AlphaGo and the Future of Work

In March of this year, Google DeepMind’s computer program AlphaGo defeated world Go champion Lee Sedol. This was hailed as a great triumph of artificial intelligence and signaled to many the beginning of the new age when machines take over. I believe this is true but the real lesson of AlphaGo’s win is not how great machine learning algorithms are but how suboptimal human Go players are. Experts believed that machines would not be able to defeat humans at Go for a long time because the number of possible games is astronomically large, \sim 250^{150} moves, in contrast to chess with a paltry \sim 35^{80} moves. Additionally, unlike chess, it is not clear what is a good position and who is winning during intermediate stages of a game. Thus, any direct enumeration and evaluation of possible next moves as chess computers do, like IBM’s Deep Blue that defeated Gary Kasparov, seemed to be impossible. It was thought that humans had some sort of inimitable intuition to play Go that machines were decades away from emulating. It turns out that this was wrong. It took remarkably little training for AlphaGo to defeat a human. All the algorithms used were fairly standard – supervised and reinforcement backpropagation learning in multi-layer neural networks1. DeepMind just put them together in a clever way and had the (in retrospect appropriate) audacity to try.

The take home message of AlphaGo’s success is that humans are very, very far away from being optimal at playing Go. Uncharitably, we simply stink at Go. However, this probably also means that we stink at almost everything we do. Machines are going to take over our jobs not because they are sublimely awesome but because we are stupendously inept. It is like the old joke about two hikers encountering a bear and one starts to put on running shoes. The other hiker says: “Why are you doing that? You can’t outrun a bear.” to which she replies, “I only need to outrun you!” In fact, the more difficult a job seems to be for humans to perform, the easier it will be for a machine to do better. This was noticed a long time ago in AI research and called Moravec’s Paradox. Tasks that require a lot of high level abstract thinking like chess or predicting what movie you will like are easy for computers to do while seemingly trivial tasks that a child can do like folding laundry or getting a cookie out of a jar on an unreachable shelf is really hard. Thus high paying professions in medicine, accounting, finance, and law could be replaced by machines sooner than lower paying ones in lawn care and house cleaning.

There are those who are not worried about a future of mass unemployment because they believe people will just shift to other professions. They point out that a century ago a majority of Americans worked in agriculture and now the sector comprises of less than 2 percent of the population. The jobs that were lost to technology were replaced by ones that didn’t exist before. I think this might be true but in the future not everyone will be a software engineer or a media star or a CEO of her own company of robot employees. The increase in productivity provided by machines ensures this. When the marginal cost of production goes to zero (i.e. cost to make one more item), as it is for software or recorded media now, the whole supply-demand curve is upended. There is infinite supply for any amount of demand so the only way to make money is to increase demand.

The rate-limiting step for demand is the attention span of humans. In a single day, a person can at most attend to a few hundred independent tasks such as thinking, reading, writing, walking, cooking, eating, driving, exercising, or consuming entertainment. I can stream any movie I want now and I only watch at most twenty a year, and almost all of them on long haul flights. My 3 year old can watch the same Wild Kratts episode (great children’s show about animals) ten times in a row without getting bored. Even though everyone could be a video or music star on YouTube, superstars such as Beyoncé and Adele are viewed much more than anyone else. Even with infinite choice, we tend to do what are peers do. Thus, for a population of ten billion people, I doubt there can be more than a few million that can make a decent living as a media star with our current economic model. The same goes for writers. This will also generalize to manufactured goods. Toasters and coffee makers essentially cost nothing compared to three decades ago, and I will only buy one every few years if that. Robots will only make things cheaper and I doubt there will be a billion brands of TV’s or toasters. Most likely, a few companies will dominate the market as they do now. Even, if we could optimistically assume that a tenth of the population could be engaged in producing goods and services necessary for keeping the world functioning that still leaves the rest with little to do.

Even much of what scientists do could eventually be replaced by machines. Biology labs could consist of a principle investigator and robot technicians. Although it seems like science is endless, the amount of new science required for sustaining the modern world could diminish. We could eventually have an understanding of biology sufficient to treat most diseases and injuries and develop truly sustainable energy technologies. In this case, machines could be tasked to keep the modern world up and running with little need of input from us. Science would mostly be devoted to abstract and esoteric concerns.

Thus, I believe the future for humankind is in low productivity occupations – basically a return to pre-industrial endeavors like small plot farming, blacksmithing, carpentry, painting, dancing, and pottery making, with an economic system in place to adequately live off of this labor. Machines can provide us with the necessities of life while we engage in a simulated 18th century world but without the poverty, diseases, and mass famines that made life so harsh back then. We can make candles or bread and sell them to our neighbors for a living wage. We can walk or get in self-driving cars to see live performances of music, drama and dance by local artists. There will be philosophers and poets with their small followings as they have now. However, even when machines can do everything humans can do, there will still be a capacity to sustain as many mathematicians as there are people because mathematics is infinite. As long as P is not NP, theorem proving can never be automated and there will always be unsolved math problems.  That is not to say that machines won’t be able to do mathematics. They will. It’s just that they won’t ever be able to do all of it. Thus, the future of work could also be mathematics.

  1. Silver, D. et al. Mastering the game of Go with deep neural networks and tree search. Nature 529, 484–489 (2016).

8 thoughts on “AlphaGo and the Future of Work

  1. “As long as P is not NP, proving theorems can never be automated and unsolved math problems will always exist. That is not to say that machines won’t be able to do mathematics. They will. It’s just that they won’t ever be able to do all of it. “

    I think you are wrong here and that the human clumsiness which made AlphaGo victory possible apply in exactly the same fashion.
    Even where the barriers are some sort of superexponential blowup the machines will be more clever and go farther than humans.


  2. Moravec’s paradox reminds me of ‘baumol’s cost disease’ —eg the fact that the productivity of a string quartet (or even theorist) hasn’t gone up though their wages have —- it takes the same number of musicians to be in a quartet to play the same number of songs as it did in the past and it can’t be automated, Though, electronic music does automate it, but its a different kind of music, so unless people’s tastes change, there will always be a ‘cottage industry’ of manual laborers in music. I bet half of people couldn’t tell the difference.there are plenty of studies on this (see simkin on arxiv—fine art vs a toddler’s drawing are indistinguishable,. People should be born with guaranteed health, basic income (peter thiel or usbig) and a PhD. Obama has a law degree from harvard , so does ted cruz.

    I was also thinking about this on a recent hike in shenandoah park. People spent half a day walking basically exactly like the native americans did centuries (or more) ago in a basically primitive fahion—meanwhjle overhead you can see planes which in the same time cover thousands of miles. One also wonders why people do this—spend time slipping on rocks in trout streams when instead they could just be sitting in a plane watching movies or playing video games. I personally sort of hate all games including sports, and most movies tho I can watch the same youtube music videos 10 times in a row, . Maybe the best movie is ‘weekend’ by jean luc godard followed up by Orpheus (Cocteau) and orphee noir—set in brazil.

    Presumably,people do this because they want to and have some ‘liesure time’ (but even birds etc seem to do things that are unnecessary —though ‘maximizing utility’ may be a non-intuitive , nonlinear task (basically moravec’s paradox) —the shortest path may not be the ‘straightest’ one. (otherwise people might not even sleep or dream. I see these barn swallows almost everyday (beautiful birds—blue on top, orange below, ) flying around all the time—sometimes they are catching insects, but a lot of times just seem to be playing around). (I once discussed with Robert trivers his theory of homosexuality—some view it as a defective mutation, others as a possible form of extended kin (or group) selection—eg alan turing, even keynes..

    My view is that maybe the big problem is ‘rate’ (or ‘history’ and ‘path dependence===wouldah couldah shouldah. )

    There are a lot of unnecessary wars and violence around, economic production and consumer spending— construction (skyscraper, stafdiums), junk food, guns…which seems unnecessary but is popular. —a preference. Its difficult to say or dictate what people shoulld and should not prefer. One would have to ask F de Waal where consensus norms come from. Do people want to fight in wars, or think about whether p=np.

    alot of people i met who joined the military did it because they decided school was no fun. they also liked the discipline. buy now, pay later.

    The Wikipedia article on p/np is quite good, especially the opinion polls. I’m in the minority camp i think–i sort of have a (fake) proof of this (based on Hilbert’s 10th problem, and a more recent one)—so I say p=np or else its an ill-posed problem. There are also interesting polls of which interpretation of quantum theory is correct (many worlds, bohm, Copenhagen…). Bohm recently had a slight uptick in support (see Basil hiley on arxiv). Maybe a Bohm-trump ticket would win. I am more in favor of a differnet interpretation (actually 2 1/2 all from europe) , related to the one by G t’hooft except his is a ‘common cause’ theory—see r f streater—lost causes in physics.).

    Also, even with automation , replacing cashiers with automatic checkout, etc someone still has to make the computers. Theoretically robots can do that too—but the questions is will they.

    The world could probably run on solar energy to a large extent next year (or maybe a few more) if people set up the sahara desert and the southwest as sun farms—which grow electrcicity for biofuel to power humyns —- and developed some batteries, but that’s not the consensus agenda (though there is a subset of society that supports that).

    Issues like food scarcity often aren’t due to lack of food, but the algorithms society uses to distribute it—eg Amartya Sen on famines in india, and the irish potato famine (due to enclosures—privatization of property).

    People may decide its easier to semi-automate—have most work in USA done by machines, biut the machines are made in sweatshops overseas—-see no evil, hear no evil etc. This also keeps people busy, and out of making trouble. I could see entire STEM acadamic PhD programs (quite the rage these days—everyone needs an app to find the closest dentist, speed trap, etc. ) ..of a sort of hierarchical form—some might think about p/np, while others basically spend 4 years copying ramanujan’s notebooks 100 times. And then they’d have top repay student debt.

    I don’t think some medical fields can be automated. (When I was in hospital they were sticking holes through my back to drain my lungs of fluid, and putting a tiny camera up one of my arteries all the way to my heart—I could even watch the movie, and make sure there wasn’t any police brutality going on in there

    That is not easily automated.

    ( Interestingly the people doing this were from the Ukraine—and they have like 5 people each doing different things—some as easy as washing the holes they put in you and bandaging them, others on what seem extremely difficult and delicate operations. (As an aside one person I talked to on my hike in Shenandoah was from Croatia, and now employed at howard hughes lab in loudon, and was educated at Pasteur institute in france. I confused that with the Curie (and also Poincare, de Broglie, and other institutes there). She said Pasteur among other things is famous for having shown that doctors should wash their hands while doing medicine based on some ‘germ theory’. (The germs I think were a punk band headed by darby crash—not Brandon darby, the hero in the movie ‘informant’ about some dude people I knew were involved with and worshipped—one could smell a rat a mile away but no one listens — he eventually graduated to work at breitbart news after setting some people up–‘agent provacateur’. Darby Crash and the germs have their own movie—‘the decline and fall of western civilization’ ; ‘black flag ‘ is in there too —some of those people were from DC—black flag is or was used to keep away insects (roaches). Its also used as a patriotic symbol by some groups like the stars and bars, or confederate flag and others. I always found interesting the ‘aids denialist’ community which was quite popular in dc for awhile—eg there is no aids virus. They know since they never seen one. See also J of Sciientific Exploration —run by H Bauer of U Va (emereti if still around) —he also found the Loch Ness monster.
    I haven’t seen too many of these recently—just water and ringeck and worm snakes. I did hear they saw a bear in silver spring and say its same one that visited NIH awhile back and may even been seen in virginia—swim across the potomac. (I used to do that—I stayed close to the bottom to stay out of the current up near great falls and only had to come up for air one time. if it floods then you have a problem—a few times police tried to rescue me sometimes with helicopters but I told them go away —I used to try to save wild animals running across the street but they would get scared and run right into traffic. ‘I wasn’t drowning until u got here and now I may since i’m letting hold of this rock and going to the chesapeake bay’—I usually only ended up a few miiles down).

    Some people in my area sortuh into that had PhDs in science from ivy league universities—one found a Bigfoot in pennsylvania (see letter to NY Times) and also knew that actually 911 was an ‘inside job’—no planes hit the pentagon, etc. This is why I respect the meritocracy. The same people have theories about ebola and rikivirus —they say its caused by food additives, poverty, colonialism etc. and they will sell you books and DVDs about this. Some books sell, and writers get paid. see ‘science for the people’ on Wikipedia.

    Regarding ‘rate’—to me one of the most interesting topics, similar to ‘rates of convergence’, and size of infinity—even if one could automate much of health care, people may prefer to keep its size at 20% of GDP. People can get sick, and then get cured like hamsters on a wheel, or the revolving door in politics. Its a time allocation problem—-how much time does one spend on prevention, and how much on cures.


  3. Carson, so you write in the last paragraph about having an adequate economic system to sustain low-productivity occupations. Are you hinting at a universal basic income system? I imagine that, at a certain point, the low-productivity occupations would be just for keeping people busy and satisfied to dissipate unrest. For the poets, musicians, blacksmiths, etc. at the bottom of the pyramid, I don’t ever see them making enough from sales alone to sustain a consistent living wage from “an honest day’s work.” But then again, is a living wage even a mandatory guarantee? I personally like the thought of people being fed and allowed to be comfortably bored, I have a few Randian acquaintances who find basic income abhorrent and are comfortable with a massive weed-out of people to solve overpopulation. But, things never work out so simply and tidily.

    Steve Hsu also thinks we will co-evolve with machines ( through machine-assisted genome editing. The notion of a competitive arms race in the future between iterating machine and edited human is entertaining. However, my gut sense is that machines are more straightforward to improve and humans have a more limited ceiling.


  4. @Wally I really don’t know but if the cost of production goes to zero then the current conception of economics goes out the window. Perhaps there will be something like a basic income but I think even that concept is too steeped in 20th century thinking. There must be some way to share the wealth. As for your Randian friends, what makes them so sure they’ll come out on top of the Mad Max world?


  5. This reminds me: I (and others I have shared your idea with) long for an elaboration of your “Humans will always have comparative advantage in NP problems, so we (or AI) should make video games embedding NP problems that humans can play to remain relevant” idea, and how it might fit in with the alternative you outline here of just generally low-productivity employment.


  6. @Rick Yes, you bring up two good points. The first is whether humans do have any advantage over AI in solving NP problems and the second is whether we should make candles or play video games. I will try to get to these in more details in future posts. The first claim was made before AlphaGo and more along the lines of since algorithms do not have an advantage in solving NP problems, this is one place where humans may have some utility. However, deep learning has shown that machines can probably mimic human problem solving strategies so that point may be obsolete. In the second point, I think one future could be the matrix where we all retreat to a VR world, and solving NP problems could be one way to help subsidize it. However, my preferred future would be not to sit around and instead make furniture or bake bread.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s