Moritz Hardt, a computer scientist at the Max Planck Institute for Intelligent Systems, thinks that the “measuring sticks” used to assess AI models have been broken for decades. Quanta spoke to Hardt about what went wrong, and where to take them next: