How AI will change our economy?

In this post, I will be presenting my views on how AI will change the wealth gap. The arguments are built on some of the foundational topics in economy, coupled with wisdom from Y Combinator videos and Ray Dalio . The goal of this article is not to educate other folks, but to streamline my own thoughts and create a language and structure around the issue so that I can think more clearly. I hope this attempt will help you formulate your own thoughts around this topic. Please let me know in the comments.

WTF is wealth?

A modern economy generates wealth from land, labor, and capital.

Land is limited, and depending on the location and usage, has different productivity levels. A farm in North Dakota is "less productive" than the land on which a modern 50 storey luxury apartment is constructed in downtown San Francisco.

Labor is diverse and their productivity is measured in $ per hour. A house cleaner who earns $20 per hour is less productive than a yard worker who earns $40 per hour, who in turn is less productive than a data scientist in Bay area with 7 years of experience who earns $400 per hour. This data scientist is less productive than the top lawyers and CEO of mega companies who earn upwards of $2000 per hour. However, there is also a risk involved. The CEO and the data scientists are W-2 workers, so they are able to earn income for entire 2000 hours per year. On the other hand, house cleaners, lawyers, and yard workers may not be able to fill entire 2000 hours of work in the year.

Capital is money, stocks, building, structure, machinery and all that fun stuff. However, in today's world, software and intellectual property (read model structure, hyperparameters used to train the best model, and the best trained model) also are crucial components of capital. The owner of the capital earns passive income due to the invested capital. For example, CEOs earn a W-2 wage but also earn dividends from the stocks they own in the company.

Capital is valued by its discounted cash flow in the future. The capital owner has to pay some cost to acquire the capital -- the money to buy can come from the savings or can be borrowed from other capital owners or banks. After some accounting gimmicks such as amortization, depreciation, discounting and what not, one computes the ROI on capital.

WTF is AI?

When I go out to buy a fruit, my usual decision rule is as follows. If a fruit is on sale or less than $3 per pound, its a good deal and buy it. This is a simple decision rule and easy to execute in my brain in real time and make the purchase.

Consider Abhishek-v2 who is a more sophisticated buyer. He has data of the nutritional value of the fruits and vegetables. He knows how much of this nutritional value will be lost due to cooking and processing. He further estimates his body's nutritional needs for the coming week. Then, he runs an optimization where he minimizes the cost of all the produce consumed subject to constraints on meeting all of the nutritional needs. This leads to a complicated decision rule for purchasing produce. This is called optimization.

Add in some more datapoints, some statistics, and some complex constraints, and Abhishek-v2 becomes a borderline AI system.

Add in some more datapoints, some more statistics, and some more complex constraints, some forecast from the future, some distributed computing framework, some Kubernetes, some SQL servers fetching various datapoints, some API provided by third parties, and we get Abhishek-v3. Abhishek-v3 becomes an AI system.

Now, Abhishek does not have the intelligence and decision making capabilities of Abhishek-v3. Abhishek-v3 is a highly sophisticated rule book, like a car manual. Abhishek-v2 is a sophisticated rule book, like a dishwasher manual. Abhishek is a mere mortal just trying to get by with minimum burden on his grey matter.

What's happening in Capital Owner's head?

The person or entity (referred to as she) that "owns" the data, statistics, numbers, optimization solver, hyperparameters, trained model, server on which the optimization is running, etc. is now the capital owner. She hired some labor to build all of this for her and paid them competitive market salary. She invested her money (capital) to build this. She turns one capital (money) into another one (codes).

She is responsible for the profit and loss statement and management decision. She is taking a lot of risk -- market risk, execution risk, obsolescence risk, employee risk, pricing risk, IP risk, etc. At the time she decided to build the AI system, she could have put the money in a CD, life insurance scheme, S&P 500, mutual funds, farm, real estate, or buy an existing home cleaning business in her locality.

Each of the options has different risk-reward profile. Building the AI system had the highest risk, and therefore commands a higher reward. Putting money in CD has the lowest risk, but the return on CD is "dictated" by the central bank/government who is not in anyone's control (really??).

Now that she has invested so much money into building a system, she will invest money in distribution channels to earn money from the capital. She will become super successful with a probability of 1% and mildly successful with a probability of 5%. Since odds are so low, the return better be good; most of the time, the returns are not so good .

Tying it together

Wealth is discounted future cash flow. AI is complex rulebook. Capital owner has decided to convert one form of capital into another highly risky asset (a form of capital) and would like a risk-adjusted return on capital.

Executing AI system does not require land -- it requires servers that take very little space in comparison to the value created. Since most of the algorithms are commoditized, the labor required to build the system is also low. The cost of labor is also low. The running cost of an AI system is the running cost of a bunch of servers, which costs peanuts. AI is cheap as hell.

In the future, tremendous value will be created at low capital cost and low running cost. Most of the cost will be incurred in employee salaries, mitigating execution risk, market risk, and strengthening the distribution channel. In other words, a highly capable manager with the knowledge of the market and capital (I am looking at you, 40 year old businessmen and businesswomen) having access to a strong distribution channel (I am looking at you, influencers) will enjoy the discounted cash flow.

So the winner is

Intelligent rich managers who own a distribution channel

They will be wealthy, because they will enjoy future cash flow after orchestrating capital to build a useful AI system.

Labor creating the code will collect a normal paycheck as most of the algorithms and servers to run them have been commoditized. Land owners will not benefit from the growth in AI.

What about others?

You are a normal guy doing a normal job. You don't own a distribution channel. You dont manage people -- actually you hate it. What will happen?

If you are in follow-the-rule business and does not require dexterity, you will be out of luck soon. This includes some creative business such as scriptwriting and music. This also includes marketing and content writing. AI already does a good job here. The headcount will surely reduce.

If you are in dont-follow-rule business, you are very lucky. This also includes scriptwriting and music, but Christopher Nolan and AR Rehman type folks who build their business by not following a rulebook. They write nonlinear scripts and create fusion music. This includes branding agencies. Very soon, their sophisticated rules will also be copied by AI, so new more sophisticated dont-follow-rule artists would be needed to generate wealth.

If you are in follow-the-rule business and but require extreme dexterity, you are lucky. This includes car mechanic and babysitters. However, the labor productivity here is generally low.

If your job is protected by the government or society due to legal, ethical or moral reason, you are in luck. This includes babysitters, teachers, doctors, nurses, professors (lucky me), lawyers, accountants, HR managers, and taxi drivers.

The Upshot in long term

Wealth inequality is bound to increase within the broad industry vertical of AI. The class mobility over a lifetime will be severely limited in the future since only "intelligent rich managers owning the distribution channel" will generate wealth in the field of AI.

The labor in AI will receive a normal paycheck. The capital owners will have massive upshot potential.

Others will make a living depending on whether or not their job requires rule-following, dexterity, accurate processing of information and protected by the society.

Outside of AI verticals, wealth will be created in the old fashioned way. There will be innovations over time, some activities will be automated, some not. And general labor productivity will continue to climb in these sectors.

Silver lining

AI has tremendous potential to reduce long term cost or increase revenue with some upfront investment in conceiving and writing codes. Organizations can use AI to reduce their long term cost, which will improve their cashflow in medium term. Since discounted cashflow is wealth, capital owners of such an organization will substantially increase their wealth. This is why FAANG owners are wealthy.

There is a catch though.

Organizations around the world sucks. This is by construction. They will not be able to leverage the full power of AI to reduce their long term costs or increase revenue. And it is good for public.

Ray Dalio says every person's expense is another person's income. Consequently, every organization's cost is someone's income. A lower revenue for an organization leads to someone's income or capital. A higher long term cost leads to higher income for people who do not own the organization.

Organizations whose management structure is not optimized to exploit AI will incur a higher long term cost and a lower revenue (in comparison to the hypothetical organization that is exploiting AI as best as they can). This will create higher income for public and reduce the wealth of the owners of that organization.

I posit that the current organizational structure prevents many organizations to cut their costs and leverage the benefits of AI. This will go on for sometime. The reason is very simple.

The people sitting at the top of the organizations have been trained in old school way. There was no AI at that time. They are fighting to manage day to day operations. They don't understand AI. They don't have the time to understand AI because they do not devote one day a week on reading articles. They don't understand the risks to their business if they do not use AI. They do not even know where AI can help them reduce their cost or improve their revenue. They are not aware of the current state of the art methods. They are unwilling to engage with subject matter experts to advise them on the current state of the art methods and how it can improve their business.

I know this as I have been trying to convince many such organization leaders to use my services as a data science consultant . And I am hitting the roadblocks everywhere. As far as I know, I am not alone. This is the story of many employees across organizations.

I also know secretly that this is good for the society at large. This is because the organizations will be wasting a lot of future cash flow due to their unoptimized systems. Remember, according to Ray Dalio , one person's expense is another person's income. So the public wins at the expense of the capital owners for now.

Note: One person's expense is another person's income, and the "another person" here may be a wealthy organization. Consider a home cleaning business who spends on Google ads for marketing without using a tailored AI for their business. Their expense is Google's income. This improves Google's valuation but reduces the valuation of the home cleaning business owner. Whether its good or bad is left to your discretion.

Conclusion

When land generated wealth (in the form of annual produce), then landowners were wealthy. When industrial capital generated wealth (in the form of oil, steel, copper and gold), then the capital owners were wealthy. Then technology, internet and IT came along and the capital owners there made a killing -- however, the capital requirement and the market risks were huge in this market.

Now, AI is coming along and creating a massive opportunity to create sophisticated rulebooks across industries, which will create incredible wealth for capital owners and middle income jobs for people providing labor.

Rule-following-not-dexterous labor will be obsolete and a bulk of the population will not know what to do about it. They will open small convenience stores and small businesses and generally suffer from low productivity throughout their life while dreaming of a house, car, and basic medical facilities to cater to their needs.

As long as organization leaders sleep on this potential, the public will benefit in medium term. In long term, we are all dead, so I do not think about it yet.