Guyon Espiner: Why shouldn’t we trust AI on difficult public policy issues?

OPINION: When the National Party admitted it was using artificial intelligence to generate fake images of nurses, thieves and crime victims in election campaign materials, there was outrage.

A year later, when the Minister of Digital Affairs of the Government
Judith Collins has proposed using AI to teach children, analyze mammogram results and process Freedom of Information Act requests, but it has garnered only a passing interest.

A few years ago, the idea of ​​an AI writing early drafts of a minister’s speeches – which Collins told me in a recent RNZ interview would lack panache and humour – would have been a bit part of a science fiction movie script.

Collins, an optimist about AI and excited about the potential productivity gains, sees little point in regulating the technology.

“It’s like trying to stop the wind. It’s already here and it’s going to keep blowing,” she said. “I think we just have to stop being so afraid and realize that the benefits are huge.”

But she should be careful what she wishes for. In this industrial revolution, machines are coming for the elite.

You would think that after our experiment with social media – which Jonathan Haidt writes about in his new book Anxious generation is blamed for the epidemic of mental illness among teens – we are perhaps a little less optimistic.

This time we can’t say we weren’t warned.

AI godfather Geoffrey Hinton, who left Google to speak freely about AI risks, believes it poses an existential threat. “They could develop a goal to take control — and if they do, we’re in trouble.”

In his book Terribly smart, Mo Gawdat, former chief business officer of Google X, predicts that by 2049 AI will be a billion times smarter than the smartest human, a difference comparable to that between a fly and Albert Einstein. How, he asks, do you stop a superbeing from squashing a fly?

We’re at a strange moment in the public conversation about AI, where those who have contributed to its development – ​​including Hinton and Gawdat – are sounding the alarm, yet most of us think we know better.

Most people I talk to about AI don’t believe it can do anything more than predict text on an industrial scale.

The European Union is taking this seriously. In March, it passed a law regulating AI.

Interestingly, the use of AI in healthcare and education, as Collins proposes, would be considered high-risk in the EU and would require a great deal of transparency and human oversight.

Other uses, such as harvesting surveillance and internet data to create biometric databases or using AI to manipulate or exploit human weaknesses, are completely prohibited under EU law.

Personally, I am excited about the potential of AI and in awe of what it can already do.

I gasped when I saw OpenAI’s Sora create visual worlds from text prompts. What would that do to our film industry?

American actor and filmmaker Tyler Perry shelved plans to build an $800 million studio in Atlanta when he saw what Sora could do.

Check this out and consider whether you still expect the same number of filmmakers to make films in New Zealand, at such huge cost.

I recently spent a rainy afternoon talking to AI chatbot Claude about nonfiction journalism.

The quality of the writing and the sharpness of the observations, revealed in a matter of seconds, were astonishing, but not what many of us expected from our rulers.

“I want artificial intelligence to do laundry and dishes so that I can do art and write, not artificial intelligence to do art and writing so that I can do laundry and dishes,” wrote writer Joanna Maciejewska in X.

We thought robots would take over manual work, but AI is coming for the creatives. It’s coming for the elite. It’s also coming for the decision-makers: CEOs and perhaps politicians themselves.

A survey by online course provider edX found that 49% of CEOs believe most of their jobs could be replaced by AI.

Some have already done so. Chinese online gaming company NetDragon Websoft has named an AI (named Tang Yu) as its CEO.

The market seemed to accept the stock’s rise. The Polish beverage company also appointed an AI (named Mika) to the top job.

In Brazil, legislator Ramiro Rosário, fed up with corruption and government incompetence, passed a law written by ChatGPT. It was a mundane law (preventing water companies from charging residents for stealing meters), but a milestone worth noting.

In Denmark, the Synthetic Party is “the world’s first AI-powered political party.”

We already trust AI for autonomous vehicles, fraud detection, and medical diagnostics.

Why not use him to solve our insoluble public policy problems and write a press release (Collins could add humor and edge)?