AI transition hinges on human choices, not just tech, argues Oxford professor

The future of work in the age of artificial intelligence will be determined by human decisions and cultural evolution, not solely by technological advancements, according to Oxford University professor Carl Benedikt Frey. He emphasizes that automation's pace and impact will depend on consumer demand, what people choose to automate, which is itself influenced by evolving cultural norms.

Frey advocates for a model similar to the 'flexicurity' system employed in Northern European countries. This approach combines a flexible labor market with a robust social safety net, aiming to support jobseekers through the transition. However, he cautions against allowing the pursuit of inclusiveness to lead to stagnation, warning that the drive for equality should not stifle "creative destruction."

“

A key message going forward is that automation will depend on technological capabilities, but it will also depend on consumer demand, what we want to have automated or not, and that, in turn, depends on culture and how that culture evolves over time.

Frey is scheduled to deliver a keynote address titled “Designing the Future of Work in the AI Transition: Conditions for Good Jobs and Sustainable Progress” at the fifth annual Hankyoreh Human and Digital Forum. The event, themed “The Future of Work and Learning in the AI Era,” will take place on June 24.

In a previous interview, Frey recalled his 2013 paper predicting that 47% of U.S. jobs were at risk from automation. He clarified that the study considered a broad range of automation technologies beyond just AI, including robotics. While acknowledging that generative AI has pushed the boundaries of what was considered creative or complex for humans, Frey maintains that the core bottlenecks to automation, complex social interactions, creativity, and navigating unstructured environments, remain relevant.

“

I think for the most instance, yes, AI and LLMs have pushed the boundaries a bit in these bottlenecks and expanded the potential scope of automation a bit. But these bottlenecks we thought of back in 2013 are still relevant today.