Vitalik Buterin Expects Harmful Aftermath of AI Usage: Details

Ethereum’s co-founder and frontman, Vitalik Buterin, has joined a discussion on the X social media platform about artificial intelligence tools and the manipulation they may conduct on people in the future when it gets better developed and improved in many ways.

Negative AI impact expectations

An X user, Eliezer Yudkowsky, published an X post, stating that now, with billions of dollars being given to improve AI capabilitie,s will likely allow bots like ChatGPT “conduct intelligent, adaptive assaults on vulnerable minds,” i.e,. manipulate their opinions. Perhaps, when people use AI as an online advisor or consultant, including psychological problems. Yudkowsky called this “AI psychosis.”

However, there is also “defence” against this future, he believes, and that is “a few people writing static webpages for free.”

Vitalik Buterin weighs in with a pessimistic view

The Ethereum co-founder left a comment under Eliezer Yudkowsky’s tweet, stating that he is open to a bet on how much harm future active usage of AI-bots can cause.

I would be open to a bet on the future extent of AI-driven psychosis, I think I’m more skeptical than you (especially the risk that it would greatly degrade a plurality of short-term decision-makers, ie. not 20 year olds, and not even voters)

Do you want to propose something?…

— vitalik.eth (@VitalikButerin) July 23, 2025

“I would be open to a bet on the future extent of AI-driven psychosis,” Vitalik said. He believes that Yudkowsky might be a little too optimistic, while Buterin himself remains rather skeptical. He said that one of the harmful things that he expects to happen thanks to AI is: “the risk that it would greatly degrade a plurality of short-term decision-makers, ie. not 20 year olds, and not even voters.”

He suggested that Yudkowsky would make any proposal as to what can be done about it; however, this proposal should be made, per Vitalik, “timed at hitting a capabilities milestone so we’re not accidentally betting on timelines.” Basically, once AI bots get to a new level where they are able to have an actual visible impact, without predicting when exactly it will happen, since the former is much easier than the latter.

Source