Economics, Sociology, Social Commentary, Political Philosophy
“Life is just like a game. First you have to learn the rules of the game, and then play it better than anyone else.”
Albert Einstein.
Why start a post by quoting somebody else? Although a great mind, Albert Einstein did not revolutionise machine learning in the same way that he did atomic theory or research into blackholes. AI was simply not his area of expertise. Why then bring him up at all?
For my answer, I would point you to the AlphaGo project of DeepMind. Yes, I know that AlphaGo has been talked about obsessively by now. A tournament that occurred in 2016 is still looked on with awe by many. Considering it's not my style to simply rehash the work that’s come before me, you might be wondering why I’ve chosen to write about it. Media attention, for the most part, has been diverted to new projects. Surely anything worth writing about AlphaGo has already been written? You would think so. Except, the logic behind slack moves was simply too enticing for me to ignore
Firstly, I should probably define what I mean by a ‘slack move’? It’s a term which originates from the tournament itself, describing AlphaGo’s seemingly erratic behaviour. During the match, there would be times were AlphaGo made ‘stupid’ moves. Specifically, they made moves that no human player would make because there were other potential moves, moves which would immediately give the player points. Hence why people thought the machine was being slack, failing to take the competition seriously enough. What nobody expected was that AlphaGo knew exactly what it was doing. See, unlike human players who measure their success rate at any moment based on their score, AlphaGo measures its success rate based on the probability of winning. Moves that appear ‘creative’ to a human player are logical to AlphaGo because they maximise the chance of winning in the end. Conversely, the most obvious moves, the ones which increase a player’s score by the highest amount, suddenly become idiotic as they actually decrease the probability of winning in the long term.
Of course, looking at long term gains isn’t something entirely unheard of. Look at the economics behind diamond mining, for example. Homespun commonsense would dictate that you extract as many of the pretty gems as quickly as possible so you can sell them. That’s not how the diamond industry operates. Chances are you’ve come across those social media posts that say there are enough diamonds that everybody could own a cup full of them. So why are they so expensive? Supply and demand. A few companies control all of the world’s diamond mines. Using their monopoly, they decide how many diamonds enter the market at any point in time. Rather than flooding the marketplace to make a quick buck, diamond sellers allow very few gems to be sold at any one time. Artificial scarcity means that people are willing to pay ridiculous amounts for whatever they can get their hands on, blissfully unaware of the massive crates abandoned in warehouses. Suddenly homespun commonsense doesn’t look so good.
Applying the same method on a larger scale clearly demonstrates how delaying production could lead to efficient resource management. Economics has this horrible connotation of being solely concerned with the generation of money. Such a definition misinterprets the field’s purpose. Economics cares not just about money but any and all forms of wealth. Wealth is necessary for society to function. It is the sum of our laptops, clothing, homes, money and whatever else you want to name. Thus far, politicians have treated Gross Domestic Product as the score through which success is measured. A 2.5% rise in productivity is a good thing because it shows one country is doing well in comparison to another. Money is great because it provides an easy metric, a simple numerical value, through which to compare ourselves with others. However, measure progress through reference to numbers is overly simplistic.
Continuing the game analogy shows how flawed the system really is. Say we take a game like Pong. To ‘win’, the machine simply has to get as many points as possible. By way of analogy, for a company to ‘win’, it has to make as large a net profit as possible. As long as the number keeps going up ad infinitum, investors.
Conversely, games like Go don't care about your score. Have you got more points than your opponent? If yes, you win? It doesn't matter whether you have one more point or a thousand. Whatever happens and you win. Winning is still the end, however, how a victory is measured is utterly different from a scoreboard. Change the metric and everything changes.
Now, how life and society can be compared to a game is still slightly unclear. Clearly, governments have ends they work towards. Year on year they want the economy to boom, employment to go down, and life experience to go up. Existence likewise has various goals. What those end goal might actually be is still subject to philosophical debate, but humanity seems to have some final end. Aristotle termed this purpose Eudimonia, for Jeremy Bentham it was maximising utility, and so forth.
Utilitarianism, or at least rule-utilitarianism, promotes something of a linear scoreboard. Happiness is the only good, and all actions can be given a quantitative value of some description, meaning that the purpose of existence is simply to crank those numbers as high as possible. Phrasing it in this way starts to show why the theory comes under criticism. Nobody wants to think everything they do is basically a more complicated version of Tetris. From a statistician’s viewpoint, however, the theory makes logical sense. Everything else in our life is measured. Calory intake, intelligence and the number of cells in your body can all be quantified. What's stopping us from just combining a bunch of other numbers to measures somebody's success at life.
Luckily, I'm not talking about measuring success in life today; some other time perhaps. Instead, I'm looking at 'slack moves' in economics. So what have I actually established? Looking at what I've said so far, it seems I'm suggesting companies should take not of AlphaGo's strategy. My observations basically amount to restating a well-known phrase. 'Slow and steady wins the race'.
Slack moves, slack economics, don’t just have to focus on profiteering though. Businesses and individuals aren't the only people that control wealth. Governments use laws to accumulate and regulate all the time. ‘Slack’ economics could be the answer to efficient resource consumption. What do I mean by this? Take a mining company. They will dig up as much ore as possible so that they can make money selling it. In this instance, money is the ‘score’. At a certain point, the market reaches a state of diminishing returns, before entering a state of deflation. Yes, shareholders are adversely affected by the decline. Guess who taxes those industries though. Governments. Especially in places dependant on a few key exports, poor resource allocation can be deadly.
In the past, I’ve made outlandish claims that robots should be our philosopher-kings. My argument was along the lines of ‘humans are corruptible and biased so we should trust an impartial machine to rule us all instead’. Indeed, that argument makes sense, but in hindsight, it is quite clunky and sci-fiesque. Instead, I’d like to amend my claim. I still believe that robots should be our philosopher-kings (if we really want to personify a machine in that way), but my justification has shifted somewhat. A machine isn’t simply better at the job because it is impartial; instead, machines are better at measuring the metric of life. If society as a whole is analogous to Go, then there are only two outcomes, win or lose. Humanity understands that bit, yet they are obsessed with ensuring they win by a big margin. To be good at life is not merely to ‘win’ but to win by a large margin. On a sociological scale, that means we are predisposed to big plays that boost our ‘score’ by a large margin for a short time. Consumerism is my case in point. Earth has x trillions amount of resources, only a small volume can be used at any moment in history if we want to be sustainable. Globally, consumption is currently over this volume. In the short term, that makes our score look great— people can have giant yachts and seven holidays a year. Eventually, though, we will lose the game. Nature will chip away until we are nothing but dust.
A machine just wouldn’t think like that. Society is Go; ergo, a machine will play ‘slack moves’, moves that may well reduce their short-term scoring potential, but will ultimately win the game in the long run. ‘Creative’ solutions, like national wealth funds, will become the new norm solely because the short-term cost of investment is far outweighed by the long term benefits. In the short term, many of us would see a sharp dip in our standard of living. Second homes and cars will be a thing of the past. People have been annoyed this year because they couldn’t go abroad on holiday. No doubt a mechanical leader would hardly see swanning off to Crete for a few weeks as a priority. As somebody on the ground, the change wouldn’t look positive. Leaders would definitely be seen as slacking off. Luckily our robotic leader isn’t there to make us feel better. Centuries from now, when the forests are green, and the Barrier Reef has replenished, our successors will thank us. We will have won the game of life. The fact our win was by a hair’s breadth will be irrelevant.