The AI outlook

AI is said – by AI itself in response to a web query – to have hit the mainstream in 2022 with the launch of ChatGPT by US company OpenAI. Since then, AI chatter has gone from a background hum to full noise. At the time of writing, the news of the day was how GPs are saving up to ten minutes per patient consultation by using AI-powered ‘scribes’ to generate clinical notes.

Also creating buzz was Environment Southland’s approval of plans for an ‘AI factory’ at Makarewa, near Invercargill, by Singaporean-owned company Datagrid. As news broke that the US$3 billion data centre had been green-lit, University of Auckland computer science senior lecturer Dr Ulrich Speidel told the Science Media Centre the thousands of graphics processing units (GPUs) housed in such facilities have two types of output.

“[There’s] data – which is what we want – and heat, which is generated when the billions of logic gates in the GPUs do their switching.” Makarewa has a cooling advantage thanks to its average air temperature of nine to ten degrees. Yet Datagrid’s facility, intended to come online by 2028, is still expected to be the country’s second-biggest electricity user.

That same year, worldwide investment in data-centre construction will reach US$3 trillion according to the Economist. In the magazine’s rosiest view, we could be entering “a new world of economic growth” of 20 percent a year in which artificial general intelligence (AGI) models are better than the average person at most cognitive tasks.

Then there’s the less-rosy view. Free software activist Richard Stallman, a veteran of Massachusetts Institute of Technology’s AI Lab, recently said “pretend intelligence” is a more apt name for AI and that ChatGPT is a “bullshit generator”.

Then there was that open letter in 2023 signed by tech titans including Elon Musk calling for a pause in training AI systems after OpenAI’s launch of GPT-4, which cited the risks such powerful systems pose to humanity. (One of OpenAI’s original backers, Elon Musk has since mooted creating data centres in space to provide AI processing power. To help get there, he has merged his rocket company SpaceX with his AI company, which espouses ‘AI’s knowledge should be all encompassing and as far reaching as possible’, suggesting his GPT-4 qualms have passed.)

Closer to home, the AI Forum first brought together New Zealand’s AI developers and users with researchers nearly a decade ago. And last July a national AI strategy was launched with then-minister of science, innovation and technology Shane Reti saying the country needed to develop stronger capabilities to “drive economic growth” but the government would be “taking a light-touch approach”.

Yet surveys of public attitudes to AI show widespread anxiety.

A 2025 study by Melbourne Business School and KPMG spanning 47 countries and more than 48,000 respondents found while two-thirds of people regularly use AI for personal, work or study purposes, less than half trust AI systems. In New Zealand, 76 percent of those surveyed were concerned about negative outcomes from AI and less than a quarter thought existing safeguards were sufficient. But two-thirds expect AI to deliver a range of benefits. 

Superpowered sceptics

University of Auckland alumni in the thick of AI development can see the technology’s upsides yet recognise the public’s nervousness.

Hamish Macalister, for one, acknowledges concerns over job security as AI eliminates many roles.

“The pace of change is way beyond any form of technological development we have seen before. That means it is too fast for typical human adaption, such as retraining and reskilling,” says Hamish, who is a co-founder of Singapore-based Transparently.AI. And with that, he notes, comes the accompanying risk of more extreme disparities in wealth distribution.

However, in the financial risk assessment field, where Transparently.AI operates, the benefits of the technology are clear, says Hamish, who completed undergraduate and masters degrees in accounting and economics at the University of Auckland in the 1990s followed by a doctorate in finance in the late 2000s. After three decades in the quantitative finance world, he started Transparently.AI with cloud computing specialist Mauro Sauco, formerly of Google, IBM and Amazon Web Services.

Transparently.AI has built a ‘digital forensic detective’, which looks for clues that companies are cooking the books.

“We use AI to solve a trillion-dollar problem: the fact that traditional auditing only catches about three percent of accounting fraud,” says Hamish, who is the company’s chief scientist. “We have a success rate of over 90 percent for prediction of corporate collapses up to three years before they happen.”

Coming down the AI pipeline he sees a new era of agentic AI, defined as autonomous agents capable of multi-step reasoning and execution.

“In finance this means an agent won’t just flag a suspicious transaction; it will initiate the investigation, cross-reference it with historical data across different platforms and draft the regulatory filing – all while you sleep. We are moving from ‘AI as a tool’ to ‘AI as a digital co-worker’,” says Hamish.

If things play out well, he sees the technology helping solve humanity’s myriad challenges, but that doesn’t mean he’s immune to the nervousness being expressed more generally about such trends.

“My co-founder and I talk about this all the time. We both have kids. We are deeply concerned regarding how this plays out and what it means for them.”
One response in the face of AI’s erosion of trust: promoting the notion of scepticism as a superpower. “People everywhere will have to develop a forensic mindset similar to what we use at Transparently.AI,” he says. 

Putting up the guardrails

Natasha Crampton, a University of Auckland law honours and information systems graduate, has arguably one of the most influential oversight roles within the AI development world. The former Microsoft Australia and New Zealand in-house counsel is chief responsible AI officer at the software giant, a key partner of OpenAI.

Based in Redmond, Washington, she leads Microsoft’s “effort to translate the company’s AI principles into practice”. Her team works with engineers, researchers and the company’s commercial divisions on guardrails that govern how AI systems are designed and reviewed.

“It’s work that sits at the intersection of technology, law and society – an interdisciplinary space I was drawn to long before AI dominated the conversation,” says Natasha, who understands people questioning AI’s costs and benefits.

“It’s also true that many earlier technological innovations – like electricity or the internet – have led to societal shifts and upheaval. It’s important to be clear-eyed about both the risks and the potential of a transformative technology like AI and to make choices that guide the technology towards outcomes that serve people and society.” 

Natasha, who was a member of the UN’s High-level Advisory Body on Artificial Intelligence, highlights international efforts being made to mitigate the risks of advanced AI systems. A UK government-commissioned International AI Safety Report released this year – its reviewers included AI Forum New Zealand – sets out to help policymakers keep up with the technology’s developments. Unsurprisingly it notes “the rate of change, the breadth of applications and the emergence of new risks pose important questions”.

Microsoft attempts to answer some of those through the AI Economy Institute, which it set up last year to sponsor interdisciplinary study of the technology’s economic and social impacts.

Natasha says the institute’s establishment is in line with the report’s recommendation that a network of such bodies be created.

“The international network of these institutes was featured at the India AI Impact Summit in February and is already creating the foundations of an international science of AI evaluation,” says Natasha.

A 24/7 junior colleague

While AI poses global challenges and opportunities, it’s already helping most of us with the day-to-day.

For Amar Virk, San Francisco-based product lead at Irish-American payments company Stripe and a University of Auckland electrical engineering graduate, his personal AI light-bulb moment occurred during a 2024 home renovation project.

“I’m not an expert in home construction and didn’t always know whether my contractors were doing things right,” says Amar. “So I’d take photos of the work and upload them to ChatGPT to double-check things or ask questions to verify what my guys were telling me.” 

As the project progressed, he increasingly relied on ChatGPT, becoming convinced the technology was a step change, not just hype.

Adoption at work was slower. “These things tend to move more cautiously in a corporate setting. But I started bringing AI into my own workflows at Stripe pretty quickly.”

Initially, he used it for drafting emails and producing documents, expanding it to help with such tasks as keeping meeting notes, tracking action items and even creating custom dashboards and doing analysis.

He believes the “computer use capabilities” of Claude, the AI from Anthropic, represent a breakthrough, where AI doesn’t just answer your questions but does things on your behalf .

“Within the next 12 months it’s clear we’re going to have what essentially amounts to a capable junior colleague that works 24/7.”

Another leap he sees is multimodality: instead of responding just to text inputs, AI will work seamlessly with text, images, video and audio. If both trends advance at the rate he expects, he is forecasting the emergence of extraordinary downstream applications in the coming year.

“In science, for example, you can imagine an AI agent that doesn’t just answer a question about a protein, it autonomously designs an experiment, runs simulations, analyses the results, adjusts its hypothesis and starts again. Around the clock, no breaks.”
He urges the AI-hesitant to try it out on planning a trip or deciphering a complex document.
“The fear tends to dissipate once you see that these are powerful but imperfect tools – they’re impressive and useful, but not magic or infallible.

“The people building these systems are, by and large, thinking seriously about safety and alignment. That doesn’t mean mistakes won’t happen – they will – but the narrative that AI development is reckless and uncontrolled doesn’t match what I see from the inside.” 

Advances in film and music

Education is the context in which University of Auckland conjoint business and arts graduate Bethanie Drake-Maples is applying artificial intelligence at Atypical AI, which she founded in 2023 in San Francisco.

She has more than a passing acquaintance with the field. Bethanie spent five years at Google AI, has a masters degree in learning science from Stanford University and is completing a PhD there.

At a fraction of the cost of human tutoring, Atypical’s AI-based tutors provide help for exam preparation that used to be too expensive for all but the richest test-takers.

“We take a unique approach, combining socio-emotional learning, psychology, human-computer interaction and cognitive extension theories to craft an experience that supports neurodivergent learners at any stage of test preparation,” says Bethanie.

She talks to people daily who are concerned about AI’s potential impacts on students, teachers and people who are neurodivergent. “It certainly is a time of unprecedented change and rapid [AI capability] releases without a huge amount of user testing.”

However, she’s encouraged that technically sophisticated research centres such as Stanford’s Graduate School of Education and MIT are on the case.

Acknowledging that international competition in AI at the economic and military level is “a cauldron” in which some sets of users will be disadvantaged, she says “many millions if not billions” of people are benefiting from the technology daily.

While AI will result in job losses in some sectors, such as computer coding, for example, it will also democratise some areas, distributing creative potential to a wider group of people. If some creatives see that as a threat, Bethanie is excited by developments in AI-generated music and film.

“We’re seeing great advances in AI film editing that I think will unlock wonderful creativity. And, of course, advances in music just seem to be an incredibly joyful application for artificial intelligence globally.”

She dreams of bringing a learning science venture home to New Zealand that meshes with this country’s cultural values, seeing “an opportunity to make education and intelligence a national strategic asset”.

“But a learning science venture in New Zealand can’t just be a ‘copy-paste’ of Silicon Valley tech; it has to be built for our unique context,” says Bethanie.