Skip to content →

Artificial Intelligence/ Natural Stupidity

Today’s News Of The Week in Review has an article about the precipitous temporary decline in UAL stock when an old news story about their 2002 bankruptcy filing mistakenly appeared on the internet and was picked up by Google as current news. The Times reporter claims that automated AI computer news scanners picked up the news, shorted United, and, following that, algorithmic trading programs detected momentum and kicked in to do the same.

I’m a little skeptical about the first part of this argument. I know there are people stupid-smart enough to use algorithms to trade first without removing stocks (like Lehman) which are going down for a good reason. But are there really people willing to take a chance on a machine detecting news and trading without a quick look at the story itself? I’m inclined to think this started with simple human hot-to-trotness before the machines kicked in.

Things like this have happened before the widespread advent of algorithmic trading, which I first heard about in a talk by Steve Ross: “An amusing example is the case of the stock MCI (see Rashes [2001]). There is another company that is traded on the NASDAQ with no affiliation whatsoever to MCI, call it MCI2, and it is well documented that its price is highly correlated with that of MCI. Important news about MCI moves the price of MCI2. It is no mystery what is happening; some investors are confused. But investor confusion and silliness is not grounds for dismissing efficient market theory. MCI has a market capitalization of $100 billion and MCI2 has a capitalization of $300 million and a small float.”

Published in Models