More than 150 years ago, philosopher and author Samuel Butler wrote a pseudonymous letter, ‘Darwin among the Machines’, to the Christchurch Press. The 1863 letter described an evolutionary progress of machines towards, and beyond, human capabilities, and suggested that a potential war against such progress might already have been  lost.

Today, we are well beyond his ‘antediluvian prototypes’ of evolved machines, and we have a clear name for that evolution: Artificial Intelligence (AI). AI is the idea that, just as physical tasks have long been largely automated, mental tasks can be too. Seemingly every day, computers have been acquiring skills such as translating natural languages, drawing, playing video games, chess or poker, naming objects and captioning pictures, recognising and producing speech and answering questions.

Increasingly, they are doing so at superhuman levels. We should expect these new learning and perceptual capabilities to lead to rapid improvements in robotics, and thus also in the physical skills machines can acquire.

Groups around the world, including my new Broad AI Lab at the University of Auckland, are starting to use advanced machine learning techniques to give computers the ability to break down problems, solve their parts, and remember and apply what was learned for novel problems. In short, we are attempting to replicate the central basis of complex human thought.

One area where AI should be particularly useful is in dealing with complexity. Humans are limited in attention, in the number of concepts we can consider at once, and in how quickly we can communicate about them. This limited attention and communication is fine for everyday situations, but some of the most important systems that affect our lives, including biology, communications in social networks, complex software systems and economics, are overwhelmingly complex: we have no choice but to oversimplify them and in doing so we all but guarantee mistakes.

With AI, though, we have the potential to fully understand the enormously complex networks of causes and effects in biology and medicine. For example, medical teams will understand a disease fully in its context of an individual patient and design a treatment that will reverse that exact case, minimising harm or discomfort.

The way ahead is human-AI collaboration, with AI systems contributing their ability to access and consider the whole of the medical and biological literature, comparable patient records, and patient and pathogen genomes, and with teams of human experts simultaneously contributing local, specialist, empathy-based and care-based guidance.

Oddly, having only humans as examples, we do not know how difficult such broad, knowledgeable reasoning really is. It’s therefore very hard to predict when it might be performed well by machines, but we can be all but certain that broad, and then general, AI will be available before another 156 years pass, and probably much sooner than that.

It’s fitting that some of this progress be made in the country where Butler’s vision first flourished, and it’s vital that New Zealand’s economy fully benefits from the 21st century’s AI technologies. But these are not the only reasons we should help to develop, adopt and guide increasingly capable AI. As a small country, New Zealand has the complexity of other societies, but is limited in the skills it can apply to problems. Where large countries may have 100 experts, New Zealand may have one, or none. And, in areas from software engineering and building to seasonal fruit picking, we simply lack the flexible, expert labour pool to respond rapidly to changing needs. AI and robotic systems increasingly offer the hope of providing large-country levels of expertise and labour pool flexibility.

But this must be done without threatening the benefits – material, social, and psychological – that these roles give people now.

One reason I returned home after decades overseas is that I believe New Zealand can lead here. We have an almost unique combination of social inclusion, rapid technological adoption, and technical and financial capability that suits us for discovering how to use AI for societal wellbeing and how to mitigate its risks.

We can explore how to move from valuing people for utility, to valuing and rewarding our humanity – our interactions with other people; our contributions to social and cultural life; our ability to appreciate art, culture and the physical environment; and our stewardship of nature and of the earth.

New Zealand has often shaped human civilisation for the better, leading the world in fair wages, votes for women, and, very imperfectly, mitigating colonisation. Let’s also help lead humanity towards a better future with AI.  

Professor Michael Witbrock is from the University of Auckland's School of Computer Science and leads the new research group, Broad AI Lab.

Leave a comment