AI agents may handle most crypto trades, but human judgment will remain

AI agents may take over much of cryptocurrency trading and a broad range of daily tasks, but human judgment and user ownership of data will remain essential.

AI agents could execute most cryptocurrency trades without fully replacing people, as individual preferences, emotions and irrational market forces will continue to require human input.

In an Oct. 14 Cointelegraph interview, Animoca Brands Chairman Yat Siu said AI agents would not operate entirely autonomously without human advice or intervention, even if they come to dominate trading.

Siu said people asking an AI the same question would receive broadly similar answers, but their subsequent questions would lead to different outcomes. Asking the system to consider China, the United States or the war in Iran reflects a person’s knowledge and preferences, producing distinct trading models.

“Our agent is going to do most of the trades,” Siu said, adding that agents would neither execute every trade nor operate entirely without human input. People may not need to approve each transaction, but could still direct an agent toward factors it has overlooked.

He said markets often produce trades that cannot be explained solely by logical valuation multiples. Memetic appeal, potential network effects and emotions such as hope and despair can influence prices, leaving room for human intuition in trading decisions.

Siu also emphasized that AI agents have uses beyond trading. Digital entities can manage calendars, book flights, reduce grocery costs and support social interaction, meaning “the metaverse is coming to you” rather than people entering the metaverse.

Blockchain could provide a foundation for users to control their AI agents and data. Siu warned that agents offered by centralized companies may be removed under their terms of service, while personal information supplied by users may be used for model training.

He described submitting sensitive materials such as résumés or bank statements to AI services without safeguards as “crazy.” Users should treat an AI service like any other third party and remove irrelevant personal details, including names, before sharing information.

Siu said the greatest risk is not only the data handed over, but suddenly losing access to knowledge and services on which a person has become dependent. If an agent that has spent years learning about a user’s health, work and market decisions disappears, he warned, it could be worse than losing a smartphone.