New research suggests that AI-driven automation may take longer to diffuse throughout the economy than many predict. The reason: It’s expensive.
A wave of studies modeling the potential impact of AI on jobs gave us a sense of the potential magnitude of disruption, but told us less about the time horizon for the changes. Now a new working paper addresses the “when” question by looking at the economics of substituting AI for humans in vision tasks, which the authors defined as tasks where “there is a use case for computer vision, and image recognition in particular” (analyzing x-rays is an example of a vision task, as is checking the pricing on retail items) . “Overall, our findings suggest that AI job displacement will be substantial, but also gradual—and therefore there is room for policy and retraining to mitigate unemployment impacts,” write the authors.
We spoke with one of the researchers, Neil Thompson, director of the FutureTech research project at MIT’s Computer Science and Artificial Intelligence Lab, about this working paper and what it tells us about AI adoption more broadly. Here’s an excerpt from our conversation, edited for length and clarity: