Artificial Intelligence (AI) and the Stock Market: Buyer Beware
You know the saying: if it sounds too good to be true, then it probably is.
Artificial intelligence (AI) is all the rage these days as tech giants race to capture market share. The promise of new and accessible technology has everyone trying to come up with the next greatest idea, and identifying market opportunities with AI is just one facet of this AI boom.
Recent advances, often demonstrated by generative AI leaders like OpenAI’s ChatGPT and Google’s Bard, are making impressive strides in automating workflows, creating code, and churning out articles for the web. But these capabilities have yet to become readily applicable to the financial markets.
Can You Ask ChatGPT to Predict Stocks?
There’s no comparison: asking a chatbot to write an essay on the fall of Rome is hardly on the same playing field as asking ChatGPT the best trading strategies or to predict the S&P 500’s performance. In fact, generative AI tools like ChatGPT provide a disclaimer that they may provide inaccurate statements, or that they stopped collecting data nearly two years ago.
So, can you ask ChatGPT or Google Bard to predict stocks? Well, technically you could ask the chatbot, but should you is perhaps the better question.
The Pitfalls of Relying on ChatGPT and Generative AI to Predict Market Performance
When it comes to recounting well-documented history, the task is relatively straightforward because there are few variables and assumptions involved. This is because the events have already taken place and most of the related information is known and recorded.
However, making a trade or investment recommendation requires predicting the future state of the markets, and despite how advanced AI technology has become in the past year, these AI algorithms are only as good as the past data they have collected. Predicting future market performance requires these machines to make countless assumptions regarding the economy, government actions, and geopolitical events that may or may not ever occur - and this hardly scratches the surface! These stock market predictions also demand that AI considers the available capital, risk tolerance, and timeline of the inquiring investor.
Striking a Balance Between AI and the Trader for a Successful Investment
One of the appeals of autotrading is how it incorporates AI-powered algorithms to reduce the risk, volatility and flawed biases that human involvement may introduce into the equation. Naturally, you’d expect that an AI bot would be the shiny new solution for making investment decisions. But there’s a careful balance that must be struck here; AI chatbots like ChatGPT exist in a vacuum somewhere in the digital sphere. The human element may not seem necessary to execute a data-backed strategy, but in reality it is essential.
The generative AI widely available to the everyday person with an Internet connection was most certainly not devised to handle the complexities required to assimilate and normalize the vast and varying data required to make a forecast regarding future market behavior, or account for the many hidden risks of trading and investing.
Of course, the dream AI-powered technology that many investors desire from AI bots like ChatGPT is likely already available to the Wall Street elite, but what about the mainstream investor?
Well, we believe generative AI is far more likely to be used for its glittering sex appeal than to be a holy grail solution for generating profits for the investing masses.
Alternatives to Stock Market Trading with ChatGPT, Google Bard, and Others
Over the past few years, countless firms have exhausted an array of artificial intelligence techniques to build investing models designed to conquer the markets. Unfortunately, success stories are rare, and some have even blown up in the process.
So, how has InvestiQuant managed to successfully and profitably implement AI-driven autotrading solutions while so many have failed?
The Keys to Successful Autotrading with AI: iQ’s Stock Market Success
Our focus is extremely narrow. Instead of trying to “boil the ocean,” we’ve chosen to focus on a narrow subset of markets (index futures) and the intraday time frame only. When pursuing statistical edges, it pays to go deep - not wide. Our AI stock trading is designed with very specific risk constraints. Finding an edge in the markets that occurs regularly and with a high degree of certainty is worthless if the risk required to capture it is too high. InvestiQuant’s advanced machine learning algorithms don’t just identify if there is a significant probability of a directional bias in the market. Rather, they answer the more important question: “Is there a clear historical bias today, with attractive profit potential, that can be traded without exceeding our maximum risk tolerances?”
We started early and leveraged our trading experience. InvestiQuant was launched in 2008 as a statistical trading service. Our strategies were based upon thoroughly researched - but rigid - rules using market factors that had historically demonstrated an edge. As the markets evolved and some of our edges waned, we struggled to incorporate new factors and improve our signals.
In 2013, we made a conceptual breakthrough and created our Machine Learning 1.0 engine. It allowed us to expand our strategy inputs tenfold by organizing them into categories that we, as traders, believed to be uniquely predictive for the intraday session. The ML engine was engineered to auto-update the factors nightly to adapt to changes in market conditions, as well as their historical performance for each market environment. By ensembling the different factor groups and making them market-adaptive, we were able to create a powerful directional bias indicator that dwarfed the best results of any single indicator, or group of indicators.
To refine our innovative approach, we collaborated with Duke University's Center for Quantitative Modeling in 2015 (the Director subsequently became a Board member and investor in our company). Though we tried, it took seven years to improve upon this first-generation approach. In fact, iQ’s Machine Learning 1.0 engine continues to drive many of our best trading strategies today.
We are nuts about data and continuous improvement. Artificial intelligence is only as good as its data. As the saying goes: "garbage in, garbage out." With the incredible horsepower afforded by advances in machine-learning technology and generative AI, this concept is especially true.
Over the past two decades of observing markets and their behaviors, we have accumulated a massive proprietary database of market factors. And we don’t just buy 3rd party data and plug it in. Nor do we outsource the tedious and laborious processes of data checking, correcting, scrubbing, and formatting. We manage our data completely in-house using our own data scientists who follow strict protocols. Why? Because we risk our own capital with the same InvestiQuant algos and strategies that we license to our clients and bad data = bad results!
When trading intraday, robustness and speed are king. In 2020, we realized that to take full advantage of our ever-growing database and IP, we needed to utilize the most advanced AI technology available. So, we embarked on a nearly 2 year R&D project. And in late 2021, we launched the iQ Machine Learning 2.0 engine. It was a significant upgrade to the ML 1.0 engine, in that it could process the real-time state of every factor in our database (nearly 1,000) and generate statistically robust predictions every minute (faster if needed)—something not remotely possible using our first-generation engine and approach.
The Future of Trading Stocks with AI
If it sounds like we have it all figured out, we don’t. However, we do have well-established AI machine-learning technology and proven processes for finding and trading real, statically-based edges using our autotrading programs. Most importantly, our strategies have stood the test of time and are not based on fanciful, over-optimized backtesting aberrations created by an amateur programmer or inexperienced quant researcher. Like all things in trading, and perhaps even more so with artificial intelligence: buyer beware!