Putting the AI in Wall Street
Imagine a world where supercomputers and AI algorithms do all Wall Street trading at lightning speed, without human intervention, and all the money in the world is managed by AI. Surprisingly enough, we aren’t that far from an age where this happens. But how will this new system function? And what gives one artificially intelligent robot a competitive advantage over another – if any?
Let’s start off with the basics – the stock market is simply a processed collection of decisions. In the past, these decisions were made mostly by human investors making choices from their own judgement. However, these crucial decisions are being made less by people, and more by automated machines trained to make smart trades based on mountains of data.
In fact, a common use of machines in investment is for Robo-Advisors, technologies made to assist with financial investments by analyzing information and choosing investments based on a client’s risk preferences.
The benefits of using machines over human judgement are clear. Currently, successful traders in the stock market make it by constantly looking at news to collect information, analyzing that information quickly, making fast decisions, and taking risks. Luckily (or unluckily), these are same exact traits that AI-enabled computers have. The stock market is staged for a large transition to artificial intelligence that will revolutionize the way trading works.
Computer Domination
Investment firms will all have their own supercomputers collecting massive amounts of data, processing it at incredible speeds, and enacting buying decisions for their clients. Eventually, as the AI get smarter from the abundant amounts of data present in the day-to-day exchanges of the market, they can begin to take advantage of normal investors in one of two ways:
1: The computers can each out-smart and out-play the normal population, unleveling the playing field. By analyzing what the average human investor will do before they do it, they can make smart buying decisions that increase their profit margins, therefore decreasing the human investor’s profits.
2: The computers can out-right collude with each other, so the market becomes computers (and their clients) vs humans. In Part Two (“Hacking the AI Race”) of my work with the Harvard AI Initiative, I outlined possibilities of large computers from around the world using small signal trades to work against the rest of the market.
Either way, only the computers and their clients will stand to benefit.
Everyone Wants In
And so everyone therefore tries to become a client of these computers. Unless they are wealthy enough to attempt to create their own trading algorithm, they will be forced to use these companies to make their buying decisions, transforming the market into a competition of computers.
What makes one computer unique from another? Other than data, there’s not much. Computers will be trained to act a certain “winning” way, as at any given range of time, there is only one way to get the most money. Many say that there are a variety investment paths one can take to be successful, and while that’s true, only one set of investment choices at any given time will yield the most money.
Imagine a chess competition of all equally-trained, identical robots. No robot would be able to outsmart any other robot, and no one would ever win the competition. In other words, the stock market will eventually be comprised of equally-footed competitors all going for the same prize. In economic terms, a perfectly competitive market – a hypothetical situation where no one earns a profit in the long run.
But can they get in?
The above situation is non-ideal for these computers and their corporate creators. Clearly, these companies would rather invest for only a select few wealthy individuals, in order to drastically increase their buying/purchasing power, and still have a large-sized market to win against.
This way, they are able to earn money – why else would they invest?
Legally, it’s perfectly okay to choose which clientele you want to serve. Morally, it’s not. But don’t fret – 2 solutions exist.
Solution 1: The Natural Way
Eventually, with computers clearly dominating Wall Street, the poorer investors who aren’t clients of these computers will unionize. Just like political revolutions in our world’s history, an economic resistance will begin. By pooling money, they can create their own AI computer that is just as competitive as the other Wall Street AIs. And they will all become clients of that new computer.
And if that one starts excluding people? Then that group will create another.
Solution 2: Uncle Sam Intervenes
By regulating that these Wall Street investing funds have to be more reasonable with their clientele selection, the US (and other countries’ economies) can completely remove this problem. The market will once again be a bunch of investment computers!
Either solution forces these computers, instead of colluding with each other, to play against each other.
But even if these solutions can stop computers from completely out-playing normal investors, they can’t stop yet another problem – large-scale hacking.
Hacking
There’s also always a risk of severe hacking. Companies who have that much money at their disposal can make investments that slightly favor certain industries or companies. The current state of AI is a black box, which means if it’s impossible to show reasoning behind investments, companies can be more shady about what they do behind-the-scenes, without major questioning.
Of course, the government and Securities and Exchange Commision (SEC) can step in to ensure that datasets and the algorithms themselves are as fair as possible across the entire economy, but with something as unclear as a neural network, it’s hard to stop all cases of fraud.
Hold on a Second…
Other than large-scale hacking, Wall Street wouldn’t be too much different from today. Investors would still invest against others, just with large companies as their agents. So a big question needs to be answered – if it’s literally the same thing, is there any benefit to putting AI in Wall Street in the first place?
In my opinion, I don’t think so – regulating Wall Street to keep superhuman AI computers out would prevent the above mess. Regulation would entail stopping the usage of smart AI in stock market analyzing computers, or heavily curtailing it.
However, would this even be effective? People still have an incentive to cheat, if the benefit is high enough. In this case, companies could still secretly use computers to get an upper edge. It would be hard for any regulatory agency to enforce such a rule.
In this case, making it a regulation only further monopolizes the cheaters who unlawfully, but undetectably, use AI to decide their stock choices. It’s important that legislators around the world decide now, before it’s too late, on their policies on AI, not just in finance, but everywhere.
Are we hopeless?
Not necessarily. It will be quite a while until Artificial intelligence can completely understand behavioral forecasting/economics to predict stock movements with 100% accuracy, although there are currently many quite successful ones. For example, a simple prediction algorithm (outlined in my book Stock Prediction with Deep Learning) achieved a 50% annual return rate.
In addition, another key variable for these dominant AI computers is risk tolerance, a somewhat touchy-feely concept that AI has not yet mastered (stay tuned for more). Let’s hope that stock traders continue to innovate on this front, and on all fronts, to stay ahead of the robots.
Recent Comments