The Productivity Committee is investigating whether technology companies should be exempt from copyright rules that will halt companies mining text and data to train artificial intelligence models.
In its preliminary report on “Utilizing the Data and the Digital Economy,” the committee used copyright as a case study on how Australia's existing regulatory framework can be adapted to manage the risks of artificial intelligence.
An important recommendation was that the federal government should do a drastic review of regulations to use AI to plug in potential gaps that could be exploited by “bad actors.”
Scott Farquhar, co-founder of software company Atlassian, last week called for a “urgent” overhaul of Australia's copyright regulations, claiming he is not a step away from other comparable countries.
Farquhar said that creating a text and data mining exemption to train large-scale language models “will unlock billions of dollars of foreign investment in Australia.”
The proposal was rejected by copyright institutions, a non-profit organization that collects and distributes royalties to thousands of copyright holders. Instead, the government advocates to create a new compensation scheme for creators of content used by technology companies to train AI models.
Sign up: AU Breaking NewsEmail
“Copyright is a great example of where Australia needs to sit down and ask, “Is our law suited to the purposes of AI?”, said Stephen King, one of the two commissioners leading a PC survey to exploit the opportunities of PCs' digital economy.
“The obvious harm is that AI companies may use copyrighted materials without providing proper compensation. On the other side, it requires the development of AI-specific tools that use that copyrighted material.
“Through the copyright collection, we might be able to say that we can approach AI in the same way that we approached copyright in other ways. Music plays everywhere. So we collect a society that is approved under competition law and act on behalf of singers, songwriters and creators.”
King said the PC is seeking feedback on other options before the end of the year before final recommendations.
“As long as AI companies have got legal copies and they pay for it, there is a fair trading exemption, including text and data mining.
In the third of five theme reports, the committee said that artificial intelligence could revive Australia's MO death productivity.
PC modeling has shown that even the most conservative estimates AI is to raise the economy by increasing an increase of $116 billion over the next decade.
After the newsletter promotion
King said this could translate into a $4,300 kicker on the average Australian real wage ten years later, and the actual profit could be much greater.
The committee advised from creating comprehensive AI-specific laws, warning that clumsy or excessive regulation risks curbing the potentially transformative benefits of the technology.
That message is likely to have been well received by the government, with productivity aide Andrew Lee supporting this approach.
However, King said the committee did not advocate for a minimalist approach to AI regulation.
“It's not a light touch at all,” he said. “What we're saying is that it makes AI easier, cheaper and faster to engage in harmful behavior, but most of that harmful behavior is already illegal.
“Think where the harm lies and see if it's covered by existing laws, and if so, make sure regulators have the resources and power to stop bad actors.”
PC's interim report supported changes to privacy rules to allow companies to incorporate an outcome-based approach rather than a “box-tic” exercise “compliant with the letter of law but not its spirit.”
The committee said the government should “support new channels to enable individuals and businesses to access and share data related to them.”
