The Albany government has decided against legislation to manage artificial intelligence, emphasizing Labor’s focus on the economic benefits of the technology with a new national roadmap and a plan to “unlock” vast datasets held by private companies and the public service to help train AI models.
Providing support and reskilling workers impacted by AI in their jobs, boosting investment in datacentres and sharing productivity gains across the economy are key components of the Labor government’s national AI plan launched on Tuesday.
The document also warns about datacentres sucking up vast water and electricity resources, AI-facilitated abuse targeting women and unanswered questions about copyright protections for artists, writers and journalists, putting their content at risk of being affected by big language models.
Sign up: AU Breaking News Email
But the government has rejected the EU example of setting up a stand-alone AI Act of Parliament – an approach pioneered by former minister Ed Husick – saying existing legislation would cover the fast-growing new technology.
“This plan is focused on capturing the economic opportunities of AI, sharing the benefits widely and keeping Australians safe as the technology evolves,” Industry and Science Minister Tim Ayres said.
Ayres’ AI plan states that the government wants to use AI technologies to “build a fairer, stronger Australia, where every person will benefit from this technological transformation”, adding that it could improve health, disability, aged care, education and employment and help create jobs.
The plan is supported by a $30 million commitment next year to set up an AI Safety Institute, which will advise the government on the technology, including whether new laws are needed in the future.
Ayres said the government would measure “success” by “how widely the benefits of AI are shared, how inequalities are reduced”, and how workers are supported.
As unions voice concerns over AI replacing human workers, the minister said technology should “enable the talents of workers, not replace them”, and he committed to consulting with unions and business on the way forward on the workplace.
The document promoted Labor’s desire to make Australia a top destination for investment, taking into account local excellence in datacentres, academic research and AI jobs.
Amid much debate about how the government should respond to the risks and benefits of AI, and calls for business to support productivity-boosting technology, Ayres indicated early in his tenure that a stand-alone function was not in his immediate sights.
“The government is monitoring the development and deployment of AI and will respond as challenges arise and our understanding of the strengths and limitations of AI evolves,” the plan says.
Labor had previously rejected calls from the business community and the Productivity Commission to support text and data mining exemptions for AI companies to help train their computer models on copyrighted material such as news reporting, books and films. The new AI plan says the government is still keen to help “unlock high value datasets to pilot AI use cases from both public and private sources” to help train those models.
After newsletter promotion
Government sources said those plans were still in early development, and did not rule out new licensing schemes or other compensation. The plan suggests that the government’s “non-sensitive” public datasets, such as those from the Bureau of Statistics, could be opened up to AI companies.
The plan states, “AI models are only as good as the data on which they are trained. Australia has high-quality and comprehensive data sets that can support AI innovations that can create value for the AI sector, provide public goods and better reflect the Australian context.”
“Both government and the private sector have high-value data sets that can support a globally competitive Australian AI sector.”
The plan also noted the need to expand new renewable energy and efficient cooling technologies to the system, noting that datacenters will consume about four terawatt-hours of electricity in 2024 – 2% of electricity supplied by the grid – with expectations that this could triple by 2030.
The plan states that datacentres, which can consume millions of liters of water, will need to develop new technology to cool their projects and reduce water use.
<a href