The successes of Apple’s Siri app have seemingly proved to the world, that artificial intelligence (AI) is not a far fetched concept and that there is still more to explore. This partly explains the renewed interest in AI research and applications in almost all sectors of the economy. The stock market is showing signs that in the future, it might dispose of the human hand and carry out all its trading based on this technology.
A bad history with computers
Although computers have been used in the stock market pretty much since the emergence of commercial computing, the relationship has not been a wholly admirable one. Many instances have led to losses due to computer failure.
For instance, in 1967, NYSE had to shift to manual pen and book recording for days after a “human error” led to a problem with the input system. This not only slowed down transactions but also affected profits of many day traders.
In 1987, NASDAQ’s quotation service went offline for over 82 minutes. The cause being squirrels. The stray rodents interfered with the electrical system, leading to partial power failure. The same was to happen in 1994, causing three trading interruptions at different periods, prompting worldwide jokes about the enmity between squirrels and finance.
A problem with a transfer to the decimal system in the period leading to the new millennium caused various delays in major markets, leading to destabilization of share prices. This also partly accelerated the dot com crash, which reached its height in 2000.
In the last fifteen years, hackers and bugs have been the major contributors to stock market faults. And no one has been spared because when anything affects NASDAQ or NYSE, it brings an impact on pretty much any trader.
Looking at it from this sense, the fact that hedge funds still intend to trust a wholly new computing system with their money is therefore somehow intriguing.
How do they tell that AI will not mess things up and leave investors holding the bag? What new thing does AI bring with it that raises chances of making it in places where others have not managed? Is AI worth the investment and effort? Is it trustworthy? What about the hackers? Is artificial intelligence a sham? These are some of the questions raised- and the revelation was interesting:
Aidyia is the most recent company to announce that it is fully running on the AI model. Ben Goertzel, the main brain behind the project, expresses his full trust on the system. He is sure that even if everyone were to die, the Aidyia’s AI would continue executing calls, bids, and responding to warning bells reliably.
But there are many more firms that are either already running on a similar technology, or have plans of launching it in the next few weeks: San Francisco Technologies, Two Sigma, and Renaissance Technologies are among the few firms that have announced they are using AI; Point72 Asset Management and Bridge Water Associates are laying down plans to adopt the system soon.
The hedge funds that have employed AI have not disclosed much about the specific results. Aidyia reveals that on the first day of running, the fund generated 2 percent profit exclusively through AI. However, it has not stated exactly how much was invested.
According to a 2015 survey by PREQIN, AI seems to perform more steadily compared to other systems involving the human hand. In 2011 when human managers made losses of -1.78%, Artificial Intelligence traders closed with an average profit of 2.47%.
However, the worldwide profit margin of AI trading so far has been rather too low and stagnant compared to the human run systems. In 2013 when human managed systems closed at 12.23 percent high, AIs only managed 7.68 percent profits.
Therefore, while the general performance of AI in the buying and selling of securities and other instruments has been positive, the human hand has still beaten it when it comes to the highs.
It would however be a big error to clamp all AIs together when judging them. This is because, just like human traders, one AI system is entirely different from the other. Like humans have different ways of looking at the market and how they interpret situations, so do these systems.
The level of complexity of each AI system also varies from one to another. The Aidyia’s Artificial Intelligence, being the latest, is one of the most complex systems, having been built on the strengths and weaknesses of its predecessors. To put it into perspective, let’s briefly compare quant systems, which the majority of firms still rely on, the Bayesian networks, and Aidyia.
The quants system relies on data scientists who collect and analyze large amounts of data, and use these to determine a funds’ course of action. While they have been very successful, there are inefficiencies within the system that at times override the gains made.
One of these is the fact that more factors that are global currently affect the stock market and this means there is much more data to be analyzed today than the previous years. To facilitate decision-making, the firms have to employ more quants and this translates to higher salary expenses.
Secondly, even with additional data scientists and more computers, the quant system is slow by current standards because in this market, time is money, literally.
Third, a lot of data goes to waste as market variables change daily. The data collected yesterday has to be discarded since it is no longer useful for today’s market.
To counter some of the limitations of quants, Rebellion Research is one of the companies that has joined the Bayesian Network. The system uses member’s computers together with a few other undisclosed trade platforms to collect and analyze behaviors of different markets. Using this information, it may automatically implement certain set actions for users.
The problem with Bayesian Network and its kinds is that they use a limited sample. They have also failed to elicit much trust, especially after the lessons of 2010 flash crash, and so most organizations merely use them for suggestions but leave all execution powers to man.
Aidyia and Sentient’s AI
The current AI brings with it new techniques in that, apart from using large market data collected from thousands of computers around the world, it also incorporates human “genes” into its calculations.
Using behavioral data collected from the best stock market traders, the AI combines all these and uses some form of “median brain” to analyze and act on the data collected. Complex as it appears, the system is quite flexible and adaptable to future changes, as programmers can add new algorithms at any time.
So does this make Artificial Intelligence an “ever winning” alternative? The assumption is that when the perfect AI comes to place, it will give a sure win to anyone who partakes of it; and of course, everyone will want to invest through it to have an arbitrage.
Will Artificial Intelligence overcome the mystery of equities market? If it does, what will happen to loss section of the stock market? Will it simply disappear without an effect? Will stock prices flatten such that people no longer earn from trading leverage? Only the future can answer some of these and more.