Bridgewater starts $2 billion fund that uses machine learning for decision-making and will include models from OpenAI, Anthropic and Perplexity

Bridgewater Associates, a renowned hedge fund led by CEO Nir Bar Dea, is set to launch a new fund that utilizes machine learning as its primary decision-making tool. The fund, which will begin trading on Monday, will start with nearly $2 billion in capital from more than a half-dozen clients.

The strategy for this fund is based on Bridgewater’s proprietary technology that the company has been developing for over a decade. Notable contributors to the fund’s models include OpenAI, Anthropic, and Perplexity, among others. Greg Jensen, co-chief investment officer at Bridgewater, will oversee the management of the new fund.

Bridgewater has been testing this strategy since late last year with a small portion of its main Pure Alpha fund. The Pure Alpha fund, which has assets under management of $108 billion, has seen a significant improvement this year, climbing 14.4% through June 26 after a decade of mostly lackluster returns.

The push into machine learning is part of a broader transformation at Bridgewater, led by Bar Dea since founder Ray Dalio ceded control in late 2022. This transition also included a major management overhaul.

The use of machine learning has the potential to change the composition of staff at Bridgewater, with an increased focus on hiring data scientists. Jensen, who has been thinking about the impact of machine learning on investing since 2012, has been instrumental in this initiative. He has personally invested in OpenAI and Anthropic.

Jasjeet Sekhon, a statistician and professor at Yale University, was hired by Bridgewater as a chief scientist for the initiative in 2018. The firm also formed a division called Artificial Investment Associate Labs (AIA) early last year.

Jensen acknowledges the limitations of the machine-learning process, emphasizing the need for human intervention for functions such as risk management, data acquisition, and trade execution. He also highlights the potential for large language models to exhibit “hallucination,” lacking human concepts like greed, fear, and understanding of cause-and-effect relationships.

Tests using the AIA systems have included predicting the impact of political events, such as the effect of a Donald Trump victory on asset prices, and calculating the impact of the Federal Reserve’s quantitative tightening process on bond prices. Jensen notes that machines are better at finding patterns across time and across countries.

.st1{display:none}See more