- How AI fits in investment management?
- Artificial Intelligence (AI) and Machine Learning (ML) algorithms have long been used by hedge funds in algorithmic trading and systematic investing. As technology companies are getting into financial business, financial institutions are also evolving to technological companies. JPMorgan Chase, one of the largest financial institutions in the world, is actually also one the largest technological companies. It has ~1000 employees in data management, 900 data scientists, 600 ML engineers, and a team of 200 people in AI research group. Processing 25-years of Fed speeches, JPMorgan found, that AI may predict policy changes, which can be used in trading. Currently, the bank has more than 300 AI use cases in production. Even Bloomberg has recently created a 50-billion parameter model – BloombergGPT – trained on market data, to perform NLP tasks in finance industry.
- Does AI help to make better investment decisions for general public?
- There has been a lot of interest in AI since the launch of ChatGPT, a deep neural network designed for processing sequential data. There is no proven track record, which is based on real market data, but a fintech company Finder.com conducted a research on the ability of ChatGPT to give an investment advice. They asked it to create a portfolio of stocks of high-quality businesses, that can compete with 10 benchmarks, among which were Vanguard US Equity Index, HSBC FTSE All-World Index, Fidelity Index World and others. ChatGPT initially warned that it was not trained to give an investment advice, but researchers were able to easily bypass it. Analyzing such parameters, as the level of debt, sustained growth in the past and factors, that generate a competitive advantage, among others, and assuming equal weighting, ChatGPT chose 38 stocks: Cisco, Microsoft, 3M Company, Mastercard, Nestle, United Health Group, Walmart, etc. In the result, ChatGPT was able to outperform benchmarks in 34 of 37 observed market days between March 6 to April 28, 2023, gaining +4.9%, while benchmarks had on average -0.8%. In the same timeframe, S&P500 gained +3%, and Stoxx Europe 600 got +0.5%.
Despite the excitement, it is important to note, that ChatGPT has not been trained on market data, although there are already many researches, demonstrating that it has statistically significant predictive power on stock returns and outperforms more traditional sentiment analysis algorithms.
- Can AI influence market behavior, distorting fundamentals?
- AI may influence market behavior, and it can trigger further spiraling reaction of the market. It has been argued, that the Dow Jones Industrial Average’s drop by 1 600 points on February 5, 2018, the largest in the history at the time, was not only caused by the fear of interest rate hikes and the U.S.-China trade war, but also due to the glitch in computerized high-frequency trade, which shorted an ETF, linked to the volatility index. As a result, the sponsor of ETF was triggered to buy a large amount of futures to balance the exposure. The selling caused others to cover their short positions and the following selling reaction of the rest of the market. However, such cases are rare. AI/ML may make incorrect decisions, but it will make it less frequently, than humans.
- What are the main drawbacks of using AI/ML in investments?
- AI/ML models are highly depended on the quality of the input data, and could be biased if certain statistical requirements are not met. Also, past performance is no guarantee of expected results, as there is a possibility that previous patterns will not persist in the future. As they require continuous refining and tuning, such models can become very complicated to interpret the results. Although facilitating to stay informed, AI may also push to be less attentive and overly rely on models, the principles of which are not fully understood.
- How AI could be applied in risk management?
- ML and AI can help mitigate risks by creating synthetic data to test all kinds of scenarios, including severe ones. It allows to conduct more granular risk assessment and mitigates human error. Though, one of the concerns of AI application is that it can get too much power. Because of the tendency of optimizing decision-making, it is entrusted to make decisions for which it will bear no responsibility. That is the reason many regulatory agencies are calling to regulate the use of innovative technology, ensuring traceability and accountability. This even evolved in new branches of risk management, using suptech (financial supervisory technology to digitize regulatory strategies) and regtech tools (technology to ensure compliance, conduct regulatory risk calculations and reduce human error).
- What are the prospects of wide adoption of AI/ML in management of alternative asset classes?
- Hedge funds, pursuing quantitative strategies, employ physicists and computer scientists to develop domain-specific models, and economic reasoning is a prerequisite for their efficiency. In private markets, the application of AI/ML has been in gathering and analyzing a large amount of unstructured data, that is accumulating at accelerating pace. Yet, unlike public markets, the private market often has incomplete, opaque data. There is less availability and accessibility of data, or there is a time lag between the reporting date and the date the data is actually disseminated for further use. Thus, there is currently less understanding how AI can facilitate investments in alternative assets, beyond traditional data management, clusterization and classification tasks. The wide adoption of innovative technology is also limited by the computing power and data storage capabilities, that are in fact, one of the niches, increasingly financed by private capital. Thus, in the near future, such investments will start benefitting business processes and decision-making in the private market at a large scale.