Author Archives: Illah R. Nourbakhsh

The Threat of Alien, Mediocre AI

Kevin Kelly has an article in the newest issue of Wired: The Three Breakthroughs That Have Finally Unleashed AI on the World. Thanks to Randy Sargent and David Brooks’ post in the New York Times for making me aware of this excellent piece.  Kelly spends far longer discussing where AI is headed rather than just what the three breakthroughs happen to be.  His characterization of AI is very similar to the one I espouse of robotics: the AI coming right now is not some singularity-borne hyper-conscious entity that bests us all; it is the somewhat useful, only moderately intelligent AI subsystems that will be infused in everything around us, whether in the Net or the physical world:

The AI on the horizon looks more like Amazon Web Services—cheap, reliable, industrial-grade digital smartness running behind everything, and almost invisible except when it blinks off. This common utility will serve you as much IQ as you want but no more than you need.

In a way, you can think of AI as an IQ booster- a turbocharger that will be added to all sorts of products and services swirling around us. But then, it’s not human assistance from a full-fledged NI standing next to us. It is code, with very limited capabilities. Consider how walking amongst robots in the park is simply unlike walking amongst humans, since robots are simply alien- we just don’t know as much about their provenance as we do about humans. Kelly says just the same thing about these AI’s that will commonly deal with us:

The chief virtue of AIs will be their alien intelligence.

He quotes a medical diagnostics founder saying that a child born today will rarely see a (human) doctor for diagnoses in his adult life- that the AI systems will be far better at this. I already have experienced poor human bedside manner in my life. How will our relationship to medicine change in this possible future, where bedside manner is altogether laughable?

One final quote, because I have long argued that robotics will give us a crisis of identity regarding what we consider human. This problem is not just philosophical, but a problem of power relationships: after all, if robots consistently replace ever-greater numbers of physical and cognitive labor, what is it that we will do, as humans, to earn income?  The same problem faces us with AI, as IQ supplied by non-humans constantly changes our relationship to our own sense of what makes humans special, and how we ought to be treating the humans surrounding us.

We’ll spend the next decade—indeed, perhaps the next century—in a permanent identity crisis, constantly asking ourselves what humans are for.

So, as David Brooks points out, all these trends points toward the concentration of knowledge in power in the hands of the few companies with the AI’s in tow. We continue the never-ending slide from human labor to human-invented capital that replaces that very labor. Except now the very labor we consider needs to grow broad, covering ever greater cognitive fields, from medecine and tutoring to friendship and companionship.   I think it’s time to reread Player Piano.

Technology and Underemployment: The Economists Speak

In PBS’ Nova Next, Tim de Chant just published an excellent analysis of technological underemployment called Navigating the Robot Economy. This piece distributes real-world examples with analysis by two excellent economists, David Autor and Erik Brynjolfsson, both of MIT. These economists take somewhat different views, positing a variety of mechanisms justifying just how automation may be a harbinger of displaced human labor, or alternatively how automation may help those (still) employed achieve even greater prosperity. Of course, both views of the future can come true, but then the question this begs is, just how will overall inequality be affected by the dynamics of ever-improving automation? De Chant does an excellent job in bringing together modern-day examples, including those of the auto industry and the medical industry, but also reminding us of the disruptive shifts in employment that the 1800’s saw thanks to the mechanized loom. The article is worth a careful read, but will lead you mostly with the feeling of uncertainty. Even worse, Brynjolfsson makes the convincing point that the next ten years will be far more disruptive than the last ten. And that, coming from a time when the past ten years have seen a total disentangling of company productivity and worker wages, is depressing indeed.

The Best Drone Article Ever, so far

I do not give such praise easily; but Benjamin Wallace-Wells just published a New York Magazine article about drones, and it is the very best in-depth article I have read on drones, ever: Drones and Everything After: the flying, spying, killing machines that are turning humans into superheroes.  I suggest reading the whole thing, and clicking on many of the example links- the videos are both beautiful and horrifying, in all the right, provocative ways.

Of Guns and Printers

Much hay has been made about 3D-printed plastic guns, and along comes a new set of articles about the same folks, this time announcing that a milling machine can make the parts that really do need to be made of metal, as reported here.  Of course, CNC machines have been around for a long, long time; so the idea that this is new is somewhat inaccurate. Yet as the price of automated milling machines comes down over the years, it is true that regulated production of metal parts, such as the receiver in a gun, becomes ever more irrelevant. The Maker movement is a fabulous form of empowerment; but we will always see ramifications that will be less inspiring and more threatening to our own freedom.  This is a one-way street we are walking down.

Mediocracy’s Reality: Economist special issue

The current issue of The Economist contains a special report on Advertising and Technology called Little Brother, edited by Alexandra Suich.  Without a subscription, they will only give you access to the leader article, but I found very useful discussions throughout about the ways in which our behavior has become digital labor, able to be collected, analyzed and shared to provide revenue streams to innumerable shadow companies without our faintest awareness.  A few of my favorite tidbits from this special issue follow; all of them show that we are well on our way to what I call Mediocracy, when our digital and physical trails will, together, turn us ever so gradually into automatons at the call of algorithms with the power to build nearly ideal remote control units for every one of us.

most marketers say they have seen more change in the past two years than in the previous 50.

Richard Edelman, the boss of Edelman, a public-relations firm, describes the media and advertising business as a “mosh pit.” Media companies are producing more content on behalf of advertisers, dubbed “native advertising.” At the same time some advertisers have taken to hiring their own journalists to produce stories, websites and videos.

Facebook and Twitter accumulate heaps of information, including ages, friends and interests, about people who sign up for accounts and spend time on sites. Some of it is collected without users being aware of it. For example, Facebook’s “Like” and Twitter’s “Tweet” buttons on other websites carry a code that enables the social-networking companies to track users’ movements even if they do not click those buttons…

BlueKai, for example, compiles around 1 billion profiles of potential customers around the world, each with an average of 50 attributes.

According to TRUSTE, the 100 most widely used websites are monitored by more than 1,300 firms. Some of these firms share data with other outsiders, an arrangement known as “piggybacking.”

RadiumOne, an advertising-technology company, puts cookies on users, normally unbeknown to them, when they click on a weblink sent by a friend.

PubMatic, a firm that helps publishers sell advertising space in real time, provides some 50-70 data points about users on desktops and around 100 on mobile, including the mobile device’s precise position.

Credit Card companies, including Visa, MasterCard and American Express, all sell anonymised data about their cardholders to advertising companies…American Express has an edge..because it actually issues the card enabling it to put cookies on users when they log in to check their statements and see where else they go online.

Companies are also keen to connect the offline and online worlds. Facebook, for example, has joined with Datalogix, a data provider, to link purchases in both spheres. Acxiom, one of the largest data brokers with expertise in the offline world, recently paid more than $300m to buy LiveRamp, a firm that helps match offline data about customers with online information.

Sometimes advertisers do not use information they have because they do not want to look as though they are spying on customers. “We can do more technologically than we’re permitted to culturally…”

Google and Facebook alone controlled over 47% of all digital advertising in America last year, according to eMarketer, and over 57% of mobile advertising.

Robots and Half-Jobs

Many news stations today have reported on Harvard Business School’s report, An Economy Doing Half its Job, by Michael Porter and Jan Rivkin. The news stories have concentrated mainly on the big picture executive summary of the report, where the authors point out that large and mid-sized firms are recovering in the U.S., but that the individual workers comprising the middle class have been thoroughly left behind.  One issue close to my heart is the state of K-12 education in the U.S., and this report does not make that picture a pretty one. K-12 continues to decline in quality, and what efforts business undertakes to help end up being piecemeal, never really positively affecting the systemic problems with education that we face across the nation. 

Fundamentally, the report is a snapshot of inequality. Some choice passages:

In the lower and middle strata of the income distribution, household incomes have remained stagnant in real terms for decades…Labor force participation in America peaked in 1997 and has now fallen to levels not seen in three decades. Real hourly wages have stalled even among college-educated Americans…Notably, all of these trends began well before the Great Recession. they are structural, not cyclical.

 

But the report does reach conclusions on the question of underemployment as well, in a section labeled An Aversion to Full-time Hires.  The leading paragraph is well worth quoting in its whole:

First, our survey reveals that business leaders in America are reluctant to hire full-time workers. When possible, they prefer instead to invest in technology to perform work, outsource activites to third parties, or hire part-time workers. For instance, 46% of survey respondents strongly or somewhat agreed that their firms’ U.S. operations prefer to invest in technology to perform work rather than hire or retain employees, while only 25% disagreed.

The report goes on to say, preferences for automation, outsourcing, and part-time hires are likely to lead to less skills development. True, the workers who run the automated equipment, the employees of outside vendors, and the part-timers may receive some training, but that is unlikely to offset the skills that are no longer developed in equivalent full-timers.

Many more details are in the report- particularly on page 22- and I recommend you leaf through the report yourself. You can download your own copy here.

 

Food: a mediocracy antidote

In this weekend’s New York Times, David Segal writes Hey Chef: Next Time, Skip the Fennel about Dinner Lab, which creates eating events around the country, mated with an attitude that invites participants to give the chefs no-holds-barred comments regarding the food they have just eaten. I perked up when Dinner Lab, Segal explains, started compiling massive data on eaters’ feedback, presuming that they would be making a bundle of money reselling this behavioral analytics information to mega-restaurant chains. As I sharpened my Mediocracy antennae, however, the surprise ending grabbed me: no chains were interested in the food preferences behavioral data. What we like to eat, it seems, is not digitally trumping the intuition of your local neighborhood chef just yet. As restauranteurs explained, food is a cultural experience. The milieu is just as important as the analytical, dry examination of ingredients and spices. So behavioral analytics does not scale to food just yet, and this reminds us that culture is a phenomenon that we may try to wave away when abstracting away from time and place; but it will raise its head from time to time and mess with our clean, digital data. Horst Rittel famously talked about urban architecture as a Wicked Problem, where the solution to a real architectural challenge is hard to create, hard to detect and impossible to successfully emulate in a new place: local variations in culture trump what we learn from a previous architectural experiment.  Segal’s example gives me hope that, for some stations of invention and analysis, human intuition still reigns as the most accurate judge of human desire.