Ethereum co-founder Vitalik Buterin expressed concern about current trends in the development of artificial intelligence (AI), particularly the race to create maximally autonomous agent-based models. In his view, prioritizing agency leads to a detachment of humans from the decision-making process, which can affect the quality and safety of the results generated by AI.
This is reported by Business • Media
Buterin: AI Must Listen to Humans
Buterin emphasized the importance of providing more opportunities for human intervention in the functioning of artificial intelligence. He highlighted that creating additional pathways for human involvement in AI processes not only enhances the quality of results but also makes the system safer.
“It frustrates me that many developments in the field of AI strive to be as ‘agent-like’ as possible, while in reality, creating more pathways for human contribution both now and in the future yields better results and is safer,” Buterin stated.
Additionally, Vitalik Buterin noted that he is more attracted to AI models that have open weights and broad editing capabilities, rather than those that operate exclusively autonomously and generate content independently. Such approaches, in his opinion, promote greater transparency and user control.
The Future of AI: BCI and New Interaction Architectures
In the medium term, Buterin sees potential in the development of brain-computer interfaces (BCI), which will allow real-time reading of the user’s emotional state during content generation and adapt the output according to their reactions. This aims to make the interaction between humans and AI deeper and more intuitive.
Vitalik Buterin’s position was supported by Tim, founder and CEO of OpenServAI. He emphasized that the future of AI lies in architectures capable of transforming user intentions into valuable actions, enabling more effective coordination of complex tasks.
“New architectures that convert user intentions into valuable actions are the only viable path for development, in my opinion. This is what we are working on at OpenServAI, applying directly to agents for coordinating complex tasks.”
Discussion participants also mentioned relevant industrial cases of agent-based AI, particularly during the recent online event Incrypted Online Marathon 2025, where David Shwed, CISO of brokerage and financial sectors at Robinhood, shared his vision for the development of such technologies.