'Forced down our throats': Why Australians are the world's biggest AI sceptics

Australians are the world's biggest sceptics of artificial intelligence, even as it becomes increasingly part of everyday life.

But across the Asia-Pacific, many people keep using AI despite their doubts, while nations go their own ways to protect against its risks.

Isabel Oar finds herself relieved these days to spot grammatical errors in an email.

It is a sign that she is reading something from a human, not artificial intelligence (AI).

"I get quite irritated if I feel like I'm responding to an email that's been written by AI," the Queensland-based teacher says.

"I feel that communication should be human. So I refuse to use it for writing."

After starting as an enthusiastic user of OpenAI's ChatGPT for coding and other tasks, Isabel has pulled back more recently, like many other early adopters she knows.

For a start, she has come to trust AI less after seeing it make mistakes in its responses and in coding. 

"I don't really ask AI questions anymore. I still Google things," she says.

However, even the search engine's use of AI summaries makes the technology seem unavoidable.

"We're having this forced down our throats. It's in everything and we're not consenting to it," she says.

The ABC has been talking to Australians over the past year about their hopes, doubts and experiences as AI rapidly enters workplaces, homes and education. 

Over that time, views have quickly changed from caution, excitement and concern to something different.

Isabel's experience is echoed in many of the hundreds of responses to a recent ABC call-out asking people how AI has changed their lives.

Here is how other respondents put it, which you can see by scrolling through their sentiments below.

"I don't trust AI; I use it daily as it gives me confidence with writing."

— Craig, Victoria

"AI has a way of being inherently addictive to use. The models are developed by companies who want you to use it more, in a similar way to social media."

— Harriet, Queensland

"I am not quite so trusting of the information served up to me and now tend to check with more than one product … I still find it extremely useful and use it daily."

— Laura, New South Wales

"I trust AI more than I did a year ago, but in a different way. I trust it as an exceptionally capable assistant, not as a substitute for experience, expertise or professional judgement."

— Mike, New South Wales

"Being told constantly at work 'just run it through [Microsoft] Copilot' is frustrating. I'm not an expert in these fields … so I'm relying on AI to produce accurate results, which often it does not."

— Ellie, Victoria

"I trust AI less than I did a year ago … not so much because of the technology but because of the lack of community care from the owners of AI companies, and the lack of lawmaking by governments to protect society."

— Darren, Victoria

"We need to tread more carefully ... The march of AI is not inevitable. It's a choice."

— Patricia, New South Wales

Australians are among the least trusting and accepting of AI in the world, in stark contrast to people in many neighbouring countries, according to a study by the University of Melbourne and consultancy firm KPMG.

Trust and acceptance were much higher in emerging economies like India and China, where AI was more likely to be seen as a pathway to prosperity, says Nicole Gillespie, a scholar on trust in organisations and professor of management at the University of Melbourne.

"In emerging economies, AI plays a stronger role in filling real resource gaps. It can help overcome shortages in skills, information and networks, and bridge barriers of language to open access to global markets," she says.

People in emerging economies were still aware of the risks of AI, and the survey found users in India and China reported slightly more experience of its harms, a result reflecting their higher use of the technology.

Rather than being more worried about AI's risks than others, Australians were simply much less likely to see the upside of AI, Professor Gillespie says.

Another survey suggests that people around the region are using AI even if they do not always trust it.

People in China, Indonesia and Malaysia had the highest rates on this measure, according to an online survey by market research firm Ipsos.

In Australia, New Zealand and Japan, more than half of people say the same. 

Data from the same report also showed that China, South Korea and Indonesia were more excited about AI products and services, compared to Australia and New Zealand, who were more nervous.

"What this tells us is that trust is not necessarily a prerequisite for usage," says Hamish Munro, Ipsos chief executive in the Asia-Pacific.

"Adoption can move ahead of trust, with people willing to engage with the technology while still questioning how much they should rely on it."

'It's not turn-off-able'

Some respondents to the ABC's call-out say they wanted to opt out of the AI increasingly present in software or were abandoning programs that did not give them this option.

Emma Gerts, a writer and librarian from New South Wales, says she quit using Microsoft Word and Google Documents as she did not trust them not to scrape her work and use it for training AI models.

She sees AI as inherently untrustworthy because of its mistakes, the difficulty of verifying its responses, and the profit motive of AI companies.

While she believed it could be useful for tasks like coding, it had had no benefits for her work, including writing emails, she says.

"By the time I got through prompting it and re-prompting it and figuring out exactly the right words to say so that it would understand what I meant, I was just like, I could have written this email already," Emma says.

"I turn AI off anywhere I can turn AI off.

"But unfortunately, it's not turn-off-able in a lot of things."

In Western Australia, digital literacy educator Gabrielle Jolliffe says she had noticed her students become less open to the technology this year, a change she attributed to their experiences receiving some of its erroneous answers.

"This is the first time I have struck amongst my students a reluctance and suspicion to use it," she says.

Gabrielle would like to see providers of free generative AI programs stating clearly up-front how they would use information they gathered.

"Just like we have labelling requirements on alcohol, on what's inside food … they need to be very, very open and transparent about how the data is being used."

Meanwhile, Emma would like to see regulations requiring that software let people switch off AI. 

"I don't think they should be allowed to force AI into a program," she says.

Such support for stricter rules is echoed in the University of Melbourne/KPMG survey, which showed most Australians backed regulating AI.

They were also less likely than people in other Asia-Pacific countries to say that current safeguards were adequate.

Professor Gillespie, from the University of Melbourne, says "thin institutional reassurance" was a factor in Australians' low trust in AI.

"Only around half express confidence in government, big tech firms or commercial organisations to develop and use AI in the public interest," she says.

From Go to 'whoa'

In Australia and other Western nations, the release of ChatGPT in 2022 is widely regarded as the event that began broad public discussions about the potential harms and benefits of AI.

For much of Asia, that moment arrived when Google DeepMind's AlphaGo program defeated the South Korean grandmaster Lee Se-Dol at the strategy board game Go in 2016, says Jian Xu, an associate professor and expert on China's AI governance at Deakin University.

"It marked an important stage in the longer trajectory of AI awareness and governance in Asia, well before the emergence of ChatGPT," Dr Xu says.

Since then, Asian nations have diverged in how they regulate AI and manage its risks.

At one end of the spectrum, China has developed one of the world's most comprehensive regulatory frameworks for AI governance, Dr Xu says.

Rather than relying on a single, overarching AI law, China has built a "layered regulatory framework" combining general laws for cybersecurity, data security and personal information with AI-specific regulations, he says.

Among them are the recent "Interim Measures for the Management of AI Anthropomorphic Interactive Services", taking effect in July and addressing risks associated with AI agents simulating human interaction and emotional relationships.

Another measure requires that AI upholds "socialist values" and refrains from generating content that incites the subversion of state power.

"The government seeks to promote AI development while ensuring that emerging technologies remain safe, controllable, politically compatible and aligned with broader national development objectives," Dr Xu says.

"This can be seen as 'AI governance with Chinese characteristics'."

At the other end of the spectrum, Singapore has pursued a "soft-law" approach balancing safety with realising AI's economic potential, according to Terence Lee, a professor and expert on Singapore's AI governance at the University of Nottingham Ningbo, China.

The city-state's government has prepared to embrace AI for more than a decade and wants the external business and investment community to see it as a place that lets businesses road-test new ideas with minimum restrictions, he says.

"Rather than passing restrictive, overarching legislation … Singapore relies on voluntary or self-regulatory frameworks, and it targets updates to existing laws (rather than rolling out new legislation), and offers technical testing tools and guidelines," Professor Lee says.

Australia's unanswered AI question

One Australian government official has called trust the single factor that will determine whether the country's AI industry thrives or stalls — more influential than chips, capital or energy.

"Why are Australians the most sceptical people on earth about this technology?" Assistant Minister for Technology Andrew Charlton said at an AI summit in Canberra earlier this year.

"Australians took up smartphones, contactless payments, and online banking faster than almost anyone. Adopting is in our DNA.

"Australians are not refusing the future. But they are asking a fair question: who is this actually for?

"And until we answer that question convincingly, scepticism is not irrational. It is the reasonable response of people who suspect they are being asked to carry the risks while someone else hoovers up the rewards."

The Australian government is hoping to build public trust in AI through reforms addressing its risks.

Among its priorities is legislating a "digital duty of care" requiring AI companies to make their products safe and proactively address potential harm, reforms to personal data protection laws, and safety testing of frontier AI systems through its newly-established AI Safety Institute.

Prime Minister Anthony Albanese in July also indicated Australia would change its fragmented approach to regulating AI.

He announced the Australian government would coordinate the design of "Australian standards" for the technology, and that legislation would be brought to parliament early next year.

Among the standards would be rules requiring data centres to minimise water use and underwrite new power supply to cover their consumption.

It was a response to growing backlash against data centres among local communities, a sentiment also reflected in audience responses to the ABC's recent call-out.

Gerard Goggin, professor and researcher on the cultural and social dynamics of digital technology at Western Sydney University, says awareness of data centres has helped bring about Australia's shift in regulating AI.

"That's something that's emerged in a way more palpable sense … in terms of environmental aspects, planning aspects, but also perhaps the strategic opportunity for Australia to harbour data centres," he says.

Professor Goggin says consumer protection, disability accessibility, and copyright would be among the most important pieces of legislation responding to the rise of AI in Australia.

There were also harmful uses of AI, such as "nudify" apps, that could and should be banned, he says.

"We shouldn't shy away from [doing] that when that comes," Professor Goggin says.

Meanwhile, in Queensland, Isabel still sees AI's benefits and believes it has a useful role for teachers, especially in speeding up tedious, time-consuming administrative tasks.

Still, she says there were benefits to adopting technology more slowly.

"You get to see the mistakes that everybody else makes and then adjust," she says.

"And I think maybe Australia needs to consider slowing things down." 

Credits

• Reporting & Writing: Doug Dingwall
• Graphics & Design: Jarrod Fankhauser
• Additional Reporting & Producing: Steven Viney