The concept of building a startup in your garage is deeply rooted in Silicon Valley lore. Apple and HP famously started from those humble beginnings. With the democratization of AI tools and platforms, it’s entirely possible that a single person or small group could develop a novel AI algorithm application in a metaphorical “garage.” Open-source libraries like TensorFlow, PyTorch, and scikit-learn, combined with affordable cloud computing resources, have made AI development accessible to many.
However, building a functional AI model is different from building a successful, scalable, and robust AI system. While the initial prototype might emerge from a "garage," refining that prototype to a level where it can compete with professional systems requires much more. The amount of resources, both in terms of computational power and domain expertise, for large-scale practical applications can be tremendous.
In this article, we’re going to explore the intricacies and challenges of building AI operations at scale. Our observations are not just based on our own experiences at Kaiju but also draw insights from our peers and partners, as well as the broader market.
The Allure of AI in Investment Management
When talking about AI in investment management, one cannot overlook the pioneering efforts of Renaissance Technologies, one of the world’s most profitable investment managers. As of August 2023, their assets top $100 billion. However, their story started long before “AI” entered the mainstream. It has consistently relied on quantitative models, many of which now incorporate advanced AI in order to predict market movements.
Renaissance, however, is not the sole player in this field. Leading investment firms, such as BlackRock, J.P. Morgan, Goldman Sachs, and Morgan Stanley, are all harnessing the power of AI to employ sophisticated algorithms for tasks ranging from portfolio optimization and sentiment analysis to risk management, and even automated trading systems.
As this revolution in investing unfolds, many are contemplating the role of AI in their own investment management. The question arises: Could an individual investor replicate such success by building their own AI models for investing? The answer is not straightforward and warrants a deeper discussion.
The Heavy Lift of Technology and Data
It’s true AI technology has become more accessible, and continues becoming more accessible. Anyone with determination can start learning about it on their own computer or phone. However, when it comes to professional-level AI, particularly for stock market trading, there’s a catch. Crafting advanced algorithms requires a deep understanding of both the underlying technology and the intricacies of financial markets. Moreover, the costs associated with building reliable AI-based trading systems can be steep.
Creating, testing, training, and validating AI models in finance requires robust infrastructure. Many developers turn to platforms like AWS SageMaker, Google TensorFlow, or Azure Machine Learning, which, while powerful, can incur significant expenses. Additionally, a solid technological foundation is crucial before scaling any AI model to handle real-time market data.
One commonly overlooked challenge is the sheer volume of stock market data. Even when aggregated, stock market data is big. Very big. To illustrate, sending 1 gigabyte of data from New York to London is fairly straightforward. However, increase that to 60 terabytes of data and now you have a problem. Firms must not only source and transport this data efficiently but also ensure it's sanitized, well-structured, and free from errors for use in their models. These requirements lead to the necessity for sophisticated data processing pipelines and storage solutions, which are both costly and complex to maintain. It also means every action needs to be considered carefully and fully understood.
At Kaiju, one critical aspect we manage is the 'temperature' of our data storage. In data management, 'hot' and 'cold' refer to the frequency of data access. The 'colder' the data, the less frequently it is accessed, and thus, the less expensive it is to store. However, the nature of AI models and trading strategies in finance, particularly during the development and iterative improvement phases, often requires frequent backtesting. This means that significant portions of our data need to be kept 'hotter', readily accessible for analysis, which has implications for our storage strategy and costs.
Beyond training and backtesting, there's the hurdle of managing streaming market data in real-time. A fully trained AI model must process this data instantaneously, necessitating a robust system to process data and feed the model accurately and swiftly. And this is all without mentioning the technological intricacies involved in automating trade executions.
The financial commitment of large firms to AI and technology is staggering and illustrates the scale of investment required to maintain the forefront of the industry. For instance, J.P. Morgan Chase invests $12 billion per year on technology. Meanwhile, BlackRock’s signature technology product “Aladdin” is an investment-management platform that leverages AI, and touches $18 trillion of global assets.
Such figures may seem daunting, and they underscore a stark reality: while AI technology is becoming more accessible, the resources required to leverage it at the cutting edge in the financial sector are substantial. However, this doesn't entirely leave smaller firms out in the cold. It's feasible for modest AI operations to operate at a smaller scale, with technology budgets under $1 million per year excluding staff. While this amount may still be beyond the reach of most individual investors, it does open the door for smaller firms to enter the AI arena and find real success.
In essence, the journey to successfully train and scale AI models for trading or financial analysis is fraught with challenges: the costs, the breadth of technical expertise required, and time investment required for data sanitization and management. While new services are emerging regularly to assist with these challenges, effectively leveraging AI for market trading remains a formidable task for most smaller firms.
The Interdisciplinary Backbone of AI in Finance
For as much as AI is about machine-based intelligence, the human element remains irreplaceable. Interestingly, as I began to type the word “irreplaceable,” my machine-based autocomplete suggested “irrelevant” instead. A humorous, yet thought-provoking moment that underscores the delicate balance between human expertise and machine capabilities, and perhaps, the irony of AI inadvertently affirming how indispensable human guidance is. Hopefully.
In the finance sector, crafting high quality AI models takes more than assembling a team of programmers. Firms need to curate a diverse group of professionals who bring a wealth of cross-domain knowledge to the table. This includes researchers focused on mathematically unraveling the complexities of financial markets, data scientists who apply technical expertise to yield actionable insights from complex datasets, financial experts with a deep understanding of market dynamics, software and data engineers who build robust technological infrastructure, MLOps specialists dedicated to ensuring seamless model operations, and support and operations staff who manage the day-to-day.
Staying abreast of the ever-evolving trends in AI, technology, and finance is a continuous learning process that requires a commitment to ongoing education and professional development. Each field has its own depth, language, and nuances. While this specialization allows for efficient intra-disciplinary collaboration, it can pose challenges when bridging knowledge gaps between fields. For instance, a data scientist’s approach to a problem might differ vastly from a financial expert's perspective, necessitating effective communication strategies. At Kaiju we place a strong emphasis on having clear frameworks for communication, including developing a series of cross-disciplinary workshops. Through the years, we've encountered firsthand the intricacies of layering multiple complex topics together.
Moreover, attracting top talent is a competitive and costly endeavor, especially when seeking individuals who can bridge interdisciplinary divides. The time and resources invested into building such a team are substantial, but they are the cornerstone of any successful AI venture in finance. Such synergy is what fosters innovation and drives the industry forward, proving time and again that the human element in AI is not just relevant, but fundamentally irreplaceable.
Circling Back, Looking Ahead
So, can anyone build AI in their garage? Certainly. While this article has highlighted numerous challenges in solo AI development, it also underscores that the most successful AI applications in finance are typically the result of collaborative efforts by large, multidisciplinary teams. However, this doesn't preclude the individual innovator. A good model, nurtured with dedication and the right resources, has the potential to excel.
Artificial intelligence is an incredibly diverse landscape, particularly in the finance sector. On the horizon, we see quantum computing poised to tackle problems of staggering complexity, reinforcement learning systems that refine their own algorithms through trial and error, and explainable AI (XAI) that demystifies the decision-making process. Innovations like AutoML and generative adversarial networks (GANs) are revolutionizing how AI is developed, allowing AI to essentially program itself. And let's not forget large language models like ChatGPT, which are already transforming tasks ranging from financial analysis to the creation of tailored investment portfolios and advanced risk management strategies.
While the idea of single-handedly building a transformative financial AI model may be somewhat romanticized, the field of AI has never been more vibrant or more integral to a multitude of industries. The technical challenges are immense, but increasingly, AI itself is providing the solutions. It's an exhilarating time to be involved in AI. So, whatever your passion or curiosity, now is the time to engage with it. The future is yours to shape.
Investors should consider the investment objectives, risks, charges and expenses carefully before investing. For a prospectus or summary prospectus with this and other information about the Fund, please call (800) 617-0004 or visit our website at dipetf.com. Read the prospectus or summary prospectus carefully before investing.
The Fund is distributed by Quasar Distributors, LLC. Exchange Traded Concepts, LLC (the “Adviser”) serves as the Fund’s investment adviser. Kaiju ETF Advisors (the “Sub-Adviser”) serves as the Fund’s investment sub-adviser.
Investing involves risk, including loss of principal. The Fund is subject to numerous risks including but not limited to: Equity Risk, Large Cap Risk, Management Risk, and Trading Risk. The Fund is actively managed and may not meet its investment objective based on the Sub-Adviser’s success or failure to implement investment strategies for the Fund. The Fund’s principal investment strategies are dependent on the Sub-Adviser’s understanding of artificial intelligence. The Fund relies heavily on a proprietary artificial intelligence selection model as well as data and information supplied by third parties that are utilised by such a model. Specifically, the Fund relies on the Kaiju Algorithm to implement its principal investment strategies. To the extent the model does not perform as designed or as intended, the Fund’s strategy may not be successfully implemented and the Fund may lose value. A “value” style of investing could produce poor performance results relative to other funds, even in a rising market, if the methodology used by the Fund to determine a company’s “value” or prospects for exceeding earnings expectations or market conditions is wrong. In addition, “value stocks” can continue to be undervalued by the market for long periods of time. The Fund is expected to actively and frequently trade securities or other instruments in its portfolio to carry out its investment strategies. A high portfolio turnover rate increases transaction costs, which may increase the Fund’s expenses. Frequent trading may also cause adverse tax consequences for investors in the Fund due to an increase in short-term capital gains. The fund is new, with a limited operating history.
This information should not be relied upon as research, investment advice, or a recommendation regarding any product, strategy, or security in particular. This material is strictly for educational or informational purposes, is subject to change, and is not intended to be a forecast of future events or a guarantee of future results. The information and opinions contained in this material are derived from proprietary and nonproprietary sources deemed by Kaiju to be reliable, are not necessarily all-inclusive and are not guaranteed as to accuracy. Reliance upon information in this material is at the sole discretion of the viewer.