AI may be the next Warren Buffett but there are challenges ahead

From a disruptor to an integral part of our lives, artificial intelligence (AI) is the defining technology of our times, one that is rapidly changing how we work, consume and invest.

Its ability to sought, analyse and interpret huge amounts of data efficiently and quickly has seen it being deployed for trading and getting investment ideas. In most advanced markets, AI trading is gaining ground and is being used to develop trading strategies.

“Globally, asset management is increasingly being defined by artificial intelligence and machine learning (ML). Funds run by computers account for more than 60 percent of US trading activity,” said Kanika Agarrwal, CIO, Upside AI, which does machine-learning-based investing.

AI is unbiased, unemotional and has no specific style of investing. A good algorithm is dynamic and can test and refine trading strategies looking at market trends. It can do what analysts do—gather information, data and evaluate them to come up with suggestions.

“It can consistently find alpha, which is difficult for human managers who thrive in some market cycles but not in others,” said Agarrwal.

“Globally trading activity (short-term, high frequency, technicals, etc.) is largely run by technology now. Even passive ETFs have surpassed active managers in AUM in the US,” Agarrwal said.

She is of the view that the next step up for AI will be fundamental investing.

“We believe the next great investors like Warren Buffett and Charlie Munger are going to be AI. Benjamin Graham and Warren Buffett have been big believers in investing using systemised rules and staying away from emotions. The best way to follow rules unemotionally is the use of AI and technology,” Agarrwal said.

While developed markets seem to be rapidly getting acclimated to this technological revolution, emerging markets like India will take some time to see its dominance in investing.

“In India, while investing is largely people-driven, I believe we will see structural shifts in the next decade as our markets mature and alpha becomes more difficult to find. We will increasingly see more products like ours trying to find different approaches to investing in using AI,” Agarrwal said.

AI’s biggest strength is its absolute dependence on data and uses algorithms to understand the market and its cycles. This is the reason AI is being used in the investment and wealth management industry globally.

Mihir K Malani, Founder of FinTech startup, Nerve Solutions pointed out while the eventual aim is always to improve returns, the process usually involves multiple steps like picking the right stocks based on historical trends, deciding upon the investment size, identifying and predicting trends, etc.

“A commonly adopted approach is to use ML models to categorise clients based on their profiles, investment preferences and risk appetites and letting the model arrive at the most suited investment strategies for them,” Malani said.

One of the biggest advantages of a well-developed AI model for investment is its ability to avoid pitfalls and predict drawdowns successfully, he said.

“Also, a good way to measure the efficiency of a model is through the number of false positives generated by it. The lower the number, the better and more trustworthy the model,” Malani added.

The challenges

Even though AI appears to be an inevitable force in investing, there are challenges ahead.

While it is easy to access large sets of structured financial data to build machine-learning models, there are several challenges associated with developing a model that works.

Agarrwal of Upside AI believes that the challenges of AI are true across industries, including investing–data quality, the quality of the model built, lack of talent in India to build these machines, solving qualitative investing issues like corporate governance and wider acceptance of the technology.

Differentiating genuine patterns and coincidences is among the biggest hurdles one may face while applying AI to investing.

“At times, mere coincidences give an illusion of correlation. Inability to identify these could lead to extremely inaccurate models,” Malani said.

Identifying the right features and then designing the model is a challenge.

“This is a common challenge often faced while developing AI models for derivative instruments. With a number of factors involved in the pricing of a derivative contract, missing out on the right features could lead to seemingly correct but erroneous models,” Malani said.

Besides, embedding geopolitical factors within a model is also a challenge.

“While incorporating price and volume information within an AI model is fairly procedural, it is extremely difficult to account for factors that lie outside of the markets but have a profound impact on market movements. Such factors often are the reason for models to fail,” Malani said.

Despite these challenges, AI is the future of investing, one that opens up a lot of possibilities for investors and managers alike.

Leave a reply

Please enter your comment!
Please enter your name here