<- home

GitHub
Thumbnail

AI & Democracy

Some thoughts on the relationship between AI and democracy.

Introduction

The advent of AI presents significant challenges and opportunities for our society. In the short term, the pervasive nature of AI and its ability to sway public opinion can potentially undermine democratic processes, such as elections. This power can be harnessed either legitimately, through targeted political campaigns, or nefariously, as seen in recent instances of voter manipulation and misinformation campaigns. As AI applications become ever more embedded in our daily lives, there is also the risk of them slowly hijacking our biological instincts and preferences, shaping personal and societal decisions in subtle, practically invisible ways that may escape regulation or public scrutiny.

In the long term, our increasing dependence on AI can lead to it exercising a slow but steady erosion of our autonomy, leaving less room for free will, diversity and pluralism, pillars of any thriving democracy. As more decisions are delegated to AI in multiple domains - from personal choices regarding media consumption, purchases, lifestyle, all the way to collective decisions regarding city management, resource distribution, healthcare or education - humans will gradually be making a lesser proportion of the direct decisions that shape their own lives.

Should the pace of AI development and integration in all aspects of life continue unabated and unregulated, the very far future could lead to a situation where humans, as a far less intelligent species, might find their interests overridden by AI objectives. Our democratic systems might struggle to function effectively in such a landscape, as the power dynamics that traditionally govern human societies get disrupted. The very promise of democracy, that every individual gets a say in shaping societal rules and decisions, could lose its meaning if the decisions that significantly influence our lives are made by AI algorithms instead of human beings.

This prospect reiterates the need for proactive regulation, ethical guidelines and public discourse around the use of AI in our societies. It is crucial that long-term societal implications are considered in parallel with the exhilarating technological advances. Democracy and AI can coexist, provided we acknowledge the challenges, act responsibly and shape AI advances to not only respect but also enhance the democratic fabric of our societies. Ultimately, it is up to us, humans, to determine the role AI will play in our future and how it will affect the democratic values we uphold.

How AI is impacting the lives of the 21st century citizen

There is no facet of our lives that will not be impacted by AI. Already, automated vehicles, gene editing, medical diagnoses, space travel, cryptocurrency, stock trading, automated factories, the internet of things, search engines, social media, entertainment, banking technology, facial recognition, fraud detection, policing technology

As AI takes more of the difficulties out of our lives, we give increasingly more control away. More and more decisions are dictated by narrow intelligence that seemingly knows us better than we know ourselves. But we pay a price for this convenience. We surrender tiny pieces of our autonomy that become difficult to get back - and in some cases, are conditioned by the AI to the point we are a product of it.

You’ve probably used the “Sign In With Google” OAuth option on a lot of websites. As a result, Google knows where you should be on a given hour of any day of the week, it knows when your miss an appointment or make a new one. By reading your email it can work out financial information like how much your pay for your phone contract, whether your’ve missed a payment or received a new transfer. It can tell when your’re likely to take a holiday and were you ‘ll go. Google knows where you live and work and when there are any irrgularities in your movements. Google knows your friends, and their friends, and how you fit into society in terms of your social life. It could probably work out your favourite bar, sports, movies and games. It’s possible that with enough information and intuition it could work out if you were cheating on your partner. How AI could lead to a better future

There is no facet of your life that will not be affected. Now, in an ideal scenario, each of these minor alterations to your life will be for the better. Your personal assistant will remove the stress of organising meetings and events. Tech companies like Amazon, will know you better than you know yourself and be able to fulfil your desires without you having to explicitly state what you want ordered to your door. Your life in AI-controlled Smart cities involve less congestion, are safer because of AI powered policing technology that negates threats before they arise through computer vision and behavioural abnormality detection, and perfectly reliable transport. Your kids, whose genome has been manipulated by gene editing AI are taught by AI oracles, uniquely designed to fit their needs. AGI diagnostic tests determine exactly what is making your sick and prescribe perfect, custom remedies. Your AI GP has obviated radiologists, immunologists, allergists, cardiologists, dermatologists, endocrinologists, anesthesiologists, neurologists, providing an all in one solution. With the map of your genome, as well as the map of every other human genome and the entirety of their medical history accessible, combined with the entirety of human medical knowledge, AI GP’s will revolutionise medicine. Your dating life will have been revolutionised, as your personal assistant communicates with the assistants of others to exchange information and evaluate potential matches. The vast amount of data they have available to them, including biometric data such as your heart rate and sweat levels, and thousands of hours of conversation, allow the assistants to accurately evaluate potential matches, optimising for the desires of each person, whether it be a casual relationship, family oriented or platonic relationship.

How AI could lead to a nightmare future

So there is the potential for good. However, the alternative, and current trajectory is much different. What if these benefits to your life come at a very minor cost. Consider the analgous situation in which I give you a gift you have always wanted like a new BMW, but I give your a papercut as well. You would accept this deal because a single papercut is such a minor price to pay for the benefit that you would receive. A single papercut will not affect your life in any meaningful way. You will continue to be able to live your life much the same as before, with the minor inconvenience of a papercut. The same is true of your the loss of freedom from the incorporation of some ANI into your life - the autonomy you lose from surrendering part of your life to some ANI may be trivial, and you may be able to continue without taking much notice of the minor alteration in trajectory of your life. However, the story is different when we consider what happens if you consider a thousand papercuts. Despite the benefits you have received, suddenly, life is unbearable. The entire course of your life changes in an instant. No longer can you work, socialise, or enjoy yourself as you once did. The same is true for a thousand minor sacrifices made to ANI. One, and life goes on much the same as before. 1000 sacrifices of your autonomy to AI that does not have your best intentions written in its machine code, and your life suddenly becomes unbearable.

How could AI begin to degrade democracy?

  • Biased AI that do not cater for the entire population exclude peoples views
  • As we lose autonomy, our elected politicians are less effective
  • General intelligence will reduce the cost of quashing democracy
  • Technology is developing at a pace faster than a traditional bureaucratic democracy can keep up with
  • AI makes rational decisions. Sometimes, rationality and humanity don’t get on.

If a technology develops at a rate faster than the legislation can handle - e.g. delayed decision making due to the democratic process and feedback procedures means an issue with technology misuse takes a long time to process, in which time another problem with different technology has arisen. The problems can spiral out of control before the legislation can keep u “Demorcay takes hold when the cost of tolerating democracy for the elites is lower than the cost of quashing democracy”

  • The biggest cost of democracy is having to share wealth
  • The most inequality there is in a society, the more demand for redistribution
  • The people with the skills needed, with the means of production will make up a smaller proportion.
  • The demands for redistribution will increase, and thus the cost will increase for the elites.

The cost of quashing democracy will reduce. With general intelligence, we don’t need to keep the bulk of society happy. You don’t need the workforce, or the army. The machines will do it for you.

So how does this impact democracy? In democractic societies of bygone eras, the governing body of the country you resided in set all of the rules that you were subjected to. Your vote meant everything, because giving it to a candidate without your best interests in mind would lead to negative effects for you. Now, as AI reduces overrules the government, slowly chipping away at its power, your vote means less and less. The government controls an increasingly lower proportion of the decisions that are made that affect you. Instead, the control is being shifted to AI. It is easy to see why this is - because humans are much more falible than machines, and generally make better decisions. However, they make decisions well based on their training, and if their training never required them to take you into consideration, your best interests will not play a part in the outcome. You can vote for whoever you like, but the dwindling power of the government will not be able to override the exponentially increasing power of the artificial intelligences that spawn seemingly every day.

Some notes on bias in AI

If a neural net is trained on a data set that doesn’t include any data from people like you, then your opinion is not going to be reflected in the decisions made by the AI. In a democracy, everyone gets an equal vote. In a world where more and more of our decisions are made by AI, whether or not you get a say is determined by whether or not your perspective was taken into consideration during the construction of the AI

If the decisions that the AI make for us are have our best interests at heart, this slow transfer of our autonomy to AI may not be such a bad thing. However, presupposing that our specific interests are going to be represented is a bold assumption. Currently, the types of people who have a say in the development of AI come from a very specific domain and background. In terms of race, background and education, they have a lot of things in common. That is, they are mathematicians and computer scientists, well educated, generally male, and career focused - they work for Amazon, microsoft, google, IBM, Facebook or Apple. The likelihood that any of them have a background in ethics, philosophy, human sciences, psychology, sociology, sustainable development or law is low. The chances that this small subset of humanity can know what is best, and account for the diverse range of people on this planet, without their explicit input, is slim. Thus, decisions are going to be, and to some extent already are, made for the many, using technology developed by the few. This is inherently undemocratic in my opinion.

Where do we draw the line in terms of decisions that we allow AI agents to make for us? We already allow them to determine what music we listen to and what films we watch, soon enough they will be driving for us and managing our companies. They recommend products to us on Amazon.

Systems in which AI is being developed

Where are we developing AI that could lead to the downfall of democratic society?

AI is predominantly being developed within two different systems.

The state-controlled environment in China, in which the development of AI is watched over by, as well as for the benefit of, the state.

  • BAT - Baidu, Alibaba and Tencent
  • Belt and Road initiative
  • PoliceCloud
  • Social Credit Score System https://foreignpolicy.com/2018/04/03/life-inside-chinas-social-credit-laboratory/

China has dedicated a tremendous amount of wealth to the AI sector, with their own sovereign wealth fund devoted to fundamental research. China also has vastly different views on privacy, meaning their citizens are exploited and mined for their data at a scale that makes the Cambridge Analytica scandal look tame. There is some argument, put forward by Baidu CEO Robin Li, that “Chinese people are willing to trade privacy for convenience, for safety, and for efficiency”.

Where do we end up if we follow the Chinese current trajectory to its logical end?

The alternative solution is the one we see in the United States, where the other big players in the AI sphere operate; Facebook, Google, IBM, Amazon, Microsoft and Apple. In America, the government strives to appease the tech titans, who in turn bow down to Wall Street. This model for developing AI is also problematic - we know full well that profit-driven companies can turn psychopathic just as easily as governments can, even if the constituent parts of the company arent.

The problem with the american system is the fact that ultimately, the companies developing AI answer to the market. Wall street and investors demand profit, and as much as possible. Anything that obstructs profit is cast aside or ignored. It would be good if the big tech companies were able to fully take into consideration the implications of their R&D breakthroughs, without having to worry about getting some profitable product to market as quickly as possible. We have countless examples of bad design decisions made by the big AI developers that likely stemmed from a lack of foresight, which originally was caused by the speed at which the companies have to move because of the incessant necessity for profit.

Which paradigm for the development of AI will lead to a better outcome;

  • government controlled authoritarian regimes like china or
  • western, profit-driven democratic (yet heavily abused) systems such as that in America?

How can we avert this degredation?

This is left as an exercise to the reader.