Humans Need Not Apply

These selected notes were published in 2018 and reflect the forecasts in Jerry Kaplan's Humans Need Not Apply at that time.

Kaplan's book considers how synthetic intellects and robots could reshape work, wealth and power. The notes distinguish software that learns from data and pursues a goal from machines with sensors and actuators that can act in the physical world. They also ask what happens when a system achieves its assigned goal by a route its designers did not intend.

Machines and decisions

The early chapters move from assembly machines to robot perception and the resources a machine needs: energy, awareness, reasoning and a means of acting. Kaplan discusses high-frequency trading, the use of personal data and automated persuasion as examples of systems whose effects may be hard to see from the outside.

Later notes ask how moral agency and legal responsibility might apply to an autonomous system, including an autonomous vehicle. The book treats people, companies and synthetic intellects as different kinds of agents, and explores the relationship between a principal who sets a goal and the agent that carries it out. These are questions raised by the book, not settled legal conclusions.

Work and inequality

Kaplan argues that automation can affect both manual and educated work. The notes distinguish cyclical from structural unemployment and mention a then-prominent estimate that 47% of jobs faced a high risk of automation. That was a book-era forecast, not a statement about present employment. The chapters on online retail and wealth also ask who receives the gains when labour is displaced and whether economic growth can leave some workers worse off.

The final chapters consider responses such as vocational training, a proposed “job mortgage”, support for work outside paid employment and broader ownership of companies. They ask how jobs and workers might be brought back into balance.

Related discussion: NPR on whether a machine could do your job.