A final report from a U.S. government task force seeks $2.6B in funding for a federal AI data-and-research hub.  The report covers plans for a National Artificial Intelligence Research Resource (NAIRR), which would be a publicly available AI cyber-infrastructure for researchers and students to access data and other resources. A White House-led task force said that the infrastructure is needed to further democratize AI, which has so far been restricted to "well-resourced" entities like Big Tech corporations. The NAIRR task force says the initiative would cost $2.6B over six years. If enough funding is appropriated, the task force believes the NAIRR could be up and running in an initial capacity within 21 months. It believes the resource could be rolled over three years across four planning stages, with "steady-state operations" occurring in the final year. What's needed: The report lists recommendations for a NAIRR infrastructure, which includes computational resources like servers and cloud computing. It also calls for a large-scale ML supercomputer that could train 1 trillion-parameter AI models. It estimates that training a 1.5 billion-parameter model would cost roughly $1.6M. The task force also recommends that an entity to oversee NAIR operations be "a distinct, non-government organization," though resource providers that are hired could be commercial and private entities. It suggests that NAIRR tap into the massive amount of federal data that's already stored in commercial clouds.