Profiles

Toby Walsh: wrestling with the good, the bad, and the ugly of AI

Pinterest LinkedIn Tumblr

“To see a world in a grain of sand, and a heaven in a wild flower,” begins William Blake’s Auguries of Innocence. At a young age, Professor Toby Walsh was Blake’s poem personified. “I remember looking at a patch of ground and thinking, ‘If I could understand what goes on in that square metre, I would know everything about the universe,’” he says.

His horizons have broadened beyond that patch of ground, but his curiosity hasn’t waned. Walsh, who is now Professor of Artificial Intelligence at the University of New South Wales in Sydney, Australia, is a world-leading researcher and communicator in his field.

Just a few years after his first childhood epiphany he experienced another. An avid reader, and with access to one of the first personal computers at his school, he realised that he could actually build the intelligent computers and robots he was reading about in sci-fi novels by his favourite authors, Arthur C Clarke and Isaac Asimov.

What I imagined when I was 11 is something that I single-mindedly have done since,” he says. “Maybe someone with greater imagination would have moved on to other things than their childhood fascination, but I never got away from it.”

“What’s fantastic now is that the rest of the world is waking up to those dreams, and realising that they are going to be an important part of our futures.”

Walsh was the first in his family to go to university, earning an undergraduate degree in theoretical physics and mathematics at Cambridge. He then embarked on a PhD in the then “fringe subject” of artificial intelligence (AI) at Edinburgh University – “the first and, at the time, the only department of artificial intelligence on the planet,” he recalls.

Who’s running the revolution?

Walsh notes that there may be as few as 10,000 PhDs in AI globally. “I think that’s a really important observation for the world to understand,” he says. “There are few revolutions that can change the planet that are driven by such a small number of people.

Unfortunately, one of the challenges is, that set of people is not representative of society as a whole. It’s heavily biased towards white men, and we’ve already seen some consequence of that bias. I would prefer the revolution to be driven by everyone, not by a small, select group of people.”

Gender balance in the sector has improved little in the last 20 years, and Walsh’s concerns about the need for a democratisation of AI is a driving force behind his considerable outreach activities. He has written two books about AI for general audiences, most recently 2062: The World That AI Made, about the possibilities inherent in a future in which the machines we have created are as intelligent as we are.

“There are many reasons why machines might have some technical advantages over human thought,” he says. “It’s already the case that in some fields of expertise, such as reading X-rays or translating Mandarin into English, machines can do tasks on a scale and a speed and a cost that humans can’t match.”

Given these advances, Walsh says, we ought to be preparing ourselves for a time when machines can match us in almost all our endeavours. “And we should think about how we use that as a force for good to tackle some of the wicked problems that plague the planet, like the climate emergency, pandemics, inequality and hunger, and make it a better place for all of us.”

For better or worse

A large part of Walsh’s work is concerned with the ethics of AI and he has campaigned hard for a ban on autonomous weapons, a cause which has been supported by technology leaders such as Elon Musk and DeepMind’s Mustafa Suleyman and which is now being discussed at the United Nations.

“We can’t prevent other people from taking our algorithms and misusing them, but we can at least demonstrate and help encourage the positive uses of the technology,” he says.

By way of example, Walsh describes how face recognition software was used by police in New Delhi to reunite over 3000 children in orphanages with their parents.

On the flipside, he adds: “Immense harm can be done with that same face recognition software. We’re already starting, sadly, to see that in China, with the persecution of the Uighurs using face recognition. But we’re seeing troubling signs of the misuse of face recognition even within our own democracy.

Walsh does, however, recognise the potential for positive government intervention. For decades, accepted wisdom has been that governments have no business to regulate technology, in case it would stifle innovation. But Walsh has no doubts that governments have a role to play, and laws such as the European GDPR have shown that good regulation is possible.

“All large industries end up being regulated,” he says. “No market works perfectly. And now big tech is the biggest market there is. Five of the largest companies by market capitalisation on the planet are tech companies, and they have immense power, immense reach.

“People don’t realise that the internet they see is unique to them. It’s the one that Google and Facebook think is in your interests, reflects your preferences, and which will try to get you to buy things and vote in particular ways. There’s plentiful evidence that social media has actually changed the outcome of elections.

“Some of it is bias on steroids, because it’s done at a scale and a speed and a personalisation that human systems could never have managed, and it’s less accountable because it is algorithms. There’s no human oversight and no human accountability to those decisions.”

Australia: world player in AI

Walsh moved to Australia in 2005. (“As a Brit, I was amazed to discover that bad weather was optional,” he says jokingly). “Australia has an excellent reputation in AI – as in other areas, we punch above our weight. I suspect few people realise we had the fifth stored program computer in the world in Australia, the first one outside the US or UK. And since that first computer, Australia has been very active in AI and robotics – we are five times world champion at robot soccer, for instance.”

Since then, his profile has risen steadily, in tandem with growing awareness of AI – in no small part because of Walsh’s work. In 2018, he led the team that prepared the first report on the implications of AI in Australia.

Walsh says he had never fielded an interview request until five years ago. But his phone has not stopped ringing since. “I spend maybe a third of my time talking and communicating, a third of my time doing research, and a third of my time talking to government about policy and doing all the other things you’re supposed to do,” he says.

It’s perhaps not so important for me to write yet another scientific paper. But it is important for me, I think, to use the platforms I have to try to inform this really important conversation. If, at the end of the day, you can’t communicate to others what you’ve done, there was no value to having done it. That’s a really important message for scientists.”

Follow Toby Walsh on Twitter | LinkedIn

Article by Iain Scott
Photo credit: Grant Turner/UNSW

Comments are closed.