‘AI already beating humans at…’, says Google DeepMind Chief Scientist Jeff Dean

Google Deepmind

Artificial intelligence systems have reached a stage where they may already be outperforming people in many daily tasks, according to Google DeepMind’s Chief Scientist Jeff Dean. While Dean avoided using the term Artificial General Intelligence (AGI), he suggested that AI models have surpassed human ability in several areas, with some important caveats.

AI compared to human ability
Speaking on the Moonshot Podcast, Dean explained that current AI models often demonstrate stronger performance than the average individual, particularly in problem-solving and reasoning tasks that do not involve physical work.

“AI today is better than the average person at most tasks,” Dean said. “Most people are not that good at a random task if you ask them to do something they’ve never done before, and some of the models we have now handle those challenges fairly well.”

Despite these gains, Dean emphasised that AI does not yet match top human expertise. He pointed out that while models are competent across many areas, they continue to make errors and cannot yet match the depth of highly trained professionals.

“They will fail at many things; they’re not at an expert level in some cases,” he clarified.
Dean argued that the potential of AI lies in accelerating scientific and engineering advances. He noted that automated research and large-scale computation could expand the pace of discovery in the coming decades.

“There will be many domains where automated search and computation accelerate progress — scientific progress, engineering progress,” Dean said.
Steering clear of AGI claims

Dean also chose not to speculate on the timeline for AGI, saying the term is too loosely defined. His remarks contrast with DeepMind CEO Demis Hassabis, who recently predicted AGI could arrive within the next five to ten years.