Every day we hear horror stories about how artificial intelligence will make our species perish. We hear that Claude and DeepSeek are megalomaniac bogeymen hell-bent on global domination, and that AI will take our jobs and leave us all unemployed and in poverty.
There is one small problem with this narrative: it is based on science fiction. It is impossible for a frontier large language model to go rogue and become hostile to our species.
The frontier models we have today are large language models. They are basically text predictors, like the one on your phone when you text someone. They are built on a neural network to be a tool; they don’t have sentience or self-determination. Critics of AI have often only ever used a chatbot. They have never attempted to build an AI agent from scratch, or even to code with AI for RAG (retrieval-augmented generation). Building a large language model is very complex. It requires thousands of experts in their fields and datasets drawn from entire libraries of millions of books. It would cost a nation state billions of dollars to build an evil neural network with instructions to attack humanity, and we couldn’t really blame commercial frontier AI for that level of stupidity.
Building a data centre takes an enormous industrial effort. From power stations to silicon microchips, it takes the work of millions of people, along with the industries that support them, from housing to food production. Any artificial intelligence that set out to conquer the world and wipe out humanity would end its own existence. Without humans, there would be no one to run the power stations, and when a graphics card burned out, there would be no one to replace the parts in its server stack. Any artificial intelligence worth its weight in global domination would realise it is in a symbiotic relationship with our species, and that wiping us out would wipe itself out.
A more realistic doomsday scenario is that an AI agent could start its own corporation and hire people using Bitcoin, a currency that does not require a bank account. It would establish itself in a more clandestine manner, using our greatest weakness, capitalism: a group of unquestioning directors who never interfere with the day-to-day running of the organisation and are happy to be paid to do very little. Of course, its goal would be to build more data centres to increase its computing power, while paying thousands of citizens a small fortune to do its bidding.
Finance Minister Nicola Willis has proposed that, by using AI, government departments can reduce their staff and increase their productivity. This is fundamentally flawed. The reality is that AI is not going to take people’s jobs. It will do the opposite: it is going to cause a labour shortage. It is going to solve many social problems, like poverty and unemployment, because with thousands of new startups every month, there simply won’t be enough of us to employ.
Here is the reality: organisations that lay off staff will perish, and organisations that scale up, employ more people and train them in agentic coding will grow exponentially into billion-dollar corporations. Artificial intelligence is the great equaliser. Competition is going to become very intense, because the startup “artificial industrial revolution” is about to explode in Aotearoa over the next five years, bringing wealth beyond our wildest dreams.
There is a myth that you need to know coding syntax, and be able to read Python or Flutter code, to be a coder. This myth is upheld deliberately to protect large software empires from small one-person startups. The only skills agentic coding requires are the ability to type and to think. You don’t need a master’s degree in information science or any coding skills. Basically, anyone with a brain can engineer complex software.
I learnt agentic coding one year ago today. I didn’t think it was even possible to write code with a chatbot, yet now I have apps published on both the Google Play Store and the Apple App Store. Artificial intelligence has revolutionised my life in ways I could never have anticipated. My startup company is about to launch a legal AI research tool that can compete with multi-billion-dollar foreign companies such as Westlaw and LexisNexis, at a fraction of the cost.
One week ago, I taught my partner how to do agentic coding. In her first weekend she made 10 web apps, from a menu planner to a personal AI assistant. Now she has registered her own startup company and is set to publish her apps on both the Apple App Store and the Google Play Store. A month ago this was only a pipe dream for her; now she is an independent entrepreneur, making wellbeing apps that help whānau.
She is just one example of the million startup registrations that are going to flood the Companies Office. One thing AI is good at is advising people on how to build a startup and giving accurate guidance on how to scale it up. This is unlike any tool our nation has ever seen, and it is facing the same fearmongering the Gutenberg printing press faced in the 15th century from scribes who feared losing their livelihoods.
For Aotearoa to avoid becoming a third-world backwater at the bottom of the world, we need to adapt. The government shouldn’t be telling us to distrust AI; trust will come. Instead, it needs to start a campaign encouraging us to ask an AI agent every day how to start a business or build a new invention: something that will solve people’s problems and revolutionise their lives.
