Buterin: AI Has Vast Knowledge but Lacks Wisdom

Ethereum co-founder Vitalik Buterin described AI as highly knowledgeable yet not very wise. Chatbots, he said, can spot references to almost anything but form only superficial judgments about what a written work argues.

His remarks were part of an X debate on whether chatbots now match top human readers. Crypto traders who feed whitepapers and governance proposals into AI and act on the summaries have reason to care about that gap.

A Harvard Course Banning Chatbots Started the Discussion

The thread originated from a Wall Street Journal article about Humanities 10, an intensive Harvard course for first-year students. Excerpts shared on X indicated that applications were up 40% this autumn, with 180 students vying for 90 spots.

Every week, students must read a weighty book, and the course bans electronic devices. Some applicants mentioned they wanted a remedy for AI-facilitated shortcuts.

One X user countered that current models are better than any human in discussing the classics because they have perused everything. However, an author objected, labeling chatbot feedback on her essays as “slop” at this point.

The Element Vitalik Buterin Finds Missing in AI Evaluation

Buterin entered the thread and differentiated knowledge from wisdom.

I find AI to be extremely knowledgeable, and so exceptionally able to detect references to almost anything, but not very wise, and so it only makes the most surface-level judgements about what point some writing is making.

This is particularly true when it's analyzing writing…

— vitalik.eth (@VitalikButerin) October 29, 2026

AI's knowledge, Buterin suggested, is simpler to identify than its restricted wisdom, because wisdom, in contrast to intelligence, may be inherently difficult to verify.

He added that the flaw becomes larger when a model sees a text for the first time. Famous works, on the other hand, are accompanied by abundant human analysis in the training set.

Buterin's warning follows a familiar pattern. Lately, he warned that AI might harm crucial areas of cryptographic security within two years. He also built a paranoid mode for AI conversations.

Not everyone is as doubtful as he is. Some investors argue that AI already surpasses the finest human intellects in mathematics.

Meanwhile, crypto groups keep delegating audits, research, and governance summaries to AI agents. A cursory reading in those contexts could fail to catch a latent flaw in a smart contract or a token's supply schedule.

Buterin's point indicates that verifying facts against a reference is straightforward, while forming an opinion about a wholly new document is not. His recent endorsement of an AI model that limits hacking shows the same tendency to keep human oversight.