The computer overlords have come to finance Holy HAL. As artificial intelligence gets more, well, intelligent, it’s playing a bigger role in the finance world -- and it’s starting to write its own rules. Although there are certain efficiencies inherent in letting a machine run the numbers, some worry that AI anarchy could lead to economic catastrophe. It’s getting quieter on the trading floor You’ll still see stressed-out guys in suits yelling about buying and selling, but more of that work is getting done on servers. Computer-run funds make up 35% of the US stock market, 60% of institutional equity stocks and 60% of trading activity. Sophisticated programs can do more than just manage portfolios They’re now starting to monitor the economy and move capital. While most of these AI programs are designed to follow rules established by their human controllers, newer programs are starting to write their own strategies… often in ways humans can’t quite follow. ‘I’m sorry, Dave. I’m afraid I can’t do that.’ Data is a big driver in determining how to move money, but some economic phenomena -- like the trade war or falling interest rates -- require a human touch to navigate. Meanwhile, some hedge fund managers have raised concerns that AI might not operate under the same strict insider-trading and disclosure laws they do -- you can’t program scruples. Other investigators have complained that the way AI chases and abandons securities can distort asset prices, and that some algorithms have already led to “flash crashes.” And because it’s fund managers who vote board members in and out of office, AI could even start playing a role in corporate governance.
While I should love 100% of any article that contains the sentence, "I'm sorry, Dave. I'm afraid I can't do that," I must object to two parts: Sure you can. Just the same as you program only legal trades, or only those that comport with margin requirements or trading hours or Authorized User regulations. Silly statement. Also silly. AI, in seeking out value, will also issue BUYS while the market is still tumbling, and SELLS while the market is still rocketing up -- it thus would be acting as oil on the roiling ocean of market action. As well, who's on the other side of such herd behavior? Other herd members -- whether 'bots or live humans -- for every transaction in a rush to sell, there is a buyer going down, and for every transaction in a rush to buy, there is a seller climbing up. Whether climbing or falling, whether human or 'bot, the concept of "blaming" one side of a market for the actions of the overall market, is inconceivably stupid with just a moment's thought. Sheeesh. Now -- all that said -- "Loved the article!"
AI is already here in one form or another, and it’s here to stay the same way that the classic computers came and never left the field. But in a such noisy and ever changing environment like securities pricing it will have a pretty hard time ( especially nailing predictions on longer time frames). Bright minds are struggling these days to find way to use ML and other forms at AI in this field. Time will tell but I think AI’s role will be limited in the foreseeable future. Self driving cars and even humanoid looking machines would probably be here earlier than a hedge fund managed by AI without human guidance and supervision. But what do I know ?! Nothing is harder to predict than the future.
When AlphaGO went up against, and beat, then the world's best GO players, the moves AlphaGO made were completely unconventional. Well, that was the reason why the players lost because they couldn't think out of the box. Perhaps we could see the same effect in the markets as well where patterns don't jive with the patterns we're familiar with.
I read a sentence like this and it comforts me. These are the opportunities created by this new technology.
@tommcginnis sure you can't blame computers for distorting prices, but can you blame them for sudden bouts of volatility? I would say hell yes...
Gary Kasparov did the exact same thing against the first Deep Blue. He started playing erratically and the computer couldn’t calculate the permutations properly.