AI Wealth Truth (06): Why 'Equal Opportunity' Is Mathematically Impossible

I. "Equal opportunity" is one of the core promises of modern society. We can accept unequal outcomes, as long as the starting point is fair. As long as everyone has equal chances to compete, the final wealth gap is "reasonable". This is the moral foundation of liberalism.

II. But mathematics tells us: equal opportunity is an unstable state. Once tiny differences appear, differences self-amplify. To maintain equal opportunity requires continuous external intervention. Once intervention stops, the system immediately returns to inequality.

III. The mathematical principle behind this is path dependence.

IV. Imagine two people start a race from the same starting line. The rules are perfectly fair. At each step, both randomly move forward or backward by one step. In theory, with enough time they should hover near the same position. But in reality they do not.

V. This is the famous random walk problem. Mathematics proves that two random walkers starting from the same point will, over time, have a distance that grows without bound. Not converging to the mean, but drifting farther and farther apart. Even if every step is completely random and completely fair.

VI. Why? Because random fluctuations accumulate instead of cancelling. If A moves right 10 times in a row, and B moves left 10 times in a row. The distance between them is 20 steps. That gap does not disappear by itself. Subsequent random walks keep fluctuating on top of this gap. History accumulates. It is not forgotten.

VII. Wealth distribution follows the same logic. Two people start from the same wealth baseline. Each year's return is random, sometimes gain, sometimes loss. The rules are identical for both. But decades later, their wealth gap may be worlds apart. Purely because their historical paths differ.

VIII. This is the essence of path dependence: history cannot be reset to zero. Every stroke of luck, every choice, every shock in your past influences where you are now. And where you are now determines how far you can go. Tiny differences at the start become huge differences in outcomes under the magnifier of time.

IX. Related to path dependence is an even deeper concept: non-ergodicity.

X. In physics, "ergodicity" means: if a system runs long enough, it will visit all possible states. For example, gas molecules eventually spread evenly through a container. If a system is ergodic, then the time average equals the ensemble average.

XI. But economic systems are non-ergodic. Over a long enough time, you will not experience every possible wealth state. You will not become a billionaire, then a beggar, then middle class again. You move along a specific path, and you never "traverse" all possibilities.

XII. Why does this matter? Non-ergodicity means: aggregate statistical laws do not apply to individuals.

XIII. Suppose 100 people go to a casino. On average, the house has a 5% edge. If you had unlimited money and unlimited time, in the long run you would lose 5%. But in reality you do not have unlimited money. If you lose several rounds in a row, you may go bankrupt and be forced to exit. You cannot keep playing and wait for "reversion to the mean". The game ends for you.

XIV. Physicist Ole Peters studied this problem. He proved that in non-ergodic systems, a strategy that maximizes expected value can still lead to ruin. Because expected value is the average across all possible paths, but you will experience only one specific path. If that path happens to be bad, you are done.

XV. This explains why some "positive expected value" strategies fail in reality. The expected payoff of entrepreneurship may be positive, averaged across all entrepreneurs. But most entrepreneurs fail. If you are the one who fails, expected value is meaningless to you. You cannot comfort yourself with the ensemble average. You only live once. You only walk one path.

XVI. Back to "equal opportunity". Even if at time t=0 everyone has exactly equal opportunity. After time passes and people walk different paths, opportunities are no longer equal. Because past paths determine current positions, and current positions determine future opportunities.

XVII. A simple example: Two equally smart kids start at the same point. Kid A's father happens to know an internship opportunity and recommends A. Kid B does not have that information and misses it. Because of that internship, A has a better resume and enters a better company. Ten years later, their positions and incomes diverge hugely. One "luck event" changed the entire trajectory.

XVIII. What does this imply? To maintain "equal opportunity", you need to keep pulling people back to the starting line. Otherwise, any tiny difference gets amplified by path dependence. Progressive taxes, inheritance taxes, education subsidies, social security. These are all attempts to pull people back to the starting line. Once these interventions stop, inequality grows naturally.

XIX. Perfect "equal opportunity" requires perfect redistribution. Each generation starts from zero. No inheritance. No family network advantages. No information asymmetry. This is impossible in reality, because people will resist. So "equal opportunity" is always a question of degree, not a state you can reach.

XX. AI makes non-ergodicity more extreme. In the traditional world, path divergence takes time. The gap between 20 years of accumulation and another 20 years of accumulation is somewhat controllable. But in the AI era, one success can be copied endlessly. If you build one AI product right, you might become a billionaire in a year. The speed of path divergence accelerates exponentially.

XXI. The "opportunity window" in the AI era is extremely short. If you started studying large language models before November 2022, you had a first-mover advantage. If you only started in 2023, you were already behind. The window may be only a few months. Miss it, and it is gone. Paths get locked in within an extremely short time.

XXII. This is a devastating blow to the faith in "equal opportunity". We think: as long as the rules are fair, outcomes are fair. But path dependence tells us: fair rules are not enough. History accumulates. Non-ergodicity tells us: you live one path. The average has no meaning for you.

XXIII. "Everyone has a chance to become a billionaire" may be statistically true. But for 99.99% of individuals, that "chance" will never happen. Because their path does not lead there. A truth meaningful at the aggregate level may be meaningless to the individual.

XXIV. Equal opportunity is not an endpoint you can reach. It is a dynamic process that requires continuous maintenance. Every moment, new differences emerge. Every moment, interventions are needed to correct them. Once you stop intervening, the system slides toward inequality. This is thermodynamics, not politics. In the AI era, the speed at which the system slides toward inequality is faster than ever.