To win the AI rush, Australia needs expert humans

The piece, titled 'To win the AI rush, Australia needs expert humans,' contends that advances in fine-tuning, post-training and distillation have reduced the cost of building and adapting smaller models. This, it argues, strengthens the case for Australia to enter AI model-building and develop the associated skills. The article argues that these technical advances make the case for domestic model-building stronger than it was previously.

The article identifies three techniques — fine-tuning, post-training and distillation — as drivers of lower costs. According to the argument, these advances make smaller models cheaper to build and adapt, opening up model development to countries without large frontier laboratories. The piece frames this as an opportunity for Australia to build domestic capability rather than rely solely on imported models or infrastructure. It argues that the falling cost of these techniques changes the economic case for model-building.

The analysis follows a recent account in which Australia said it did not want to be AI's landlord link. The new piece shifts the focus from infrastructure to human capital and model development. It argues that winning the AI rush requires developing expert humans alongside any technical investment. The piece suggests that expertise, not hardware alone, will determine whether Australia succeeds.