New research from MIT FutureTech shows that human workers may be safer from short-term AI automation risks than previously thought. Although AI, particularly computer vision, could theoretically automate many tasks, the study suggests that only a quarter of these tasks are economically viable for firms to transition to AI.

The study indicates that the rate of AI adoption in the medium term will hinge on two main factors: the rate at which AI deployment costs decrease, and the speed at which companies develop AI-as-a-service platforms for widespread economic use.

For example, if privacy concerns can be addressed, one could easily imagine a few AI providers reading x-rays for much of the health system.

The authors anticipate the rise of numerous AI-as-a-service platforms, driven by the economic advantages of large-scale AI deployments. This shift will necessitate firms outsourcing to AI providers many of tasks traditionally done internally.

Paper: Video:

@MITFutureTech Viable for now ... until it goes to scale.

@MITFutureTech And At what point does your digital assistant create more work than it saves you?