Microsoft and OpenAI are planning to announce Tuesday that they are helping to launch an AI training center for members of the second-largest teachers' union in the US, according to details about the initiative that appear to have been inadvertently published early on YouTube. The National Academy for AI Instruction will be based in New York City and aims to equip kindergarten up to 12th grade instructors in the American Federation of Teachers with "the tools and confidence to bring AI into the classroom in a way that supports learning and opportunity for all students," according to the description of a publicly accessible YouTube livestream scheduled for Tuesday morning. The YouTube page also lists Anthropic, which develops the Claude chatbot, as a collaborator on what's described as a 22.5 million initiative to bring free "AI training and curriculum" to teachers. The three AI companies and the union did not immediately respond to requests for comment about the information released on YouTube. On Monday, Microsoft and the union declined to share details ahead of an announcement planned for Tuesday morning in New York.

Understanding and manipulating the causal generation mechanisms in language models is essential for controlling their behavior. Previous work has primarily relied on techniques such as representation surgery -- e.g., model ablations or manipulation of linear subspaces tied to specific concepts -- to \emph{intervene} on these models. To understand the impact of interventions precisely, it is useful to examine counterfactuals -- e.g., how a given sentence would have appeared had it been generated by the model following a specific intervention. We highlight that counterfactual reasoning is conceptually distinct from interventions, as articulated in Pearl's causal hierarchy. Based on this observation, we propose a framework for generating true string counterfactuals by reformulating language models as a structural equation model using the Gumbel-max trick, which we called Gumbel counterfactual generation. This reformulation allows us to model the joint distribution over original strings and their counterfactuals resulting from the same instantiation of the sampling noise. We develop an algorithm based on hindsight Gumbel sampling that allows us to infer the latent noise variables and generate counterfactuals of observed strings. Our experiments demonstrate that the approach produces meaningful counterfactuals while at the same time showing that commonly used intervention techniques have considerable undesired side effects.