Like a baby taking its first tentative steps, New Zealand’s political firmament is finally talking about how to manage the arrival of artificial intelligence.

Analysis - Labour has published an AI policy and Opportunity has offered the outlines of one. There was an attempt at a discussion about it in the first leaders debate.

The most interesting contribution so far, however, has come from ACT leader David Seymour. In an interview with RNZ's Guyon Espiner on Friday, Seymour reminded us of his background as an engineering graduate.

"I was at engineering school 20 years ago and they told us that around 2035, if computer power kept growing at the exponential rate according to Moore's Law, which it largely has, then we would get to a stage where intelligence could self-reproduce and it might start to do things adverse to us that we can't control."

His credentials established, Seymour engaged on the issues AI presents for New Zealanders thoughtfully and more concretely than others have to date. He said:

  • There should be "somebody in government" in charge of AI policy for New Zealand - a near-enough endorsement of Labour's 'Office of AI' proposal.
  • His office has used AI in a "dummy run" to cut through a lengthy policy development process in "a very short space of time".
  • He has no issue with AI drafting Cabinet papers - the important thing is that they're read and understood by ministers.
  • The public service is wrongly resistant to adopting AI technologies because it fears being displaced.

These are all worthy of scrutiny as part of a broader debate about the role of government in AI implementation here.

But the biggest and most controversial position Seymour took was that it's unrealistic for this country to think it can influence the way the AI industry is regulated.

We "certainly could" take a national position on things and even regulate in our own context, he said.

"But I suspect that our best play is to be a fast follower as those debates play out overseas."

 

This is where the biggest faultline between the parties of the centre-right and those of the centre-left is likely to emerge on AI in the months and years ahead.

Labour's spokesman Reuben Davidson earlier told Espiner that AI likely represents the "most significant industrial revolution of our generation" and that future governments "have to help navigate what that looks like for New Zealanders".

Both Davidson and Opportunity's Qiulae Wong said several times and in different ways that the government needs to get active on AI, taking leadership on standards for its use.

The trouble is, their rhetoric is not quite matched by the ambition of their ideas.

In other countries, a much more active role for government in AI policy means things like:

  •  Counting AI as part of the national strategic infrastructure, and planning and investing in it as such.
  • Mandating certain sensitive data sets to be held entirely on-shore, or in a digital embassy.
  • The creation of state-funded national AI models for things like ACC claims or tax returns.
  • AI-specific alliances with friendly nations to pool resources.

If you squint, you could just about claim any one of these counts as the "fast following" Seymour advocates for New Zealand. But it seems unlikely he or any of the parties in the current coalition would be up for these ideas. Rather than the quest for full-scale AI sovereignty they imply, something more like AI resilience appears to be what he has in mind.

Here again though, the policy development by the parties seems a bit thin. All parties are keen to be seen as boosters for the benefits of AI. After nodding to that, most of the time is spent describing the risks and suggesting some regulatory responses to control for them.

But what about the effort that's required to achieve the benefits? In Thursday's debate, Green Party co-leader Chlöe Swarbrick was keen to point out her party's position of a one-year moratorium on data centres. It's a straightforward piece of populist policy and it may well be sound, but what if building more data centres is actually a plank in an AI resilience plan? Could preventing even local firms from building data centres just end up meaning even more of our data sits offshore?

The public appetite for understanding AI technology has been evident for many months. We are shifting to an expectation of leadership on the matter.

Our multi-party system should be well-suited to tackling the issue with a spectrum of ideas. Let's hear more of them.