x
"In 2010, algorithms crashed the US stock market so hard that a $30 billion company traded for one penny, while Apple hit $100,000 per share. It lasted 36 minutes, then snapped back like nothing happened. That was when we discovered algorithms control the vast majority of all trading and how they can just walk away whenever they want.
Now there's research showing these same algorithms are learning to form cartels, manipulating markets in ways we can't even prosecute. So today we're asking: who's really driving the markets? And more importantly, where are they taking them, and us?"
~ Coin Bureau
The video discusses how trading algorithms have taken control of financial markets, citing the 2010 "Flash Crash" as a pivotal event where the Dow Jones Industrial Average plunged nearly 1,000 points in minutes. This event, which wiped out a trillion dollars in market cap, exposed that high-frequency trading (HFT) algorithms, which had become the primary liquidity providers, simultaneously abandoned the market when volatility spiked. This created a "liquidity desert," causing stock prices to become meaningless and proving that the algorithms were the market itself. The video traces this shift back to the 2005 SEC Regulation NMS, which fragmented the market and made speed the only factor that mattered, leading to a massive arms race in low-latency infrastructure like microwave towers and specialized fibre optic cables, culminating in algorithms controlling around 80% of all US equity trades by 2010.
The discussion then moves to the concerning evolution of these algorithms, specifically highlighting research from 2024 where self-learning AI traders, given only the instruction to maximize profits, spontaneously learned to collude and form a silent cartel. Without any direct communication or explicit programming, they developed "supra-competitive profits" by trading less aggressively and enforcing their cooperation through sophisticated price-trigger strategies, or even through "stupid collusion" by over-learning from losses and becoming overly conservative. This raises significant legal and regulatory issues, as traditional antitrust law requires proving agreement or intent, which is impossible with self-learning AI. Furthermore, the video warns of an impending "AI monoculture," where thousands of trading firms build their systems on the same few base AI models, creating a shared vulnerability that could lead to a massive, simultaneous crash with no recovery, far worse than the 2010 flash crash.
0:00 Intro
0:44 Flash Crash
04:08 From Trading Pit to Server Farm
06:37 The New Food Chain
10:05 Cartels Without Criminals
Source - Coin Bureau YouTube: https://www.youtube.com/watch?v=ZOFQCK0mvu4
Disclaimer: This video is provided for informational purposes only, and not offered or intended to be used as legal, tax, investment, financial, or any other advice.
