Artificial intelligence is to trading what fire is to cavemen.
An Anonymous Trader
Without a doubt, the artificial intelligence boom we're experiencing these days will transform virtually every industry and economic sector. Every day, we hear news about advances in this field that make names like ChatGPT, Dall-E, or Stable Diffusion seem familiar. Of course, the world of trading and financial markets is no stranger to this revolution, and we'll likely see profound changes in this sector in the coming years.
Aware of these advances, at X-Trader.net we will try to delve deeper into these topics throughout 2023, focusing primarily on the applications this technological revolution will have in trading. To do so, in this first informative article, we will look at what Artificial Intelligence is, what applications it can have in finance and trading, who are some of the main players in this field right now, and what risks and implications all this may have for financial markets in the future.
So, we begin this exciting journey into the fascinating world of this new paradigm that could possibly change everything in trading. I hope you find it interesting.
Listen: What is this Artificial Intelligence?
Artificial Intelligence (usually abbreviated as AI) is a branch of computer science that studies how to make computers capable of performing tasks that require human intelligence, such as learning, reasoning, or language comprehension.
In other words, artificial intelligence allows devices to think and react like humans to perform specific tasks, but without human intervention.
The origin of this branch can be traced back to 1956, when John McCarthy, Marvin Minsky, Nathaniel Rochester, and Claude Shannon organized a conference at Dartmouth College to explore the possibilities of developing "intelligent machines." It was at this conference that the term "artificial intelligence" was coined, establishing this new branch as an independent field of study.
Within Artificial Intelligence, there is also another branch that has been talked about a lot in recent years, including here at X-Trader: Machine Learning (abbreviated as ML), which focuses on the development of algorithms and techniques that allow computers to learn automatically from data, without being explicitly programmed.
While it's true that artificial intelligence techniques haven't experienced a major evolutionary leap in the last century, the truth is that the first twenty years of the 21st century have seen exponential growth in their development, largely due to the increase in computing power but also to the development of new and powerful algorithms.
All of this translates into a sector experiencing exponential growth: to give you an idea, according to a report by Allied Market Research, the global AI and ML market size was valued at around $65.48 billion in 2020, but is expected to reach $1.5 trillion by 2030, representing an annual growth of 38%.
As you can imagine, this new paradigm presents tremendous potential and countless use cases within the field of finance, and more specifically, trading. Let's take a look at what potential uses this technology can have.
The Evolution of Trading
Long gone are those days of trading in the pits where traders would throw buy and sell signals at each other while shouting, in what seemed like a kind of ceremony between savages where no one understood each other (that's what it seemed like, in reality they did understand each other perfectly).
However, things have changed dramatically, and market trading is now conducted entirely electronically, allowing for the connection of algorithms that interact and operate automatically. Back in 2013, we reported on this website that automated trading accounted for around 70% of the volume traded in US equities.
All of this obviously doesn't seem innovative to us now: surely many of our readers will remember how in the early days of this website we talked about creating trading strategies and automating them with Tradestation (when this platform hadn't even become a broker yet!) or from a spreadsheet through DDE against the Interactive Brokers API.
However, with the arrival of AI and ML, things are changing dramatically. These algorithms make it possible to analyze large volumes of data in a matter of seconds and quickly extract patterns and trends that a human would otherwise be unable to grasp or discover, even after several years. AI also helps automate multiple trading processes and reduces human intervention in routine tasks, introducing a clear advantage: as long as the algorithm is properly developed and coded, machines don't tire or make mistakes, as humans do. All of this, of course, contributes to saving time and, above all, costs, something that hedge funds certainly haven't overlooked, as you can see in the following graph:
While there are countless applications, we can say that, in summary, the main applications of AI and ML in the world of trading would be the following:
Pattern detection and analysis: This is undoubtedly the field likely to see the most growth, as the possibilities are almost endless. Using neural networks and other complex learning models, AI and ML experts try to identify the factors that explain stock price behavior in order to determine their future performance and develop trading strategies based on them. Once a specific pattern of behavior is identified, its predictive capacity is assessed, and if its quality is good, it is implemented in the form of rules that allow a trading strategy to be built.
Risk Assessment: Just as AI allows us to extract patterns predictively, similar elements and data can also be used to anticipate risks, thereby avoiding scenarios or mitigating actions that could increase the risk of a financial portfolio.
Sentiment Analysis: As we mentioned earlier, AI attempts to replicate the behavior of human intelligence when performing tasks. One of these tasks is language understanding, which is reflected in the field of AI through Natural Language Processing (NLP) techniques. This type of technique allows for tasks such as analyzing market sentiment using words contained in tweets or news. Other uses of AI in this field include tasks such as entity recognition (for example, identifying whether a particular country, company, or currency is mentioned in an article) or text classification (determining whether a document is a balance sheet, a contract, an agreement, etc.).
Synthetic Data Generation: Data undoubtedly fuels the engine of AI and machine learning, so having large volumes of it is vital to develop models with a certain predictive capacity. However, if we don't have enough data or we want to create new data that mimics the behavior of existing data, how do we do it? No problem: AI is capable of learning the characteristics of what already exists and generating new data that appears realistic. Generative Adversarial Networks (GANs) are often used to address this task, with which we can, for example, create new historical data with characteristics similar to those of any real financial market.
Fraud Detection: ML methods can also be used to detect fraud in financial transactions. Not surprisingly, the SEC and even our own CNMV are working in this field, trying to detect signs of insider trading with algorithms.
Disruption in the Trading Industry
Let's move on to another question: which companies are breaking the mold by applying AI to create new products and services in the trading sector? While it's difficult to compile the names of all the companies joining this revolution, here are some worth considering:
Trading Technologies: we couldn’t have started this little roundup in a better way. As all our readers probably know, Trading Technologies (TT) is the legendary company behind the X_TRADER trading platform. In 2017, TT acquired Neurensic to develop its AI division. Thanks to this, TT currently has an AI platform that identifies complex trading patterns on a massive scale across multiple markets in real time, taking advantage of its entire data infrastructure. The company thus offers its clients the possibility of building their own algorithmic trading platforms on top of this technology, automating the entry and exit of positions, and reducing the market impact of large orders, as well as the risk of manual errors.
GreenKey Technologies: This is a company specialized in automatic voice recognition and natural language processing acquired by VoxSmart in 2021. The technology created by GreenKey Technologies allows extracting relevant information from voice and text data, so that it can be used, for example, to analyze a speech by a central banker in real time, or even to record a meeting and obtain key data from the conversation.
Kavout: offers K Score, a predictive equity scoring system that utilizes broad and diverse data sets, including fundamental data, prices, technical indicators, and alternative data, and delivers a score between 1 and 9, generated by combining statistical analysis techniques and ML-based classification algorithms. Based on these results, Kavout offers various services, such as stock buy and sell recommendations or the creation of customized portfolios.
Numerai: Founded by Richard Craib in 2015, Numerai is a project that includes some very relevant figures in the world of quantitative finance, such as Howard Morgan (co-founder of Renaissance Technologies) and Marcos López de Prado (need an introduction? :D). The way this company works is truly disruptive: Numerai offers huge amounts of databases for free and invites data scientists from all over the world to compete to detect anomalies in this data. Data scientists participating in Numerai enter their model predictions weekly, which are combined with those of other participants to make stock investments using a meta-model. In return, each participant can receive rewards by investing in their model, using a token called Numeraire (NMR), receiving remuneration when the prediction made is correct. The more predictions are delivered regularly and the higher the success rate, the higher the reputation on the platform and the greater the rewards that can be obtained. And all of this without needing to share any details about the model used, and without being able to know the predictions of other participants, making it impossible to reverse engineer the models. With all of this, Numerai gains an unparalleled competitive advantage over other hedge funds by combining the predictions of thousands of data scientists.
Auquan: In this case, we are looking at a platform that offers data science-based solutions to solve investment problems at the institutional level. To do so, Auquan uses a model similar to Numerai, relying on contributions from a global community of more than 13,000 data scientists from around the world, some of whom work in the private sector (Google, Uber) or academia (MIT, Oxford University). This way, the client simply provides raw data to these professionals for analysis. This is undoubtedly a truly intelligent solution, given that creating and/or expanding internal teams of such professionals is expensive and time-consuming. Therefore, it is not surprising that Auquan's clients include financial sector companies such as Optiver and Neuberger Berman.
IntoTheBlock: Of course, in the field of AI, we couldn't miss a service related to another of today's leading sectors: cryptocurrency. IntoTheBlock uses AI and deep learning techniques to generate price predictions for multiple cryptoassets. In this process, this company's models use both spot quotes, derivatives prices, and on-chain metrics.
Trade Ideas: At the core of this company's business is Holly, an AI that analyzes massive data sets, both structured and unstructured (market activity, news, social media, etc.), on all US stocks every night through approximately 40 strategies, each powered by multiple algorithms. The result of this analysis is the daily presentation of 5 to 7 market-beating scenarios. Risk limits are set for each strategy, so traders are alerted when limits are reached, either upside or downside, intraday. Additionally, the execution of these trades is automated with brokers such as Interactive Brokers or ETRADE.
Imperative Execution: With a team of experienced traders, analysts, and engineers, Imperative Execution enables optimal execution through its IntelligentCross product, which uses AI to optimize trading of US equities. Imperative Execution's services are available through virtually any institutional-grade US broker, and it receives over 200 million investor orders per day.
Sentieo: Among other services, this company provides an AI-powered search engine that allows users to find financial information and company news, compiling both internal and external content into a single shared workspace that is systematized using natural language processing techniques. This allows any analyst or investor to access a wealth of fully organized and structured information, significantly increasing productivity.
Two Sigma: in line with what Numerai or Auquan do, Two Sigma offers the service called AlphaStudio, which has a team of more than 1,600 data scientists around the world, who work with a gigantic database of (pay attention!)… no less than 71 Petabytes (in case you don't get the idea, a Petabyte is equivalent to 1015 bytes) to make financial predictions using AI and ML.
The Future of Trading: How Will AI Impact Financial Markets?
As we've just seen, the use of AI in trading has only just begun, and hundreds of potential applications can already be counted. However, the impact all of this may have on the behavior of financial markets in the future is truly uncertain.
For example, proper use of AI will likely improve prediction accuracy and reduce execution costs, which will undoubtedly lead to improved market price formation. But beware, this can be a double-edged sword. Imagine a scenario in which, in a given market, only AIs that could reach a consensus on a company's valuation operated. If that were the case, there would likely be little market movement. And any movement that occurred would be the result of an immediate reaction to new information flows absorbed by the AIs in a matter of milliseconds.
On the other hand, incorrect use of AI can lead to poor decision-making that generates significant losses for investors. It's far from enough to simply input data into a machine and start making money. On the contrary, there are several risks to consider when using these techniques by people without training in this field. The following are particularly noteworthy:
Black box risk: AI systems can be difficult to understand and interpret. This can make it difficult to identify errors and assess the impact of changes.
Risk of overfitting: A data scientist who is not well versed in AI and ML concepts can force their model to overfit to the data supplied for training, leading to the wonderful and promising pattern detected not actually existing, resulting in inaccurate predictions when the model is faced with new data that it has not seen before.
Data risk: Any flaws or biases in the data provided to an AI will inevitably lead to inaccurate predictions. We must not forget that AI learns based on what we teach it. If the data is flawed or unsuitable for a particular algorithm, the results can be simply disastrous. Hence, a large part of a data scientist's job is to collect, preprocess, and curate the data they will use.
What is almost certain is that there will be an increasing number of AI-based financial products and services, whether in the form of algorithmic trading strategies or robo-advisors, further democratizing access to the world of markets, making them more accessible than ever.
This, in turn, will bring a significant increase in activity and trading volumes on the markets, which will undoubtedly boost the profits of intermediaries, who will have to specialize to adapt to this new reality, offering a wide range of services such as easily integrable platforms, data APIs, etc. Precisely in this area, the saying "renew or die" will be more valid than ever. Any broker that fails to adapt to this new paradigm will be doomed to disappear.
Another transformation that the growing use of AI in finance and trading will bring will be the transformation (or, outright, the disappearance) of certain jobs. For example, the speed with which AI performs analysis, generates forecasts, and creates reports will likely end up making the role of the financial advisor redundant, whose "human" forecasts will likely lead to suboptimal returns (unless they also use the services of an AI and don't tell their clients :D).
Conclusion
We are undoubtedly living in truly interesting times, with humanity possibly on the verge of a huge evolutionary leap thanks to the innovation brought about by AI. Visually, as Tim Urban pointed out in his visionary article, The AI Revolution: The Road to Superintelligence, we are at a point similar to the one in the following graph:
Trading will not be immune to this impact, as traders, through AI, will be able to create increasingly sophisticated strategies thanks to widespread access to increasingly comprehensive databases and more complex algorithms. In fact, in the future, traders' results will likely be differentiated by the knowledge and quality of their AIs, as well as by how they obtain and transform data.