Expect a multi-faceted impact on New Zealand's economy from AI, Treasury note suggests

Our government's financial and economical advisers, the Treasury, have weighed in on the potential effects of artificial intelligence (AI), with a cautious approach.

Which is not to say the "conversation starter" note from Te Tai Ōhanga isn't useful, as it is. Actually, the note isn't necessarily the official view of the Treasury, but more the thoughts of the staffers who wrote the paper. That said, it's hard to imagine an official Treasury view would diverge greatly from the conclusions drawn from the studies and research referenced in the paper.

Looking at AI from an economic perspective, the authors of the note, Harry Nicholls and Udayan Mukherjee, ponder if the technology will increase productivity, displace labour and if so, what kind of jobs will be replaced, and some of the regulatory approaches taken overseas.

In terms of regulation, it's a reasonably safe bet that New Zealand will wait and see what other nations do, Australia in particular, and the bigger economies of the European Union bloc and the United States. There's not much point in being first cab off the regulatory rank when so much of how the difficult to grasp technology is used is still up in the air.

This is particularly so for intellectual property rights, around which there are some serious court room brawls coming up. AI needs access to human created material to be able to imitate our intelligence better. AI companies thinking they can just freely snag fresh data and use it to train their models that are then rented out to users on a monthly subscription basis wasn't ever going to go down well with intellectual property rights holders. There are most likely trade agreement clauses to re-read here, and that's before we start thinking about the undesirable effects of a technology that's scarily good at imitating humans.

Unless we're all going to standardise on North American norms, the cultural aspect of AI will start to chafe and become increasingly relevant to users elsewhere in the world. Recently, Jensen Huang of graphics cards repurposed as AI accelerator hardware maker NVIDIA, currently trading the top spot with Microsoft and Apple as the world's most valuable company, pointed to the need for regional large language models (LLMs) to capture the culture, history and intelligence of the world outside the US.

Which is easy for Huang to say, but not at all clear how to achieve practically and in a respectful and ethical fashion.

AI in the local context will possibly be more interesting from an economic point of view for New Zealand, than using the technology via one of the big providers. For generative AI, that could mean training smaller models on local data at the edge, to create unique usage scenarios and applications. This is what local developers are excited about, while raving about how good AI is at generating code for the boring stuff their apps need. 

One thing seems for sure though: officialdom will have a hard time keeping up with the fast moving AI technology. The image generated for the Treasury note is from OpenAI's DALL-E 2 which is no longer available, having been replaced by the newer DALL-E 3.

Likewise, the authors asked Google Bard if their jobs as policy analysts were secure. Bard has been Gemini since February this year, and is a rather different kettle of generative AI fish these days with new under-the-hood technology powering it.

Also, asking an AI if your job will go is pointless. The publicly accessible AIs are heavily filtered by humans earning pittance for cleaning up the output. That's because having a chatbot without empathy and ability to reason take stock of a profession or trade's hope of survival and provide true answers is terrible PR.

On the other hand, ask managers wanting to widen margins while removing the pain of dealing with human staffers and you'll get a better answer. It's not going to be a nice one as it'll always be "yes, if the automation tech is good enough".

Unfortunately, "good enough" is a subjective term that doesn't necessarily translate into "highly valued output". 

That AI could sink the value of output, rendering it nearly worthless, is a real risk that tends to get drowned out in the current euphoria around the technology. One good example of that is this story from the BBC, about how AI ate copywriters' jobs. Adding insult to injury, the machine generated output was of such poor quality that the copywriters who'd lost their jobs were called back in to work in the "AI humanising mines" to improve on it, but at ridiculously low rates. 

The above needs to be offset against other scenarios like customer service representatives, a thankless gig if there ever was one, being supported by AI when taking voice calls. Speaking to complete strangers about often complex matters is never easy but with AI sitting there in the background, listening in with voice recognition, and anticipating what information needs to be found and the actions to be taken, a CSR's job can be so much nicer.

No more awkward pauses putting already irritated customers on hold while the CSR tries to find the right material, while scrolling frantically in multiple applications. In theory the result is happy customers and CSRs, with the latter being more productive as well. Being able to do your job well to help people is that magic job satisfaction sauce that reduces staffing levels and turnover, so happy employers as well.

There you have it then, that annoying "it depends" ambiguity that technology is steeped in. Maybe even "it depends on if people will start to use AI a lot" as uptake locally seems slow, ditto overseas.

As the note authors put it: "while the economic impacts of AI are still emerging as the technology continues to develop, it seems likely there will be a multifaceted impact on New Zealand’s economy."