During a session at the 2026 Inclusion·Bund Conference, 2011 Nobel economics laureate and New York University professor Thomas Sargent issued a cautionary note, stating that the massive investments in U.S. AI are based on frameworks whose ultimate outputs remain unverified. He observed that while AI has become a central driver of economic growth, the path of investment, corporate decision-making, and policy formation is riddled with a profound level of uncertainty; the expansion of capital expenditure does not, in his view, guarantee a corresponding clarity or security of future returns.

This gathering featured economists, industry experts, and investors discussing AI's relationship with productivity and investment strategy. Sargent argued that to comprehend the economic shifts underway, one must accept a specific reality: we are in a phase where AI’s importance is obvious, yet its tangible impact remains uncharted. To illustrate, he drew a parallel to the scientific evolution of the 17th century, contrasting the work of Kepler and Newton. Kepler discerned planetary motion patterns from vast data sets without grasping the underlying physical cause, whereas Newton established the foundational principles behind those movements.

“Today’s AI is a bit like Kepler,” Sargent explained. He points out that while modern AI excels at detecting patterns and making predictions from enormous data pools, it does not equate to comprehending the causal mechanics of the world. Current progress stems from unprecedented volumes of data, computing power, and increasingly intricate models. Yet, the true frontier lies in transitioning to an equivalent of the "Newton stage" where the technology can infer underlying structures and generalize beyond its training sets and recognize its own boundaries. He confirmed that we do not yet know when this progression will occur.

Sargent noted that the arrival of the AI economy does not necessarily herald a predestined future. On the contrary, as AI integrates deeper into finance, research, and public policy, the ability to manage this pervasive "uncertainty" is becoming the defining economic challenge. He warned against the misconception that neural networks are model-free learning systems. Every algorithm embeds specific parameterized models and loads of implicit assumptions about the world. A model performing well in familiar environments may falter when faced with conditions outside its training data, highlighting the key distinction between risk and uncertainty in economics. Risk implies unknown outcomes with roughly estimateable probabilities, while uncertainty suggests we cannot even foresee the range of possibilities—a category into which many AI-induced changes fall.

This concern connects directly to Sargent’s research on "robust control" theory, which seeks to address such decision-making dilemmas. The approach avoids searching for a single "correct" model, aiming instead to ensure that decisions remain reliable even when the model is flawed. It asks whether a choice can withstand errors in assumptions, not just whether it works when the model holds true. This principle, he suggests, is vital for the nascent AI economy—where productivity changes are certain, but the degree of improvement, the location of value creation, the restructuring of industries, and the final distribution of gains are entirely unresolved.

Concluding his remarks, Sargent reinforced the need for "humility." He emphasized that enterprises, investors, and regulators must accept the limits of their frameworks, allocating room for error and establishing resilient mechanisms to handle multiple possibilities. He framed this as the flip side of the AI revolution: it is not just a tool for prediction, but a teacher reacquainting us with the scope of our ignorance. In this light, the rarest commodity for leaders is not increasingly accurate forecasting, but the ability to acknowledge the unknown, reserve caution, and make prudent choices amidst opacity. While machines may be getting smarter, humanity’s first task may be to admit how little we still know.