AI evaluated as normal technology; billions in funding redirected to other industries

Frontier AI capability measurement improved; improvements cannot be compared to prior measurement

Neurosymbolic systems advance; LLMs remain functionally useful despite theoretical collapse

AI existential risk probabilities are too unreliable to guide policy decisions

AI-generated voices impersonate living musicians on streaming platforms; attribution optional

New research attempts to establish science of agent reliability metrics

Language models now train other language models; distributed scaling continues