the average worker, which has increased from 70 in 1990 to more than 300 hundred in 2005–again, according to them, because of information technology. Specifically, such technology allows for more direct, widespread knowledge and control over decision-making. CEOs are more in touch–and more responsible–with the workplace rather than working through chains of assistants. (To be clear, these are not new arguments and Brynjolfsson and McAfee do not present them as such, pointing to work on this from the early 1980s onwards, especially the work of Sherwin Rosen. Inequality per se has been studied for decades, eg Thomas Piketty and Emmanuel Saez’s influential 2003 paper used tax data to look at income inequality [Harvey cites it, among others].) Robert Reich calls this “share-the-scraps” economy rather than the sharing economy. Without redistribution of wealth, there’ll be no one able to afford to buy the “iEverything.”

But Brynjolfsson and McAfee do try to relate all this to technology, robots and automation. Although they are largely technological optimists, believing that benefits take take to come through with innovation, they do deal with “spread” of wealth being unequal. So it’s not that “robots are going to eat all the jobs” to use the common descriptor. Marc Andreesen (inventor of the Mosaic web browser) makes the case clearly against this by arguing that this technology puts the means of production in everyone’s hands:

What that means is that everyone gets access to unlimited information, communication, and education. At the same time, everyone has access to markets, and everyone has the tools to participate in the global market economy.

What the consequences of that will be is less clear, as is the question of whether “everyone” gets equal access to this information tech (presumably Andreesen is wildly in favor of net neutrality [ans: he is and he isn’t!]). But even if they do, if wages are so low you can’t be a consumer, or if the Internet does not unleash vast creativity but rather vast numbers of lolcats and people eating chips on their couch if they don’t have to work… Of course the producer-consumer is intriguing all the same and explains a lot of our work in New Maps for example, but it is not unproblematic.

All of this is very interesting. But what’s the other side if robots aren’t going to eat all the jobs? Well, we’ve already met it in the work of David Harvey and authors such as Jacob Hacker and Paul Pierson, who argued in their book Winner Takes All Politics that it was politics that ate all the jobs. Or, for them, US politics. Or for Harvey, neoliberalism as a deliberate political project. Hacker and Pierson go through some of the same inequalities but hardly mention technology, except to dismiss it as the cause. What they refer to as skill-biased technological change (SBTC) supposedly has led to wealth concentrations among those who are well educated and technologically adapted. Except, they argue, it hasn’t. The people at the very top are no better educated than those they’re (rapidly) pulling away from.

Second, they argue that if SBTC occurred in America it should occur elsewhere too, especially where places are more connected to the Internet or have tech transitions. But they argue, against some of the people cited above, that it hasn’t happened elsewhere:

there is more inequality among workers with the same level of skills (measured by age, education, and literacy) in the United States than there is among all workers in some of the more equal rich nations

They provide the following graph to illustrate how different is the US case:

Instead of robots then, Hacker and Pierson place the blame on American politics.

For Harvey, as you probably know, neoliberalism is an explicit political project aimed at restoring and ensuring the wealth of the few at the expense of the many. He makes some remarks about technology in Brief History, but devotes rather more to it in his book Seventeen Contradictions and the End of Capitalism (which will be an author meets the critics session at AAG with my colleague Sue Roberts). Harvey’s “contradiction” regarding technology has already been mentioned, namely that technology gives capital controls over labor. “Robots do not (except in science fiction accounts) complain, answer back, sue, get sick, go slow…” Yet the replacement of social labor with robots “makes no sense, either politically or economically” (p. 104). This is his contradiction.

This is a contradiction for the same reason Reich points to: there’ll be nobody to buy all these new products which result from the productivity. It will have “catastrophic effects” says Harvey. And robots won’t only eat entry-level jobs but also high-paying skilled jobs (he cites university professors).

Finally, let me say a few words about drones or unmanned aerial systems since I suspect that might be why I’m on the panel. Sue and I are putting the finishing touches to a long-ish paper on commercial and civil drones, and one of the points we touch on is that drones may represent the kind of technology that is usually seen as especially impactful/disruptive since it is general and applicable in many different sectors. Harvey discusses this. The estimates for the market size are truly staggering: one market research firm, the Teal Group, has estimated it to be worth some $90B in the next ten years, globally. This will eclipse the amount spent on military drones. But drones are not being deployed in the US early and often: Sue uses the term the “post-permissive” age to describe how the skies are no longer easy to fly in as they were over Iraq and Afghanistan, but are contested spaces. In the commercial sector too we are seeing a contested and struggled over environment, what with regulation, competition, and the need to create a market for UASs where one does not exist (what are they for? How can I make money?). On top of that is public distrust, the surveillant capabilities that are being resisted, and so on.

Drones may therefore be a fairly unusual kind of robot due to this general application nature. We’ll definitely have to see where they develop and under what conditions (and with what geo technologies) but as they get smaller and more autonomous they may be pretty significant.