The Alignment Problem from a Deep Learning Perspective

Richard Ngo, Lawrence Chan, Sören Mindermann

In coming years or decades, artificial general intelligence (AGI) may surpass human capabilities across many critical domains. We argue that, without substantial effort to prevent…