An early warning system for novel AI risks - Google DeepMind
AI researchers already use a range of evaluation benchmarks to identify unwanted behaviours in AI systems, such as AI systems making misleading statements, biased decisions, or repeating...| Google DeepMind
