Vitalik Buterin Identifies a Skill Where Humans Have an Advantage Over AI

Vitalik Buterin points out a limitation of AI: despite its extensive knowledge, it struggles to assess the depth of a text, especially when encountering it for the first time. For the co-founder of Ethereum, this difference between knowledge and wisdom deserves particular attention.

In Brief

• On October 29, Vitalik Buterin describes AI as "extremely knowledgeable," but "not very wise," particularly when evaluating a text.
• According to him, models face more difficulties with a new piece of writing that lacks the human commentary present in their training data.
• He finally emphasizes an essential difference: one can easily verify a model's knowledge, but measuring its wisdom is much more challenging.

AI Knows the References, But Does It Understand the Ideas?

Vitalik Buterin regularly expresses his reservations about AI agents, as he did in October with his "paranoid mode." On October 29 at 4:11 UTC, he published a new reflection in response to writer Ruxandra Teslo.

The discussion followed an article from the Wall Street Journal about Humanities 10, a first-year course at Harvard that prohibits students from using phones and computers for a year.

In his post on X, Vitalik explains that AI can mobilize knowledge on almost any subject, yet it does not demonstrate deep judgment. It risks stopping at the most superficial aspects of a text without fully grasping the scope of its argumentation.

This remark comes amid a debate about the quality of exchanges with chatbots. Some consider them better conversationalists than most humans for discussing the great classics. Conversely, a writer believes that their comments on her essays often boil down to hollow responses.

The First Reading

For Vitalik Buterin, this limitation becomes particularly evident when the model encounters a text for the first time. Indeed, a famous work benefits from decades of critiques, analyses, and commentaries that AI can find in its training data. A recent essay, on the other hand, lacks the same historical context.

The model must therefore rely more on the text itself, without being able to draw on as much pre-existing analysis. This difference is particularly relevant in crypto, where teams use AI agents to examine white papers, audits, or governance proposals.

However, some developers are already warning about the flaws that AI could exploit without fully understanding the implications. Under these conditions, summaries produced by a model should not replace human verification, especially when technical or financial stakes are high.

Wisdom, a Quality Difficult to Measure

Vitalik Buterin finally raises a more subtle issue: a model's knowledge can be verified quite easily, unlike its wisdom. One can check a date, a reference, or a figure. In contrast, judging the relevance of reasoning requires more than a simple factual check.

An episode from July illustrates this distinction. A contest organized around identifying the author of an anonymous Ethereum document rewarded, on July 6, the team of Franklyn Wang from Co-Invest. After examining 27 texts, they achieved a result associated with a probability of about 20%.

This experience reminds us that artificial intelligence can recognize stylistic characteristics without necessarily grasping the ideas behind a text. The same caution applies when analyzing code or technical documents: identifying patterns does not always mean understanding the intentions of their author.