AI and liberal democracy

By Adrian Zidaritz

Original: 02/09/20
Revised: no

It may seem obvious to many that liberal democracy has been just the right ideological formula for economic and social success in the West, and therefore our AI should just adopt it and strengthen it. But even though it may be obvious, an obvious fact would still need to be formally specified in order for it to be programmed into AI; as a software program, AI only understands the language in which its specification is written, some sort of formal language, not a natural language. So in order to have AI promote liberal democracy, we would have to completely understand what we are asking for, because a formal language does not allow for ambiguities. The idea that AI will force us humans to sharpen our understanding of political and economical terms will appear many times in this website. But before we can aim for a formal description, we would have to agree on a description in plain English. Even that task is monumental, let alone the formal version. We will exercise a bit of humbleness and set a more modest goal, at least in the first version of this article, to make more precise the relationship between AI and liberal democracy, a very tenuous relationship at this time.

Liberal democracy, with all the variations it has embraced, is both a form of government in the West and the ideology behind that form of government. It is based on the principles of a representative democracy, economic freedom, rule of law and civil liberties, and a separation of power between branches of government. In most cases, these principles are written down into a Constitution and this Constitution is of special interest to us. Our task of writing down the formal rules of liberal democracy for AI would be made much easier by all the thought processes that went into the creation of the Constitution, and the task would be in theory reduced to a translation from English to a formal language. We will see that some of the principles of liberal democracy, especially the representation principle (= elected officials represent the interests of groups of people), may be significantly challenged by the advances in AI.

There is a deep paradoxical nature to the relationship between AI and liberal democracy. Because of its stronger support for the X-Quartet, liberal democracy appears to offer us a better chance of getting AI right, better than say an authoritarian socialist form of government would have. But by its nature, AI works against liberal democracy in many ways. It exacerbates income inequality as we have seen in the article The Future of Work and potentially creates social unrest. It rewards centralization of power through its hunger for centralized data and centralized processing of this data. We will see below that it also works against the core tenets of market capitalism. We have no idea what the end result of AI might be, but if liberal democracy is worth preserving, some more unorthodox economic and political theories will have to be developed which will show us how to achieve that preservation while developing AI.

Although there may be slight differences between the interpretations of the principles of liberal democracy by the Republican and the Democratic parties, those distinctions are less important than the commonalities, so we will look at liberal democracy as the common denominator for both parties. We borrow from David Frum's article The Republican Party Needs to Embrace Liberalism that "in a democratic society, conservatism and liberalism are not really opposites. They are different facets of the common democratic creed. What conservatives are conserving, after all, is a liberal order. That truth has been easy to overlook in the friction of partisan politics." The hope that our AI systems will also conserve that liberal order, in other words that they will optimize their tasks always within the constraints set by that order, seems remote at this time. Currently, quite the opposite seems to be the case, as our AI is not aligned at all with our liberal democracy.

The video started with the game of Go, and Go serves as a reminder that liberal democracy is not the only modern possibility. What would China like their AI to conserve? Presumably, the rules of their Constitution, which was written down in 1982 under the leadership of Deng Xiaoping, and which promotes "a socialist state under the people's democratic dictatorship". The 1982 Constitution (with periodic revisions and updates made to it) was meant to promote the modernization of China, and from that viewpoint it has been a stellar success. At the same time, it places the final authority in the hands of its Communist Party and the individual freedoms are not as strongly enshrined as in a liberal democracy like ours. Nevertheless, it may well be that China will not embrace liberal democracy and still be an effectively organized nation, providing for the well-being of its citizens. Before we embark on criticisms of other Constitutions, let's remark that our own Constitution was also not meant by its creators to be an immovable Bible, it was meant to be revised and updated periodically by Congress, accounting for new needs and presumably for progress. But it turns out that those updates rarely touch sensitive issues, and meantime we have more guns than people (and our children have to go through active shooter drills in school); the same Constitution that propelled us forward can also be used to justify that unjustifiable fact.

So, AI on the blue dot will be developed under (at least two) different ideologies and implemented under the constraints coming from within very different systems of government. Seen from above, humans have carved the blue dot into nation states, with weapons aimed at each other, enough of them to blow up the whole thing and incessantly vowing for a costly supremacy. Borders, passports, hate, sickness, abject poverty, mountains of trash in oceans and on land, sectarian violence. The Mother blue dot looks with dreamy eyes at all this carving, aware of the lost possibilities, and with tearful eyes when looking at the realities and the immense suffering which this carving has brought with it. Before we return from poetry to prose and rejoin our main theme, we can wonder if, among all this talk of intelligence, we should pause for a moment and evaluate our collective silliness. And after that pause, we may turn the music back on and view the beautiful blue dot in all its majesty, because sometimes the clouds are covering the borders and the silliness:

Modern liberal democracy Includes both capitalist and socialist features

One point needs to be made from the outset and it has to do with the continued dominance of the "capitalism versus socialism" debate on the national stage. These "-isms" of economic organization have lost much of their original stark differences. It is true that if we look at the political pair (democratic, totalitarian) or the economic pair (capitalism, socialism), we see that a certain parity has developed in our consciousness about these pairs, i.e., the view that somehow there is some balance between the implementation and the importance of capitalism versus socialism, and the same balance about the political pair. This is mostly due to the long period of the Cold War, when these pairs were indeed balanced between the West and the Soviet Union and its allies.

But this isn't the case any longer. Socialism (i.e., the state owns and plans the economy, and it controls production, not individuals) appeared late in history, and it rested on artificial recipes, trying to diagnose and repair the perceived weaknesses of capitalism, by centralized authority, and without considering our diversity as humans. It has been a failure. In turn, capitalism in liberal democracies has been re-based on the concept of free markets (although capitalism can exist without free markets and free markets can exist without capitalism, as we see in China). We may be better served probably if we ditched these "-isms" altogether and referred to modern successful economies as free market economies, i.e., the natural state of economic organization, which does consider our diversity as humans. Free market may be Occam's razor for modern economics and one may make the same statement about liberal democracy on the political side. While simpler conceptually, both free markets and liberal democracies need lots of effort to maintain them. By switching to the couple (free market, liberal democracy) we can then make some progress and ask for example: At this moment of the economy would supply-side or demand-side methods work better? What should the Federal Reserve do at this moment? ... and most importantly for our topic, How will AI impact ALL these questions?

The index of economic freedom, funded by the Heritage Foundation and the Wall Street Journal, is a yearly ranking of the world's nations with respect to the economic freedom their people enjoy. Will you be surprised if the U.S. is not among the top 10 nations!? Below you will see the index for 2017, and also a video presentation of the evolution of the index. All nations have some governmental intervention in the markets, no nation has a pure free market economy. Nevertheless, the economists who are behind the calculation of this index generally agree that there is a positive correlation between free markets and economic well being.

Based on our fixation with the way AI will alter the world economies, two nations with high indexes stand out: Hong Kong and Estonia. Hong Kong is an amazing experiment in human history: the closest to a free market (it's been at the top of the rankings for more than two decades) and a neighbor to Shenzhen, mainland China's version of Silicon Valley, where AI is bound to make a formidable stand within the most watched area of China. Hong Kong has almost no government regulations, almost no tariffs, strong property rights, and with one of the highest GDPs in the world as well, it has a vibrant access to global trade. Estonia, a former Soviet republic with a Communist government until the 1989 collapse of Communism, has had an remarkable turnaround and will be one of our best examples for the use of technology to provide intelligent access to government services; we'll dedicate some time to its experiments and their implementations in the article Identity and Trust. On the other end of the spectrum, there are countries which are considered "repressed"." These countries have virtually no economic freedoms. To no one's surprise, the most repressed is North Korea (ranked 180th), with Venezuela (179th) and Cuba (178th) rounding out the bottom of the rankings.

While socialism appears dead (with a few exceptions, likely to be temporary), debating the worth of the socialistic degrees of various policies (welfare, education, health care, etc.) within a free market economy is essential discourse, especially as they relate to AI. So while socialism is out, socialistic degree is very much alive and it will get livelier in the age of AI; the key word is degree. The confusion between social programs and socialism is a source of much pain in the U.S. national discourse. There is no socialism in the U.S. (and especially not in California, despite the noise to the contrary), and even the most progressive (and reasonable) on the left do not advocate it. Bernie Sanders' form of democratic socialism is still free market, but with more generous social programs and lower income inequality. As we will see soon, the Nordic model (the term refers to the Scandinavian countries), which is very often referred to in the U.S. these days, is another example of free market with more generous social programs and a lower income inequality.

Nowadays, there is an unclear treatment of these politico-economic alternatives in the media. In particular, the strictly economic terms of socialism and capitalism are being used in all sorts of inappropriate contexts, loosing their precise economic meaning. If we do not agree on their meaning and on their modern replacements early on, bringing AI into our discussion will only exacerbate that confusion.

Let's clarify the "-isms" once and for all, so we can clear a path to bringing AI into the discussion. The clearest description of these two economic systems I could find is in the following video, but also notice that the video zeros in on the free market concept. I burst into laughter when the speaker mentioned that one of the first noticeable items that disappear in late stage failing socialist countries is the toilet paper. So true! I visited Romania in 1983, while the country was still part of the communist heaven, and the main airport in Bucharest was out of toilet paper. One caveat though: the speaker brings in the idea that government benefits de-incentivise people; as we will see below, that is NOT the right idea, especially with AI around and the absolutely unavoidable Universal Basic Income (UBI) which comes with it.

It seems obvious that a liberal democracy needs a large middle class and it needs the members of this large class to exercise critical thinking. Liberal democracy represents a mature organization of society, we take this without proof: on the economic side, this is the result of society's success of providing enough material wealth to support its large middle class; on the political side, the large middle class has enough leisure time at its disposal to exercise critical thinking necessary to participate meaningfully in decision making. With AI lurking around, the need for this critical thinking has an added poignancy.

Fundamentally, the difference between U.S. and the Nordic countries is given by the different position of the needle between taxation and welfare benefits. The Nordic countries use higher taxation to provide for larger welfare benefits than those in the U.S.. That does not make the Nordic countries socialist. The same different positions of the needle hang between the Democrats and the Republicans in the U.S.. Just like with the Nordic countries, that does not make the Democratic Party a socialist party. (Even the Medicare for All proposal or the tuition-free college proposal will not change the free market foundations, and they are unlikely to pass anyway.)

Adopting policies with a higher socialistic degree will not turn the U.S. into a socialist country; the scare tactics surrounding the use of the word socialism seem to be working unfortunately. To be clear, we are not advocating for specific policies here, that's not the point of this website, we just make the argument that a debate around them is very healthy and should not scare people away. But far more importantly than any ideological preaching, many people in Silicon Valley who work in AI, feel strongly that AI will undoubtedly force us to adopt much stronger social programs. For the simple reason that a sizable fraction of our citizens will simply be unemployable and defenseless in the age of AI. The purely capitalist solution of pushing them to sleep under bridges will not scale. Here is Eric Weinstein, bringing more weight to the argument that the government and Silicon Valley MUST work together when crafting these future AI policies.

Liberal democracy on the march

Since the emergence of the tax and welfare state after WWII, the debate has been significantly altered. As we have seen above, our national debates in the U.S. and Western Europe have shifted from absolute irreconcilable differences between tax and benefits to a matter of degree. The degree of how much to redistribute wealth from the haves to the have-nots, i.e., how much tax to levy on the wealthy and on corporations in order to maintain decent lives for the poor/young/sick/old. That shift towards the tax and welfare state has been very successful. By using that socialistic dial, one can move between more wealth redistribution to less wealth redistribution, turning it left or right and see what works the best. Without abandoning either liberal democracy or free market. So, in the end, we can conclude that modern states like U.S. and those of Western Europe have found different levels of synergy between free market and liberal democracy, but nevertheless a fundamental synergy. That conclusion is supported by the continuous march of liberal democracy on the world map since 1789, the year of the U.S. Constitution, the U.S. being the oldest constitutionally governed nation:

Notice on the map above that in the Americas (North, Central and South), only one country still shows red today, but even in that country ...

AI should boost your engagement in politics

By lowering cynicism and boosting engagement, AI should help democracy. Well, how can it do that? Here is an app that matches your preferences with political parties. But this is just a small step. As more data is collected about you and stored in your digital twin, those preferences can be deduced. After they are deduced, you should be able to log in to your digital twin and confirm the matches between your preferences and political parties. A system like VoteCompass gives us a far better gauge of voter sentiment than polls do, based on a smaller sampling of the population. It is unlikely that events like the Brexit or the 2016 Elections in the U.S. would have taken us by surprise if we had this kind of engagement. Vox Pop is the term used for interviewing on the street and here is a presentation of Vox Pop Labs, a powerful tool for democracy, helping to keep people informed. "A robust democracy requires an informed vote". (Yes, another reason for sharpening our critical thinking.)

Government on a Chip

It is conceivable that we may adopt a more direct form of democracy, which does not need representatives in Congress. This would be a departure from the current representation in our liberal democracy, as enshrined in the Constitution. Let's bring in the idea of the brain biometric, which we presented in section about fMRI in the AI Versus Human Intelligence article. Our digital twins, through these brain fingerprints and all other information that they accumulated, already will have stored all that is knowable information about our desires and motivations; the digital twin in the national graph would do a much better job at representing our interests than any representative would. Imagine, quirky as it may look now, that Congress would consist of a lonely computer in the basement of a one-room government building. Lawyers (humans for a short while) would still propose legislation, but the voting would be done instantaneously on the national graph. You can push this kind of scenario onto the remaining two branches of government, executive and judicial, to get a sense of the magnitude of the politics that would be involved on the way to that stage. It will be fierce for a while, but the one-room (or one-computer or one chip) government will most likely happen, as the natural convergence of liberal democracy; that is, if we don't extinguish ourselves one way or another, which as of now is the more likely outcome. We will half-jokingly refer to this possibility of a direct liberal democracy as Congress-on-a-Chip.

And if we don't extinguish ourselves, do those of us working in AI have a responsibility to aim for social good in all our work, and make sure that if Sam rises and gets elected, Sam can be trusted? (Sam appears below)

One idea that we do promote is that democracy has a better chance of setting the correct initial conditions for a benevolent AI than an authoritarian form of government. Our collective, active intelligence, should be superior to the intelligence of any one individual. But there are clouds in this direction. Right now, the pendulum is swinging away from democracy in two of the most powerful nations on Earth, China and Russia. And it has also started to flirt with a swing away from democracy in humanity's last remaining hope for a benevolent AI, the U.S.

It is unfortunate that the world political pendulum has been swinging away from democratic and towards authoritarian in the three most powerful countries in the world. It so happens that these countries are the ones with the most advanced AI and the consolidation is likely to increase. AI under the control of an authoritarian regime in any of those three will be of grave concern for that regime's population. AI under the control of authoritarian governments in all three would very problematic for all of us, leading to questions of no lesser importance than our very survival. It is therefore not too early for each of us (whether in the U.S., Russia, China, or any other place) to understand what AI is, to understand how it can be used, to be prepared to support democratic institutions that control AI and to guard the integrity and truthfulness of the data in whatever capacity we can.

These two main arguments are not completely independent. Authoritarian regimes can directly control the goals of an AI system and at the same time manipulate the data that this AI system uses, so that the resulting optimization is lined up with the regime's agenda. It seems natural that a democratic control, with its reliance on a larger pool of brains, and a largely distributed effort to guard the truthfulness of the data, should be superior to an authoritarian one. Superior in the sense of securing our survival. But are these the only two political systems that will battle each other? Here is an introduction to one other puzzling but possible outcome from AI, the fully automated luxury communism:

AI techniques are particularly well suited for establishing the wanted socialistic degree for various social programs and establishing the correct taxation levels in support of those programs. Through its emphasis on data, AI can constantly monitor this data and make adjustments to serve the public good. One could argue that the use of AI to solve these problems would meet the approval of both Democrats and Republicans. For the Democrats it would allow stronger and widely accepted social programs, for the Republicans it would mean a very small government, moreover a government whose rules of the game are open and clear for all to see. This would be the extension of the Congress-on-a-chip idea, as well as Federal-Reserve-on-a-chip, Social-Security-Benefits-on-a-chip, and so on.

As opposed to the old way in which technology was created, by which applications were designed and afterwards tested and modified, AI allows the development of incipient quick algorithms which subsequently improve their models based on the large amounts of data collected. These algorithms could eliminate, for example, the degree of arbitrariness which is often associated to how decisions are made by administrative judges who approve/deny a particular social service to an applicant. Exceptional cases in which the system either approved or denied a certain benefit by error, would be used to make future improvements, which improvements would benefit all the future applicants, not just the particular individual whose case triggered the modification. All governmental services will most likely go that route.

It should be clear from the outset that social programs and taxation go hand in hand with the formally specified morality for AI. AI would make decisions within that formal system of morality to allow/deny certain benefits and to establish taxation levels. It is very likely that new categories of jobs will appear, for example we would need Data Scientists who would specialize in various areas of the government, like social security benefits, disability benefits, Medicare, unemployment assistance, Universal Basic Income, etc.