The polarization of the discourse on the speed of AI progress & impacts is unhelpful. AI as Normal Technology is a project that advocates a strong view in this debate while also trying to improve mutual understanding between people who disagree. We don’t think those goals are contradictory.

In our latest, we included a “Why we could be wrong” section. We stand by our empirical findings, but we acknowledge that choices about what to measure and how to interpret the results inevitably involve judgment. We welcome other views on the state of AI reliability. My ideal state of AI discourse would be quick convergence on facts paired with robust, good-faith debate on matters of interpretation.