Dalio claims that no AI has sufficient criteria to be followed without question.
He also maintains that in investments, what is widely known has little differential value.
Ray Dalio, founder of the Bridgewater Associates hedge fund and manager of its family office (DFO), published an analysis last Thursday in which he argues that artificial intelligence, by itself, is not enough to make effective investment decisions.
According to Dalio, AI must operate in conjunction with logical and understandable criteria developed by humans.
In the text, Dalio argues that even the most advanced AIs lack sufficient criteria to be followed without question. The investor points out that human understanding remains essentialparticularly in markets where value added is a zero-sum game: what everyone knows, he argues, has little differential value.
Dalio suggests that the right approach is to develop what he calls “principled thinking.”: a process in which the investor examines and systematizes his decision criteria, documents them, subjects them to historical tests and automates them so that they operate in conjunction with human reasoning.
The model proposed
According to Dalio, The ideal system works like a computerized chess program– Makes decisions independently, but always with logic behind each visible and debatable move. The investor indicates that this allows the human and the AI to correct each other and align their reasoning.
The analysis specifies that these criteria should not be derived from data mining—that is, identifying historical patterns and assuming they will repeat themselves—but from logical understandings of the cause-and-effect relationships that govern markets. Dalio maintains that this approach allows us to process more complex relationships. with greater speed and without emotional biases.
The founder of Bridgewater points out that he has been developing this process for 50 years, in which he applies the most recent AI technologies. Dalio warns that those who are not at the forefront of this model of integration between human and artificial intelligence will be at a competitive disadvantage.
For Dalio, the key is not to choose between human judgment and the processing capacity of AI, but to build systems where both feed each other, with the logic always exposed and subject to review.
AI in prediction markets: trend with documented risks
Dalio’s warning coincides with a trend already verified in cryptocurrency markets. In February 2026, a user identified as Argona reported that an artificial intelligence agent converted $50 into nearly $3,000 in 48 hours trading on Polymarket, the prediction markets platform.
The bot scanned between 500 and 1,000 markets every ten minutes and executed orders when it detected price deviations greater than 8%.
However, the popularity of these systems also attracted fraud. As reported by CriptoNoticias, criminals began to distribute malware under the guise of bot installation tutorials for Polymarket, using commands that downloaded malicious code capable of emptying cryptocurrency wallets.
The attackers displayed verifiable on-chain balances to build trust, although the download link had no technical relationship to the displayed accounts.
The pattern illustrates Dalio’s central argument: automation without verifiable criteria not only limits performance, but exposes users to risks that autonomous systems, on their own, are not in a position to anticipate.