Every few months someone asks me a new version of the same question: now that we have ChatGPT — or soon, AGI — will AI finally beat the market? It is a fair question, and the honest answer disappoints anyone hoping for a magic oracle. AI is a genuinely powerful tool. But the market is not a puzzle waiting to be solved, and that distinction is everything.
The market isn’t the pyramids
You can point a reasoning engine at a fixed problem — how the pyramids were built, a chess position, a math proof — and it can grind toward an answer, because the problem holds still. Markets do not hold still. They are an adaptive system made of people (and now machines) all trying to outguess each other. The moment an edge becomes widely known and exploited, it decays. A market that could be permanently “solved” would stop being a market. The goal was never a final answer; it is a durable, testable edge you re-validate as conditions change.
“I asked ChatGPT for 10 proven trading techniques”
Try it and you will get ten confident, plausible strategies — moving-average crossovers, RSI bounces, breakout rules. The problem is not that they are invented; it is that they are the same folk-wisdom techniques published for decades, and most do not survive an honest, costed back-test. An AI trained on the internet fluently repeats what people say works, not what actually does. Confidence and correctness are different things — the exact trap human traders fall into too.
Even a party trick can look psychic
There is a web game, Akinator, that guesses the character you are thinking of with eerie accuracy — not because it is psychic, but because it is narrowing probabilities from millions of plays. Markets fool us the same way: enough patterns and data, and something always looks predictive in hindsight. AI makes it easier than ever to find those illusions at scale. Without out-of-sample testing, a powerful model just lets you overfit faster and more convincingly.
Where AI actually helps
None of this makes AI useless for trading — it is excellent at the work around the edge: writing and checking code, summarizing research, speeding up the testing loop, catching your own logical mistakes. What it cannot do is hand you a permanent edge, because the thing that produces edge is the same as ever: a clearly defined rule, tested on data it never saw, traded with disciplined risk. Use AI to test ideas faster, not to outsource your judgment to an oracle.
The unglamorous truth
The trader who quietly does well in an AI world is not the one with the cleverest model — it is the one who still insists on testing every idea with costs, sizing every position to survive, and following the rules when it is uncomfortable. AI changes the speed of the work. It does not change the work.
If you want the durable version of “what actually works,” start with tested analysis vs. technical analysis and how to back-test honestly.
Prediction has a terrible track record
Lord Kelvin — a genuine genius who helped write the laws of thermodynamics — declared heavier-than-air flying machines impossible, shortly before two bicycle mechanics settled the question on a North Carolina beach. Expert prediction fails constantly, and not just in physics: gold was supposedly headed to $5,000 on a collapsing dollar, and “peak oil” guaranteed ever-rising crude. Forecasting a complex adaptive system is a pseudo-precision business — which is why my systems don’t predict at all. They react. A rule doesn’t claim to know the future; it claims that when X happens, doing Y has carried an edge across decades of tests.
What actually changed when machines arrived
Markets have absorbed technology shocks before. My own research puts a structural break at 1982, when stock-index futures launched: the S&P 500’s short-term character changed, and systems built for the pre-futures market broke. High-frequency trading rewired liquidity in the 2000s. AI is the next wave, and it will surely arbitrage some patterns away — the fragile, over-fit ones first. What it can’t remove is the behavior that creates the durable edges: leverage still gets forced out, fear still overshoots, and a machine trained on markets made of humans is still trading against humans — plus other machines answering to human risk officers.
The edge that survives is the one you re-validate
So my answer to “will AI beat the market?” is that the market isn’t a thing you beat once — it’s a game that redeals. Whatever tools you use, the process doesn’t change: hypothesize, test honestly, hold data back, watch the live results, and retire what stops working. AI makes the testing faster. It doesn’t repeal the discipline — and it definitely doesn’t hand an oracle to whoever types the best prompt.
Educational only; not investment advice. RelaxedTrader is not a registered investment advisor. Past performance is not indicative of future results.
