There’s Conjecture’s Priorities for the UK Foundation Models Taskforce. This is more of a big picture list of actionable suggestions, without much discussion of reasons for choosing them, though most of it is self-explanatory. Some related ideas were discussed in The 0.2 OOMs/year target and just yesterday in Compute Thresholds: proposed rules to mitigate risk of a “lab leak” accident during AI training runs.
There’s Conjecture’s Priorities for the UK Foundation Models Taskforce. This is more of a big picture list of actionable suggestions, without much discussion of reasons for choosing them, though most of it is self-explanatory. Some related ideas were discussed in The 0.2 OOMs/year target and just yesterday in Compute Thresholds: proposed rules to mitigate risk of a “lab leak” accident during AI training runs.