Interestingly, some of the timidity came from Gates himself. Asked what lay beyond the internet, he replied that computers might one day think, but that almost no progress had been made and that it might never happen.
Predictions about the internet once attracted skepticism—much as today’s forecasts about advanced AI do. Photo / Getty Images
The question Gates shrugged off became the life work of Shane Legg, once a Rotorua schoolboy.
Legg taught himself to program on a Dick Smith VZ200. He then went from Rotorua Lakes High School to the University of Waikato and the University of Auckland. His 2008 doctoral thesis, written in Switzerland, was titled Machine Super Intelligence. Two years later, he co-founded DeepMind, now Google’s artificial intelligence (AI) laboratory, and suggested the term artificial general intelligence (AGI) to describe machines that can replicate all the functions of human cognition.
Since 2009, Legg has put the odds of AGI arriving by 2028 at 50/50. He has never shifted the date. In a recent interview, he confirmed it and guessed two more years. Almost in passing, he mentioned that today’s AI knows things about the small New Zealand town where he grew up.
That is more than most New Zealanders know about Shane Legg.
Of course, you might say that anyone working for one of the big tech companies would say things like that. Besides, forecasts in this field have embarrassed their makers for 70 years, after all, though as the Letterman clip shows, the errors run in both directions.
Shane Legg, the New Zealand-born co-founder of Google DeepMind, has consistently predicted the arrival of AGI since 2009, long before such forecasts entered the mainstream.
Legg, though, made his prediction 17 years ago, when it sounded absurd. It sounds less absurd every year. I believe him, broadly, though I would not bet the house on the particular year. But even a serious chance that he is right is worth preparing for – and people cannot be trained overnight.
Officially, New Zealand is doing fine. Responding to a survey, more than 80% of businesses said that they use AI, up from about half in 2023. But counting as an adopter everyone who has asked a chatbot a question is like counting everyone with a library card as well-read.
In July 2025, New Zealand became the last country in the OECD to publish a national AI strategy. The tardiness shows a certain lack of urgency in Wellington’s view of the subject. Incidentally, whether a country actually benefits from a government strategy on AI is an altogether different matter.
Most firms and ministries have licensed a clever AI assistant, restricted it tightly, trained hardly anyone and declared themselves transformed. When an unskilled user makes a beginner’s mistake, the headlines go after the technology and nobody asks about the training.
There is a perverse logic to this. No manager loses a job for restricting a tool, but careers end over a subordinate’s blunder with one. What looks prudent to individual managers leaves whole organisations unprepared.
In Britain, Legg recently told the leading universities’ vice-chancellors that every faculty must work out what cheap machine intelligence means for its field. I wonder whether Legg’s views attract similar interest in our education institutions, where the main concern seems to be students cheating in exams.
He says the conversation must shift from asking AI to plan a holiday to asking what it does to the structure of the economy.
If a job can be done entirely through a screen, a keyboard, a camera and a microphone, Legg reckons, advanced AI will have a role. Lawyers, accountants, analysts, consultants and policy advisers all pass his test. So, awkwardly, do columnists.
That list undersells what is coming. Legg compares it to the Industrial Revolution, which replaced human muscle with machinery. This time it will be thinking. And thinking is what most of a modern economy runs on.
For 40 years, I have been the annoying early adopter, the boy with the computer magazines, attending computer fairs and watching computer shows on television. Nearly every technology I eagerly anticipated (voice recognition, video calls, virtual reality) arrived later than promised and sometimes did less than I had hoped. But artificial intelligence is running ahead of schedule.
Dr Müller got his future. People now work from Bali, Tuscany or Wānaka on connections faster than anything he imagined on that school trip. It no longer matters where you are when you work on your laptop.
Once AGI arrives, the question will be who still needs to sit in front of one at all, and what a majority of humans will do with their time.
Catch up on the debates that dominated the week by signing up to our Opinion newsletter – a weekly round-up of our best commentary.