Artificial Intelligence and the Death of the Free Market

In classical economic theories, the “market” is an abstract entity, used by everyone, but owned by no one. It could be regulated by a government by enacting the rules and regulations of market transactions, but even the government did not own the market (unless it was the sole producer and consumer). In fact, nobody could say what the market really was, because every individual was part of the market.

But things have now changed. The markets are not free anymore. They are owned by the likes of Amazon, Google, Facebook, and other technology platforms. All these companies will tell you about the power of the free market that led to their rise. They won’t say that they are now the market, that everyone else is a player in their control and they use that power to profit from all market transactions.

An Illustrative Example

Take Facebook for example. As you scroll down your Facebook feed, the software tracks the rate of scrolling. If you pause on something for a second, Facebook knows that you are interested. If you scroll faster, then Facebook knows that you are not interested. When you scroll randomly, then Facebook knows that you are anxious. When you dwell on negative posts, Facebook knows you are depressed. If you focus on positive posts, then Facebook knows you are in a happy mood. It uses all these behaviors to create your personal profile.

If you are depressed, you might not tell your family, employer, or friends about it, because you feel shy or embarrassed. But you will do things on your computer or phone by which Facebook can know your mental state, personality, likes and dislikes, thought patterns, and so forth. You think you deserve privacy, and you maintain your privacy by not telling your friends, family, and employers about your thoughts and feelings. You don’t know (or don’t care) that Facebook knows more about you than anyone else, simply because your finger movements create a personal diary available only to Facebook. So, they can read your personal diary, and use it to determine your nature and state.

As Facebook figures out that you are depressed, they will show you more depressing pictures. If it figures out that you are happy, then it will show you happy pictures. Many of the things that you see in your feed are promoted content. Which means someone is paying in advertising to make you see it. Classical advertising used newspapers, TV ads, radio ads, and so on. The prospective buyers of a product were 1% of all those people who read newspapers, watched TV, or listened to the radio. Software platforms like Facebook change that. Now, by tracking your personality, desires, and mental state, then can target prospective customers with products more effectively. By targeting, the advertiser has to spend comparatively less on advertising, as they are getting more “bang for the buck”—more views for less money.

Of course, it is not always easy to decode your mental state and personality from your finger movements. Therefore, Artificial Intelligence is touted as the technology which will create a system to determine your mental state and personality from the observation of your finger movements. By improving AI, Facebook will get a better grasp of your personality, moods, and thoughts, and target you better.

The Foundation of Artificial Intelligence

The specific technology underlying modern AI is called “Neural Networks”. It is a network of nodes created in software that is divided into many layers. The logical nodes in each layer connect to the logical nodes in the next layer, and the connection is assigned a numerical weight—i.e., something between 0 and 1. When you give some input to a neural network, it triggers some nodes in the first layer, which then triggers the next layer nodes based on the connection weight. These triggered nodes then trigger multiple nodes in the subsequent layer, and this process continues for many layers until the final layer has been triggered. The greater the number of layers and nodes in a neural network, the more sophisticated it gets, which gives the neural network its power. The triggered nodes in the final layer are called the “output”, and quite like the input, they are also a sequence of bits. Thus, the input is a sequence of bits, and the output is also a sequence of bits. The difference between Neural Network AI and the classical models of AI, which rested on programming logic, is very subtle.

In classical AI, people were trying to represent the knowledge of the world, and then program a logical procedure (called a program) to use that knowledge. The problem was that each domain of knowledge required a unique knowledge representation and program. For example, a surgeon needed a different system of knowledge and programs than an engineer. This made AI very expensive because it required expert surgeons and engineers to devote their time to producing a useful knowledge representation and programming the logic of their actions before AI could be effective. The Neural Network approach solves that problem. Now, you produce a single hardware and software system, and you give it to the experts—e.g., doctors and engineers—and they can “train” that system, which is a euphemism for setting node weights.

In this training, the doctors or engineers don’t need to tell you about their mental models of the world, how they think about problems, or how they go about solving those problems. They can do all of that in their heads, and just tell the neural network about the solution for a given problem. When enough problems and solutions are provided, the neural network develops its own understanding of the doctor or engineer, which may be quite different than the doctor’s or engineer’s own mental models, but they produce the same results. Thus, the AI system is not identical to the doctor or the engineer, but it can act just like the doctor or engineer. The term “artificial” in AI is used to indicate this difference; the doctor’s or engineer’s intelligence is “real” while the neural network has “artificial” intelligence.

A neural network model for checking the grammar of sentences in the English language is a set of weights—between 0 and 1. English language neural network models are freely available on the internet and you can download such a model from the internet, install it as an input to your neural network software, and create an AI system capable of processing the English language. Effectively, you can have commodity computers, and commodity neural network software, and the only thing that matters now is the neural network model.

In this new age of AI computing, computer hardware, and software, are slowly becoming irrelevant. What is relevant is “data”—which is a euphemism for a neural network model—produced from a lot of training data, either given by experts or surreptitiously collected by observing human activities. When a model is created by experts, it is very expensive. But when it is created from the observation of human activities, then it is very cheap (for the collector). This cheap form of AI is dominant today, while expert-driven data models (such as to be used in self-driven cars) are still under development. However, the gaps between these two types of AI systems keep narrowing.

In the self-driven car neural model development, for example, a car is driven by the neural model, unless it makes a mistake and a driver takes control. If the driver does so, then it indicates that the software was doing something wrong. That is training for the software—to learn from the driver’s actions—and reduce the time that the driver has to take control. If that metric is constantly on the decline, then the AI model is improving. This process can take years if not decades because the rate of improvement can get slower with time.

How Artificial Intelligence Is Deployed

Facebook for example can have many neural network models—one for each user, another one for people in a city, another one for people in a country, another one for doctors, another one for perverts looking to harass women, another one for drug dealers, another one for people undergoing medical treatment, and so on. The underlying hardware and software are the same—they implement a neural network. This greatly reduces the cost for the company doing the Artificial Intelligence because it can deploy the same hardware and software many times. The difference is simply in the observed data that is fed to “train” these models—sometimes continuously to evolve the model.

For example, the base model in Facebook is of each user. As you scroll your screen, you create a time series of hand movements, which is then used to create the neural network model of that user. Facebook constantly experiments with new users by showing them a wide variety of things. If you scroll past that post, then the neural network is trained to not show you such posts. If you dwell on it, click on it, like it, share it, or comment on it, then the model is trained to show you more such things. Once the model is trained sufficiently, then it can produce an output for any given input. The input can be a certain news article, and the output would be a number that indicates whether you will like it, share it, comment on it, or ignore it. So, a news article is now fed into user-specific models to decide if it will produce “engagement”.

This per-user neural network model can also be used to create “meta-attributes” such as a doctor, engineer, drug dealer, sick, male, female, heterosexual, homosexual, etc. by comparing your responses to other neural networks previously trained for the above-noted attributes. For example, Facebook can get doctors into a room, give them a series of posts, and check their reactions. Through those reactions, they can train the “doctor” model. If your behavior is similar to the behavior predicted by the doctor model, then you are classified as a doctor. Each of these is a weight between 0 and 1, which means that in the perfect scenario, a doctor gets the weight 1 against the attribute “doctor”.

In AI language, this is called “classification”, which means that the AI system is putting you into many classes or categories. Once per-user behaviors and their meta-attributes (that classify these behaviors into categories) are determined, then these models are continuously improved through more data. For example, if you were initially classified as a doctor, but your behavior is slightly different from the standard doctor model, then the software will take your behavior into account as that pertaining to doctors, and modify the doctor model.

Thereby, there will be models of all doctors, all engineers, all perverts, or all drug dealers. Then, there can be intersectional behavioral models in which doctors in each country have different personalities, or drug dealers in different cities behave differently. There is literally no limit to how many attributes are incorporated in modeling and how many intersectional models can be created by their combination.

Again, you can input a certain news article into that model, and determine whether all doctors or all engineers will like, share, comment, or ignore that article. Based on that likelihood, Facebook can decide whether it should show that group of users that particular news article. Facebook can decide how effective this exposure to articles has to be. For instance, the advertiser who pays a lot of money to Facebook for publicizing a certain news article can get better treatment—i.e., they get more relevant and interested users who will like their article. Conversely, the advertiser who pays a little money to Facebook can be given worse treatment. Thereby, simply by the power of money, some content can be made “viral”, and that popularity attracts even more people to it due to social validation. Facebook can also incentivize new customers by giving them much higher benefits initially and then tapering off their benefits slowly—so that they will have to pay Facebook more money in order to get the same benefits. It all depends on which behavior Facebook thinks makes the most money for them.

This money-making process is also a neural network model, which can be different for countries, cities, engineers, doctors, salesmen, and so on. And there can be intersectional models too. Thus the same amount of money produces different effects for different users.

Effectively, once you make your neural network software, all you need is a lot of computer hardware to instantiate that software many times and designate each such hardware-software combination as a machine that is training a different neural network. These machines then automate all business decisions to maximize the machine owner’s profits. The personal, national, community, religious, and professional classification or modeling, help Facebook to target users selectively. And advertisers pay for that selective targeting of users, which helps Facebook make money, improve its software and models, and constantly manipulate both advertisers and users to increase its own profits.

Why Governments Can Do Little

As we have noted above, the neural network is just a set of numbers. If you give this network an input—e.g., a news article—the output is also a set of numbers. You cannot do anything with numbers alone; you have to be able to give these numbers an interpretation.

For example, if a neural network produces three numbers—0.9, 0.8, and 0.75—you cannot make anything out of the digits themselves. To know what these numbers mean, you must know that these three numbers correspond to an attribute tuple such as {doctor, American, rich}. This interpretation of numbers is embedded in software, proprietary to the company that creates the neural network model. Thus, the tuple of the three numbers above means that there is a “rich American doctor”. Without that interpretation, numbers mean nothing.

Now, we have to know that software is “private property”, which means that unless some national interest is involved, governments cannot look at a company’s software. At best, they can say: This data was produced in our country, by the citizens of this country, so we have rights over it. But that doesn’t affect Facebook, because Facebook can give the government terabytes of numbers—which is the “data” produced by the citizen’s behavior and constitutes their behavioral models—and the government cannot figure out what those numbers mean. Effectively, the government cannot use that data the way that Facebook can, because the data is useless without the software interpretation.

So, all laws about data privacy are effectively useless. The government must have laws to compel Facebook to “cooperate” with the government which means giving them the interpretation of data rather than the data itself. Facebook will happily comply because they now know that the government depends on Facebook. The government asking for Facebook data is leverage for Facebook over the government. If the government doesn’t give Facebook concessions, Facebook can turn off the data tap, say that they don’t know, or give misleading information. The government cannot know whether Facebook is telling the truth or not. The governments know this very well; they cannot fight daily battles. If they antagonize Facebook, then they can get some data or interpretation, but it could be useless or misleading.

The Usurping of the Free Market

Effectively, the users who provide the data by their finger and thumb movements, are helpless. Their behavior is being modeled like guinea pigs under an experiment—feed this and you get this behavior, feed that and you get another behavior—and use that feeding to create a model of the guinea pig. The businesses who produce advertisements are helpless because they don’t know to whom Facebook feeds their advertisements, and their profiles—i.e., whether Facebook is truly giving the bang for their buck. They just have to trust that Facebook is doing a good job, without the ability to verify what they are doing. Finally, the governments are helpless, because they have created privacy laws over software ownership, that give Facebook the power to do what it wants and keep the government on the tenterhooks.

I’ve used Facebook as an example. This is not unique to Facebook. It also applies to other large platforms like Amazon, Apple, Google, Microsoft, etc. They all track clicks, the text you type, and the fingers you move. They may also selectively share the data they have with others—for a price of course—and they just have to hide what they are doing in obscure legalese called the “terms of use”, which nobody reads, but even if they read it, they can do nothing about it, except quit the platform. If many users move to a platform, then by quitting you only isolate yourself. Effectively, you are locked into the platform, and manipulated by the platform, to help someone else profit.

The classical free-market system was about producers and consumers, regulated by governments. Economics professors stated that the market was so complex that no single entity could control it. The market had to be kept free because that was the most efficient system of market operation. The inability of any single entity to control the market was the primary reason for the collapse of communism.

But now the erstwhile thesis that the market is too complex to control is disproven. The market can be controlled—by algorithms, software, and machines—by monitoring everyone constantly. Thereby, there is a fourth player—large software internet platforms—more powerful than producers, consumers, and the government in the classical free-market system. It is no longer a free market, because the software platform is the market within which everyone plays, and it is owned by a private business. One business owns the market, and users, businesses, and governments are tenants in the market. At best, their freedom comprises voting by their feet and quitting the tenancy, to their detriment.

The Cluelessness of Current Debates

There are many clueless people who still think that we are operating in a free market system. They talk about the excesses of large technology platforms and regulations to control them or consider this the choice of the free market that should not be tampered with. They either don’t know economics or technology, or both. Their ignorance is overshadowed only by their hubris about technological progress.

Yanis Varoufakis calls this “techno-feudalism”, reminding us of the times in Europe when the feudal lords owned the land, and workers tilled that land for the feudal lords. In India, this was called the Zamindari system, in which the Zamindar (the landowner) had power over the workers who tilled the land. But this analogy is deficient; the reality is much worse, because there used to be hundreds of feudal lords, and the government still had power over them (although the government may not have exercised that power). For instance, the worker could move from one feudal lord to another, and the government could take the land away from the feudal lord. Both these powers are now gone. You cannot quit Facebook, Amazon, Google, Microsoft, Apple, etc. to go to another platform, because the alternatives either don’t exist or by changing the platform you lose most of your friends. Nobody can compete against their existing power. And the government is helpless.

So, you might think that the correct term is “techno-communism”, where the technology platform is the communist government that controls everything—the people, businesses, and the bureaucracy. But even that term would be deficient because the communist rulers still had some form of socialism. The principle for large software platforms is simply increasing the profits for themselves and their shareholders.

You might like to say that the technology platforms are autocratic, with no constitution and rule of law except the profit motive, the goal to advance personal power, and then use it to increase profits even more. But even that analogy would be deficient because even autocrats had some responsibility to maintain law and order within the state. They had to ensure that the economy would not decline so drastically that it could lead to a revolution. The software platforms don’t have that responsibility. They are just responsible for increasing their profits.

Thus, no comparison with the past is truly accurate or applicable to the current dilemmas. It is not capitalism, feudalism, communism, or autocracy, but the worst aspects of all of them. It has emerged in the heart of the technology industry, which prides itself as the beacon of progress and prosperity, operating under the principles of the free market, and democracy, as opposed to closed societies, or hereditary wealth.

This is important because people, businesses, and governments remain clueless about how to handle the problematic situation. They often look to the past and try to compare the situation to capitalism, feudalism, communism, or autocracy, but that is not enough. The government can use its powers to break down the tech behemoths, but that won’t solve any problem because the separated entities will develop cooperative agreements, then cite increasing dependency on each other, and then merge back again. Any laws created to stop cooperative agreements, partnerships, or mergers will affect everyone else adversely. So, things that the government can do will either make life worse for everyone or if they try to be cautious about these problems, then they will effectively keep perpetuating the same problem endlessly.

Cyclical Changes in Nature

Modern thinking rests upon the idea of continuous progress, linear time, and a world free of contradictions. Logic is free of contradictions; physics uses linear time; society progresses continuously. This is the dogma or doctrine under which every single subject operates. However, with this type of thinking, we can never understand or diagnose what is actually happening. For instance, if we had left behind feudalism, communism, and autocracy, and embraced free markets, capitalism, and democracy, then why is the past returning at present?

To understand these things, we have to rethink everything about logic, physics, and society we have thought of so far. One way to do that is to imagine the current economic state in terms of the yin-yang dynamic. It involves contradictions, cyclical time, and iterations.

In the yin-yang picture, there is a small black circle inside a white patch and a small white circle inside the black patch. Black and white are opposites, which we call “duality” in Vedic philosophy; they cannot exist on their own because they are defined by their opposites. These small black or white circles inside the bigger white or black patches are the in-built contradictions within each dualistic position.

For example, free-market capitalistic democracy was defined as the social opposition to feudalism, economic opposition to communism, and political opposition to autocracy. But most people did not understand that there is a seed of feudalism, communism, and autocracy built into corporations. A corporation is a feudal fiefdom in which the employees work for feudal lords—i.e., the upper echelons of corporate management—laboring for minimal wages while the feudal lords profit disproportionately from the work of a corporate employee. A corporation is communistic because the upper management has complete authority over the allocation of work, resources, and people, and the opinions of the employees do not matter; neither is the opinion ever solicited nor should it be given. A corporation is autocratic because anyone who shares their opinion, and questions the upper management or their policies, actions, or decisions is fired from the corporation.

This seed of autocracy, communism, and feudalism that exists inside corporations grows with the passing of time, until it swallows the free market, capitalism, and democracy, and undermines the basis on which such corporations were created. In the absence of this foundation, fewer new corporations are created, and even those are either absorbed into the same system or destroyed by current corporations.

In Vedic philosophy, we describe the process depicted in the yin-yang dynamic by the terms “dominant” and “subordinate”. When democracy, free market, and capitalism are dominant, then autocracy, communism, and feudalism are subordinate. This subordinate entity is the small circle inside the dominant bigger patch. However, as time passes, the smaller circle grows in size and swallows the bigger patch, such that the small thing is now big, and the big thing is now small. The situation is now the opposite of what it was before.

The reversal of the dominant-subordinate positions creates a cycle of change in which history repeats itself. That is, even as something is subordinate right now, it exists as a small seed within the bigger thing, and it will be dominant in the future. Therefore, we might despair about the decline of society, economy, and politics in the short run, as we are effectively returning to the feudal, autocratic, and communistic past. But this too is not permanent. The situation will reverse again, which means that corporations will be destroyed from within.

Self-destruction can occur in many ways—(a) employees quit these corporations, (b) they divulge the corporate exploitative practices undermining the corporation’s power, and (c) the selfish attitudes of the corporate leaders demotivate the employees and they stop working for the corporate interests. Basically, the destruction comes from within, and corporations will be undone by their members.

As this process of undermining from within begins, sometimes there are appearances of corrective actions to arrest the decline. But this is temporary and superficial. Corporate leaders, for instance, will often pretend to have become more democratic, and socialistic, or favor the free exchange of ideas and opinions. But that is a charade. Behind the superficial display of openness is the search for troublemakers to be rooted out and replaced by compliant and subservient employees who will toe the corporate line. The troublemakers will be fired from the corporation on whimsical grounds, or opportunistically as cost-reduction, resource redistribution, etc. The net effect of such a change is that the corporation’s productivity, innovation, and zeal are destroyed. And that destruction results from the need to have complete power over the corporation. This desire grows with time, and therefore, once the process has begun, it has to go to its extreme and self-destruct.

Philosophical Implications

Dualism is often portrayed as the battle between opposites, but that is only half the story. The other half is an inner struggle between the opposing tendencies within, which produces an inner contradiction. The inner contradiction is the more important cause of change.

Due to these cycles of change, there is neither eternal hope nor eternal despair. Everything produces its own opposite because the seed of opposition exists within itself. This cyclical change is itself despair to the intelligent person because he can see that whatever good I’m doing now will be undone in the future. An intelligent person wishes to escape this cycle. The less intelligent people like to see progress and they despair lack of progress or temporary decline. The hopeful person suffers more than the hopeless person, and therefore hope is not a good thing. Hopeless existence is also not good. Only the escape from this cycle of endless change can be recommended.

Therefore, if you are despairing at the current predicament, then there is nothing to worry about because it will change eventually. Of course, you and I could be dead by then. By there is still nothing to worry about, as the soul which is not liberated from material existence, is reborn. So, you could be reborn to enjoy life again, temporarily. If you are thinking that change is always progress, then you will be disappointed.

Thus, it is said that the enlightened soul becomes detached from worldly events. He takes solace in devotion to the Lord and gives up the desire for materialistic progress. It is not that this progress never happens, but only that every progress is followed by a regress, and then again by progress. If you look long enough, then you can see why materialistic progress is futile. However, since people are so attached to this materialistic progress, therefore, we can spend time discussing the nature of this progress and describe why it is regressing rather than progressing. That may seem depressing to many people, but out of that despair can come detachment followed by enlightenment.

That enlightenment involves a non-dualistic reality in which the opposites are not separated and opposed to each other. Rather, they are combined in a way that despite the existence of variety, there is no contradiction between the varied things. When contradictions are resolved, then time becomes linear progress. And the world of conflicts is replaced by that of cooperation and mutual love.

Unless we understand how this world is dualistic, we cannot understand how the non-dualistic world is better, and why we should seek it. The discussion of this dualistic reality, the illustration of its problems, and the cause of those problems is also a useful topic if it illuminates the problems, detaches us from this world, impels us toward the search for another kind of reality, and explains both realities perfectly.