The AI Century

The natural selection of economics has relentlessly pushed businesses and societies toward maximum efficiency.  It’s the principal force that fuels the start-up business models of today as they redefine and make more efficient the economics of tomorrow.  It’s this drive to maximum efficiency that’s given rise to disruptive, AI-driven business models that include Amazon, Google and Uber.

While business models have been made more efficient by AI – ruthlessly so – its broader societal impact is now unfolding in parallel.  AI is such a powerful force that it threatens to disrupt the foundation of the modern economy: employment. This impact calls for a re-examination and re-imagination of what constitutes efficient and sustainable business models for the future.  We’ve entered the AI Century and must define how we’ll actively coexist with AI lest our jobs be replaced by it.

 

Open the Pod Bay Doors, HAL

It’s not just techno-phobic Luddites who fear the impact of AI-driven automation on the job market.  Stephen Hawking worried AI may become “a new form of life that will outperform humans”.  Its negative effects on employment have been warned by UC Berkeley’s David Patterson and Stanford’s John Hennessy, joint winners of the 2018 Turing Award.   PwC’s Richard Berriman wrote in March 2017 that up to 38% of US jobs could be at high risk of automation by the early 2030s, with particularly heavy impacts in the fields of transportation and storage (56%), manufacturing (46%) and wholesale and retail sales (44%).

Similarly, McKinsey Global Institute’s November 2017 report stated that 60% of today’s occupations have at least 30% of constituent work activities that could be automated, and “that between 400 million and 800 million individuals could be displaced by automation and need to find new jobs by 2030 around the world”.  These numbers align with the earlier conclusions of Oxford University’s Carl Frey and Michael Osborne, who had found that computerization puts “about 47% of total US employment at risk”.

There are, of course, voices of optimism on the impact of automation.  John Maynard Keynes wrote in 1930 of the leisure and abundance that science would win for mankind.  But he also warned of a companion “technological unemployment” that would arise.  Technological unemployment is a product of “task encroachment”.  I.e., the steady encroachment of automation-driven capital into job functions previously considered within the unassailable realm of human capital.

Automation through AI is proving itself adept at encroaching into more and more human endeavors every day.  In addition to self-driving vehicles and trouncing humans at Go, AI is making its impact felt in areas as diverse as talk-therapy, assembling Ikea furniture, and creating original images.  As Elon Musk has said, we can now assume that there’s no limit to AI’s capabilities.

What held true in 1983, Nobel Prize-winning economist Wassily Leontief’s lament that, “More and more workers will be replaced by machines.  I do not see that the new industries can employ everybody who wants a job”, holds doubly true today.  Given AI’s relentless advance, our best-case outcome might be the “Robocalypse Now” scenario laid out by economists David Autor of MIT, and Anna Salomons of Utrecht University: an automation-driven polarization of job supply into high-skill (the jobs that control the automation) and low-skill (the jobs displaced by automation) categories.

Within Robocalypse, the “challenge is not quantity of jobs”, but the “quality of jobs available to low- and medium-skill workers”.  Bank of England chief economist, Andy Haldane, had reached a similar conclusion for automation leading to “a hollowing-out of employment, a widening distribution of wages and a fall in labor’s income share” resulting in “a permanent re-shaping of the labor landscape”.  Or, as Oxford economist Daniel Susskind wrote, “a remorseless displacement of [human] labor, a continual fall in absolute wages”, culminating in a pessimistic future where “labor is fully immiserated”.

Ouch.

If you were optimistic, the bad news is in.  AI-driven automation will lead to either a) income distribution that’s even more polarized than today; b) mass unemployment; or c) both.  In short, we’re screwed.  Unless we do something.

 

Steam Drills & Dark Satanic Mills

Technological advancement isn’t a series of random spikes, but an increasing and accelerating function.  AI’s capabilities continue to grow, and scaling it to more tasks is gated in practice only by the availability of (ever cheaper) computing cycles.  If machines prove more cost-effective than humans, then it’s only a matter of time before economics gets its way and humans are rendered obsolete in more and more jobs.

Productivity is measured as economic output divided by the inputs required.  If automation yields high productivity in ever more areas, but with the need for ever fewer humans, then our economy will be exceedingly productive, but with no prosperity.  Ferdinand Lassalle’s Iron Law of Wages had placed a floor on wages at the level required for human subsistence.  What happens when the wage floor asymptotically approaches zero as a consequence of Moore’s Law?

There’s unfortunately no rule that says that technological advancement always improves life on the planet.  There should therefore be no surprise concerning the malign impacts that AI may have on the job market; and job growth will likely not rebound post the AI revolution.

Unlike past technological revolutions, the AI revolution is different because humans are eliminated from the economic value chain.  We are neither the brains nor the brawn powering autonomous, AI-enabled systems.  With jobs increasingly eliminated, the inevitable fate of commoditized humans seems to be servitude in the monopsony of the gig economy.

 

Voltaire & FDR: Both Right

“Demoralization caused by vast unemployment is our greatest extravagance.  Morally, it is the greatest menace to our social order.”

Franklin Delano Roosevelt, Fireside Chat, September 30, 1934

Robots don’t eat at local restaurants, shop at the mall, or shop online. They don’t watch TV, borrow money, pay taxes, or insurance premiums.  Humans do.  If unemployed or underemployed people represent the bulk of the buying public, then tax receipts and our consumer-driven economy will decay in unison.

One remediating proposal that’s been made is to tax the companies that benefit from robots.  The proceeds from this taxation would be applied to helping those displaced by automation.  Another proposal has been the provision of universal basic income (UBI).  UBI may have promise, but it’s still early to know how or even whether this will work.  UBI may have the effect of either giving people so little money that it’s meaningless, or so much money that it eliminates any incentive to get educated, work, or be engaged with your fellow citizens.  We don’t know the long-term effects on families and societies of “giving money for nothing”, and these effects may be profound.

Voltaire said that, “Work keeps at bay three great evils: boredom, vice and need”.  Though UBI might address the monetary needs that employment currently fulfills, it will do nothing to keep us from the first two evils.  As members of human society, we need to be engaged in it.  It’s perhaps for this reason that Franklin Delano Roosevelt, the US president during the Great Depression, engineered the New Deal to include the Civil Works Administration and the Works Progress Administration.  Money would be meted out to the unemployed, but work that served society would be expected in return.  Universal basic income was tied to universal basic employment.

Tying UBI explicitly to public-serving work may ultimately be a necessity.  Simply redistributing wealth via UBI will not make unemployed workers engaged members of society, make them productive, or give them skills (or identity) that will make them employable ever again.

 

Iron Man Beats the Terminator

If the traditional model of employment becomes obsolete, where does the future lie?  We can’t passively stand by and await the defining of a new jobs landscape by the natural selection of economics.  We have to proactively define what these new jobs will look like.  Otherwise, AI-based disruption will eliminate too many jobs; if humans aren’t explicitly factored into the jobs equation, then the invisible hand of economics joining with the unrestrained hand of AI will ensure that we’re implicitly factored out of it.

The Terminator robot, introduced in the 1984 movie of the same name, was a fully-automated assassin.  The 1963 comic book hero, Iron Man, was, by contrast, a human (a tech entrepreneur, no less) augmented by technology.  Within this difference lies the opportunity for innovation, business models, start-up opportunities and, most importantly, the jobs of the future.  Iron Man is our best hope.

“Iron Man” innovation –  i.e., technology innovation that augments, not replaces, humans – requires a fundamental redefinition of the word “productivity”: it can no longer be founded simply on reducing costs.  Human employment is not sustainable within this calculus.  Instead, in the “new productivity”, humans need to become part of the innovation equation, with business models that place human beings at the heart, tying mind together with machine.  If productivity remains defined solely by the bottom-line then the cold efficiency of the fully-automated Terminator will replace us in job after job.

Although AI-driven vehicles may provide benefit where safety or environmental impact are concerned, they also carry the potential for eliminating a few million US jobs with no clear prospect for re-employment for those affected.  The same displacement might occur for hundreds of thousands of agricultural field workers due to automated harvesting.

Technology exists to make life better.  If it doesn’t, then we shouldn’t pursue it.  In the 20th Century we learned that the blind pursuit of economic efficiency (cheap products!) can result in the greater cost of environmental damage.  Perhaps it’s now time to frame the impact of automation on humans in similar terms.  Perhaps the next generation of technology start-up needs to be explicitly human-centric.

 

Disrupting AI’s Disruption

Iron Man innovation requires a focus on mankind’s foremost challenges, and not just on the ruthless math of output divided by cost.  No society on Earth is over-resourced in education, healthcare, infrastructure or renewable energy.  How do we now drive new combinations of technology with human skills – including intuition, communication, emotion and empathy – to design next-generation start-ups to attack these areas of need?  How do we deploy Iron Man on a grand scale?

For example, as start-up entrepreneurs and investors, our focus is better placed on impactful Iron Man opportunities in healthcare – AI augmenting healthcare workers in dermatology, cardiology, radiology or surgery – than on Terminator solutions for call centers, flipping burgers or restocking store shelves; just because we can do the latter doesn’t mean we should.  If “augmenting” Iron Man start-ups can be successful in the industrialized world, the positive impacts may be even more strongly felt in developing countries with pools of lower-skilled workers at risk of displacement by Terminator automation.

There are those who will point out that we’ve experienced industrial revolutions in the past and come through those stronger than ever.  But technology itself is the wild-card factor here.  By way of example, we survived two world wars utilizing every weapon that technology afforded us, emerging more prosperous than ever before.  Because of technology’s advance, life on the planet, let alone human civilization, certainly won’t survive a third world war where we utilize every weapon that technology affords us.  So just as we can’t look to past history as a guide in warfare, nor can we do so in economics.  Technology has become too powerful and disruptive a force.

 

It Takes a Village

I’m not an economist, nor am I social scientist.  Colleagues from my start-up days will probably argue (rightly!) that I wasn’t even that good of a computer scientist.  But I call on the most thoughtful from all constituencies – enterprises, start-ups, working people, investors, academics, government –  to work toward a new definition of “productivity” for the AI Century, one that explicitly incorporates the opportunity of jobs and prosperity for all, and not just the T-800 series of robots.

We don’t yet know what the Iron Man technologies and jobs of the future will look like.  Crucially, we do have a little time before us: AI technology is still brittle at its edges, and there are considerable cost impediments in rolling it out. We must begin the Iron Man conversation now, before we pass automation’s tipping point.  Revolutionary and disruptive it may all sound, but to echo Margaret Thatcher, “the heresies of one period [become], as they always do, the orthodoxies of the next.”

VCs and entrepreneurs stand at the source of innovation.  If change is to be effected, it can begin with us.  We pride ourselves on seeing things before others and building bold business models that anticipate the future.  Iron Man gives us our next great business opportunity.

Posted in Applying AI and tagged , , ,