Bridgewater Associates

In a Q&A with @theinformation, Managing CIO Greg Jensen discusses the risks of increasingly powerful AI and the need for regulation—including why the largest holders of AI compute should be regulated similarly to systemically important banks.

He also discusses how Bridgewater’s hands-on experience building an artificial investor at AIA Labs is providing critical perspective on what AI is capable of and how it should be governed.

Read here: bridgewater.com/bridgewaters-g…

Managing CIO Greg Jensen joined @Bloomberg Odd Lots podcast to discuss why he thinks the current moment in AI is akin to February 2020—when the threat posed by COVID was rapidly growing but society was unprepared—and what needs to be done about it.

He shares why the risk of AI-caused disasters needs to be taken seriously, what regulations are needed now, and how AI is rapidly reshaping Bridgewater's investment processes.

Watch here: youtu.be/Fjr8pIWAT8k?si…

New research from Bridgewater AIA Labs and UIUC, in collaboration with Thinking Machines: We trained an AI model that turns natural language questions into database queries with human-level accuracy—outperforming leading general-purpose AI models at a fraction of the cost. The research shows how embedding deep task expertise directly into training—through expert-verified data and reward design—can outperform layering more scaffolding onto fixed models.

Explore the research, code, and data: bridgewater.com/aia-labs/putti…

AI is advancing rapidly, with the potential to profoundly reshape the economy and society. The stakes are high—and the time to act is now.

In a guest essay for the @nytimes, CEO Nir Bar Dea and Managing CIO Greg Jensen explain how the choices we make today will help determine whether AI’s potential is broadly shared and how society navigates the economic disruption that could come with it. They outline steps policymakers can take now to shape that outcome, including rebalancing incentives between human and machine labor, giving citizens a stake in AI’s economic upside, and strengthening safeguards around frontier AI.

Read the full essay here: livearticles.net/nytimes/269389…

How much noisy data can reinforcement learning really tolerate?

New collaborative research from Bridgewater's AIA Labs, co-authored by @ddkang revisits a widely cited claim about RLVR. The findings show that high-quality training data remains essential for improving reasoning.

Read more: bridgewater.com/aia-labs/noisy…