In case you weren't aware, machines are now reading the news. Actually, a few thousand machines—probably more—may have already read and scanned what you're about to read for any hint of tradable information. Of course, this isn't just limited to blogs or news disseminated on the web, but social media too.
Late last year, a company decided to launch "streams of data from Twitter and the securities discussion site StockTwits that are 'normalized' and ready to be fed to computers for analytical processing...designed for use by hedge funds and high-frequency traders."1
Now, you've got to admit, machines scouring the web for tradable information is a bit sci-fi. Are they just collecting information though or buying and selling on it instantly? The answer is both.
You see, the notion of A.I. taking over the stock market, or even the entire financial markets as a whole, may seem a bit far-fetched for some but for long-time floor traders of the world's largest stock exchange, it's simply part of doing business.
Yesterday, for example, well known market commentator and Director of Floor Operations for UBS, Art Cashin, pointed out that a hypothesis began circulating among those on the floor of the NYSE that an instantaneous and uniform selloff around 10:00 a.m. was the result of a "computer using artificial intelligence" to parse a highly anticipated speech from Federal Reserve Chairman Ben Bernanke for any indications of providing further stimulus to the markets.
Just like the supercomputer Watson beat out the previously undefeated champions on Jeopardy, once a certain configuration of words triggered a high probability decision, a massive flurry of trades were fired off in a few billionths of a second before you or I could hit the sell button.
Whenever A.I. was debated in the past, detractors would often cite human language as the one thing machines would never be able to understand. With all the nuance, metaphor, and complexities of human speech, there was a temptation to draw a line between sentient and non-sentient life based on the intelligence required to decipher our often cryptic and emotive form of communication. With Watson, that line was blurred once again.
Keep in mind though, IBM's Watson was more a publicity stunt than anything. It wasn't like the language barrier was broken the moment Watson was developed. Although it may have awoken the public to just how far machines have come, Watson represents a level of technology that's already been flying under the radar for many years on the web and, of course, the stock market.
In all honesty, we probably know very little of just how intelligent and extremely pervasive algorithms have become. Over a year ago, a company announced the ability to execute trades within nanoseconds!2 According to Wikipedia, "One nanosecond is to one second as one second is to 31.7 years." Just let that seep in for a moment. How many millions of trades can be made in one second? Too many.
Although speed isn't a substitute for intelligence, in terms of trading, it's a pretty good proxy. It's not merely how fast a trade can be executed but, even more important, the volume of information that can be collected and processed simultaneously.
This is probably the answer to why most high frequency trading never actually results in a trade. Astonishingly, "95% to 98% of orders submitted by high-speed traders are canceled as the firms rapidly react to shifts in prices across the stock market."3
Let's think this through for a moment. How fast does a stock react to changes in the market at the level of milli, micro, or nanoseconds? When you're trading at that speed, are stocks really moving so fast that a computer can only execute 2 out of every 100 trades to make a profit? I think the chances of that are pretty slim.
More than likely, this reflects similar processes taking place on the web every day: tracking, data collection, and hacking. I'll explain.
The Atlantic just ran a story titled, "I'm Being Followed: How Google—and 104 Other Companies—Are Tracking Me on the Web." In case you're not aware that this is going on you should probably read it and take it to heart. This is the quote that is most relevant to our discussion:
Every move you make on the Internet is worth some tiny amount to someone, and a panoply of companies want to make sure that no step along your Internet journey goes unmonetized.
If smaller retail companies are willing to pay big money for data collected and analyzed by complex pattern recognition software and predictive algorithms, how much more do you think Goldman Sachs, JP Morgan, or other multi-billion dollar companies have been paying to use these same technologies tracking and predicting stocks?
Goldman spends tens of millions of dollars on this stuff. They have more people working in their technology area than people on the trading desk...The nature of the markets has changed dramatically.4
Mind you, this was a quote made in 2007. A lot has changed even since then.
So, it's probably fairly obvious that high frequency trading isn't just being used for trading, per se, but for the purposes of tracking and collecting data by millions of algorithms modeling the complex movements of assets in real-time. What's more? This process of flooding stocks with unexecuted orders can actually be used to manipulate the market.
As I pointed out in Markets, Murmurations, and Machines, the global think-tank International Economy—whose editorial advisory board includes former and current presidents of the European Central Bank along with various Federal Reserve Chairmen— ran an article stating:
...high frequency trader interaction with computerized algorithms of large-cap financial institutions is providing opportunities for high-speed, virtually undetectable market manipulation. Where there is opportunity to “shape” the market for advantage, it is likely that such opportunity will be exploited.5
Specifically, how is this done? Well, on the internet it's called a distributed denial of service (DDoS) attack where a large number of traffic is directed toward a single site or server—usually by individual users or a botnet—until it crashes. The equivalent to high frequency trading (HFT) denial of service attacks is quote stuffing. "A tactic of quickly entering and withdrawing large orders in an attempt to flood the market with quotes that competitors have to process, thus causing them to lose their competitive edge in high frequency trading."6 There you have it: hacking the stock market, HFT style.
Going back to the original question of my title, whether or not you believe A.I. has truly taken over depends on how you define A.I. If we measure intelligence relative to the best humanity can offer in certain areas, then, as with chess, Jeopardy, and now the stock market, A.I. has clearly outmatched us. If we take a more metaphysical approach and define intelligence in ways that are hard to measure—like self-awareness—A.I. rests comfortably in our minds as no more than a myth.
That is, until we realize that myths often speak of universal truths.
 Machine-Readable Tweet Streams for Algos Arrive
 Wall Streets Need for Trading Speed: The Nanosecond Age
 SEC May Ticket Speeding Traders
 Algorithmic Trading, Wikipedia
 The Marginalizing of the Individual Investor: The inside story of flash crashes, systemic risk, and the demise of value investing