Ethereum co-founder Vitalik Buterin recently commented on the development of artificial intelligence, stating that AI is gradually surpassing humans in more tasks in ways people may not notice. He believes that the previous external tendency to measure the gap between humans and AI using a single standard is inaccurate.
Don't just look at "smart"
Buterin said that evaluating AI progress cannot rely solely on whether it is smarter than humans or on economic output alone. He believes a more appropriate approach is to assess how many types of tasks AI can reliably accomplish and which scenarios these capabilities cover.
He gave the example that humans naturally perform differently across various domains. The same is true for AI—it may excel in certain cognitive tasks but still differ in physical abilities, emotional understanding, or complex judgment.
From a few specialized areas to more domains
Buterin also mentioned the machine age around 1500, when water wheels and windmills could only outperform humans in a few specialized tasks, while humans still held the advantage in most areas. He believes that today’s AI is following a similar path, but at a much faster pace and with broader scope.
He said that AI is not replacing humans all at once, but rather continuously entering new domains and gradually approaching or even surpassing human capabilities in one area after another.
- Buterin believes that advancements in AI should be measured by the scope of tasks they can cover.
- He emphasized that a single "smart" metric is insufficient to describe AI changes.
- He compared the current expansion of AI to the historical evolution of machine capabilities.
Measure progress by the number of tasks completed
In Buterin’s view, the evolution of AI is less about whether it has already surpassed humans in every area, and more about whether it can continuously enter more domains that were previously dominated by humans. As model capabilities improve, this range will continue to expand.