Scott Farquhar thinks Australia should let AI train for free on creative content. He overlooks one key point | Artificial intelligence (AI)

Australia should take over the U.S. style copyright to allow artificial intelligence Sow up all creative content Or risk scattering in the industry in Australia, according to the Atlassian founder Scott Farquhar.

Farquhar, Managing Director of the Tech Council of Australia, CEO, ABC's 7.30 program from ABC On Tuesday: “The entire AI use of mining or searching or data is probably illegal for Australian law, and I think that harms a lot of investments in Australia.”

This was due to the fact that Australia have no fair usage exceptions that are encoded into copyright law, as the United States do.

Farquhar's claim is overlooking that this is not a defined topic in the United States and could have devastating effects on the creative industry.

Companies that develop AI, including Atlassian, Google and Meta, want an exception to text and data mining insert into copyright To make AI able to train all human work in the long run without paying for it.

_{Register: Au Breaking News -E email}

Farquhar's argument is that it is not theft of people's work, unless the AI is used to copy an artist directly, e.g. B. creating a song in your style.

“I think people would say that hey, when people sit down with a digital companion, an AI song creator and work with a AI to create something new for the world, that's probably a fair use.”

According to Farquhar, the advantages of large voice models prevailed the problems that were raised by AI, their data for the free work of other people.

His argument depends on whether this AI then creates something “new and novel” that is described as transformative in copyright – something new.

He said he would not have a problem with anyone taking what they had created and used as long as it was “transformative”.

“If someone had used my intellectual property to compete with me, then I think that this is a problem with me. If they have used all the intellectual property of the entire software in the world to help people write better software in the future, I think that this is fair use.”

The US law is not set for AI training that is fair use. The US copyright office found In the May report in May In the case of generative AI training that there are dozens of complaints that challenge AI companies, use fair use as an excuse for training models with great language in copyright-protected work without payment.

In the United States, there are also factors to check whether something is used fairly:

• Whether the use is commercial or not.

• The type of work protected under copyright.

• The amount of work protected under copyright.

• The effect of use on the market for or value of the copyrighted work.

In the US case law, the transformative nature of what has been made of it is listed in a May publication However, it is not the only factor that has to be taken into account, since the effects on the market of copyrighted work are the key. According to the US Copyright Office Report, the Supreme Court of the United States described this twice as “undoubtedly the most important element of fair use”.

“Copying, which is involved in AI training, endangers considerable potential damage to the market for or the value of copyrighted work,” the report said.

“If a model can create much similar outputs that replace work in the training data directly, this can lead to lost sales.

“Even if the results of a model are not significantly similar to specific copyrighted work, you can water down the market for work that contains in its training data, including the generation of materials of similar material more similar to this.”

The authority's office no longer heard to recommend the legislative intervention in the United States, and found that voluntary licensing was already underway with some AI companies and would allow the licensing that AI innovations continue to “progress without undermining the rights to intellectual property”.

Farquhar's argument that the manner that generative AI uses copyrighted works would stack if it were properly guaranteed that the entire use would be transformative and that the markets from which they come would not be influenced.

In many industries, AI could have devastating effects. In the messages, for example, AI summary in Google search already means that people are less often clicking on stories for information, and recommendations of AI chatbots compared to the crawl in the AI, one side crawl, even worse.

To argue that fair use for AI is something due to the US law that is aimed at Australia, overlooked that it is hardly rejected by law and is fiercely fought hard in the dishes. If you give the Tech company to the Tech company on behalf of the innovation for a new industry, it could run at the expense of many other industries.