There are articles circulating today about Ray Kurzweil and a talk he gave at a conference recently. The part of his talk that seems to be getting the most attention is the use of Artificial Intelligence to pick stocks.
If AIs become vastly better than humans at picking stocks, and so become the preferred method of selecting investments, the first thing people will do is start figuring out how to diddle the AIs programming to make their stocks more valuable. We may be in a second Diebold issue here … won’t people eventually suspect that their gains (and much more importantly) losses in their stocks are being manipulated by the corporations with the best hackers.
Wall Street’s long history of corruption is not going to help the situation, either.
Grr… that was a stupid, stupid article written by an ignorant buffoon about an ignorant buffoon. First things first. AI is already being used in the stock market. A lot. Neural nets, GAs and all that fun stuff has been known about since the 60’s. All those PhDs investment banks hire have been building better ones ever since. Theres still a lot of work to be done but this is not by a long shot a new trend.
The other thing is that at a certain level, until we get strong AI, theres a whole lot AIs can’t do. AIs might be great at fiddling with numbers but they really suck at digesting news articles or figuring out historical trends. How do you tell an AI that the CEO was recently found in bed with a hooker and has been forced to resign in disgrace? How do the algorithms handle the eventual fallout? It can’t because it can’t know the social mores of American culture or the personalities of the new CEO. It’s not a matter of algorithms, its a matter of not having the right data in the right form.
What AI’s are really being used for is to help magnify the power of the human intellect. Being able to plot what if courses and do technical analysis and point out unusual events. They act as a intelligence magnifier but if you have dumb intelligence, then you’re going to get crap results, AI or no AI.
People have always used the past to predict the future of stocks.
Chartists, they are called.
And they are right half the time, just like your coin toss.
Technical analysis (i.e., chartists) are a pretty sorry lot. I thought the whole idea was loopy when I was studying for my Series 7 exam, and the final death knell was the point that, if you charted the digits of pi (a random series), you would come up with several “head-and-shoulders” patterns – something technical analysts consider one of the most reliable signals to sell a stock.
I doubt AIs will be particularly good in picking stocks simply because it depends on the myth that the stock market is a rational signal and that all stocks are fairly priced. This just isn’t so – look up the history of stock bubbles and busts. People don’t behave rationally – the hold stocks too long, for instance. It is basically rules by chaos theory, not rationality.
Now, the one issue is that AIs might create self-fulfilling prophecies. If a bunch of AIs misread the signs and determine a stock will be dropping, they will start selling, which will cause the stock to drop. But human investors can then pick up these stocks at bargain prices.
What you have been describing throughout your post is expert systems, not artificial intelligences. Artificial intelligences are sentient, they figure out things on their own. They will not need to be spoon fed data the way expert systems do. There are no AIs in existence as yet, that I know of.
Expert systems are considered a form of AI as are Neural Networks and Genetic Algorithms which was what Kurzweil is talking about. What you are referring to is Strong AI and I know of nobody in the field who means strong AI exclusively when they say AI.
AI in stock picking goes far beyond technical analysis. The problem with these algorithmic solutions is that they only work if you’re the only one who knows about them. Once everybody knows about them, then everyone is doing it and arbitrage removes any profit that you could make. As a consequence, technical analysis that you know about is almost always worthless.
I’m not completely sure but I believe you can comb through historical stock data and pick the points when certain technical analysis techniques were first developed because the returns on the strategy suddenly drop to 0. Since these strategies are discovered many years before they become widely known, it’s not unreasonable to assume there are other strategies known only to certain investors which do work.
Correct, and I have to say the stalemate already occurred on Black Monday in 1987, computers were not the whole reason for the crash, but they were an important factor. Because of that, a circuit breakers system was implemented, which electronically stops stocks from trading if they plummet too quickly.
I have been working at an ag school wa-a-a-a-a-y too long, apparently, since I read the thread title and couldn’t for the life of me figure out how Artificial Insemination could influence the stock market…
“The market” is the sum of it’s participants, surely? So if a significant number of marketplace decisions are made by AI, then how does one AI manage to outperform the other AIs to ‘beat the market’?
And unless someone comes up with an omniscient AI, then I would expect even a very powerful strong AI to do about as well as a bunch of really smart people - i.e. it will perform very well until it gets clobbered by something outside the range of what it was expecting. I don’t believe there is such a thing as a ‘perfect’ strategy that will cope with every possible market event.
Funny how the trend is now to get right out of the market.
Personally If I had the interest and the opportunity, I would go in for snapping up ‘closed’ funds and using their resources to buy out viable but vulnerable companies.
I would think a more pertinent question would be - what happens when the demographically large boomer generation starts drawing on their investments en masse - not all at once, but consistent net outflows?
Like the story of the golem who dug a ditch 20 miles long, computer programs have the potential to be very stupid in the way humans cannot and in a very short time span.
I’m glad you mentioned this, as I also thought of LTCM as a great counter-example to the idea that AI could make money in the market. Remember that the Nobel prize-winners and former Salomon Brothers men there lost almost $5 billion, and might have lost far more if the Fed hadn’t stepped in and bailed them out.
First of all, it’s spelled Salomon Brothers, and that wasn’t the company that lost all that money. The Salomon Brothers reference was to John Meriwether, a former high-level trader at Salomon, who was one of the founders of Long-Term Capital Management, the hedge fund to which slaphead referred.
Second the reason that the Fed stepped in was because they were afraid of a larger collapse in the market as LTCM liquidated its investments. The Wikipedia reference mentions that LTCM had over $120 billion in borrowed investments, against less than $5 billion of its own money. I’m fast getting out of my depth in discussing this sort of stuff and will definitely be out of my depth if I attempt to discuss their derivative investments, which scare even Warren Buffett.