Pure-mathematics background, working on AI safety and governance.
Work and writing
-
What Australia should ask of frontier AI labs that train here
Australia should attach two low-cost conditions to frontier AI training here: labs should share the safety information they already give other governments, and let Australia's safety agencies test new models early. Labs that meet them should be rewarded.
-
Which representations do small models learn for finite-group multiplication?
Small models trained to multiply in a finite group learn only a compressed, near-minimal slice of the group's representation theory, far short of the whole regular representation. To work out which slice, and to tell the competing accounts of the algorithm apart, the project builds matched pairs of groups those accounts cannot distinguish, so any difference the models show counts against a theory.