Robots and AI have long been a staple of dystopian fiction, either as merciless overlords, or as downtrodden slaves. Sometimes, they just take all the jobs.
Professor Simon Lucey says that far from Artificial Intelligence (AI) being a futuristic nightmare, it has the potential to put Australia at the centre of a global revolution involving plenty of new jobs and benefits.
“AI is such an opportunity for us,” Lucey told The Brilliant.
Lucey isn’t simply speculating about a possible future – he’s creating it. In October 2020, he joined the Australian Institute of Machine Learning (AIML) at the University of Adelaide, where he is going to help drive Australia’s role as a world leader in machine learning and AI.
“Australia is a low-complexity economy,” says Lucey. “We only really create a couple of things, so it’s essential that we diversity and complexify our economy.”
Boundless possibilities
Previously, Lucey was Associate Research Professor at the Robotics Institute at Carnegie Mellon University in the USA, as well as Principal Scientist at Argo AI, a company developing self-driving cars. Two things drew him back to Australia. The first is that the Federal Government has just contributed $20 million to AIML’s new Centre for Augmented Reasoning, which sits next door to an innovation hub that includes Australia’s Space Agency.
The second is the University of Adelaide’s AI group. “We’re ranked number two in the world for machine perception,” says Lucey. “Not in Australia, the world.” That expertise should mean that Adelaide can attract companies and talent from all over the world.
The problem, however, is that few in the world know it.
“The message that a globally ranked place like AIML exists here in Australia must be sung from the roof tops,” says Lucey. “It is also imperative that this scientific excellence results in tangible outcomes for everyday Australians.”
It’s important work. Despite its history of extraordinary scientific discoveries, Australia isn’t particularly good at commercialisation; although it’s the 11th richest country in the world, Australia is only 23rd in innovation.
What Lucey brings with him is not just world-class excellence in AI research, but experience in creating university and private enterprise partnerships, so that research can be translated to practical applications for commercialisation.
While at Carnegie Mellon, for example, Lucey worked on technology to turn a smart phone into a 3D laser scanner that maps faces, which was sold to an American company that launched a product to “allow you to virtually try on glasses”. Another project was a system to improve training for basketball players. “You could get training feedback of how the balls go through the hoop,” says Lucey. “It turns out, you have the have the optimal angle to get through the hoop, and if you can train yourself to do that, you can dramatically improve your shooting percentages.” That technology is now used by the NBA. He’s also made major contributions to driverless car technology, and his various projects have attracted funding from companies like Disney, Apple, Samsung, Adobe and Facebook.
One of Lucey’s current projects is to develop technology for a smartphone to reconstruct objects in 3D using only its camera and AI. A user could take a picture of an arbitrary object, then send the 3D scan to a friend on the other side of the world, who could print the result in 3D.
“I really enjoy sitting between industry and academia,” says Lucey. “I really like seeing my stuff go out there and be used.”
A chance encounter
When Professor Tsuhan Chen, a renowned expert in machine learning, came to Australia, Lucey – a then-graduate student at the Queensland University of Technology – was charged with collecting him from the airport. “His expertise was getting machines to lip read,” recalls Lucey.
Rather than the venerable old man Lucey was expecting, Chen turned out to be young and charismatic. “I was amazed by the quality of the presentation,” he says. “It was edge of your seat.” Lucey flew to Sydney to hear his next talk, and Chen invited him to visit Carnegie Mellon University in the USA.
There, Lucey discovered a highly competitive environment where graduate students are immersed in coursework and exposed to the best thinkers in the world. The experience left Lucey with the belief that Australian STEM graduates are often unprepared compared to their peers elsewhere. He also thinks they don’t realise that with extra post-graduate work, they could “command a job anywhere in the US or Europe, working for top companies and actually doing the research, rather than just coding and software development,” he says. He wants to see a revolving door of the world’s best coming to Adelaide, while Australia’s best head overseas, in a never-ending exchange of expertise and ideas.
Another thing he’d like to change is the lack of diversity. “If you want powerful young women in AI, you need to have powerful female mentors,” he says. An idea he’d like to see adopted in Australia is a program like Girls of Steel, a program for high school girls to get together and build robots.
The USA is also where Lucey learned the benefits of collaboration between university researchers and industry. “They have hybrid appointments, where you are 50 per cent of the university, 50 per cent of the company.”
Above all, Lucey wants to see students grapple with the “why” of AI, not just the how. He says that robotics and AI today is in the same place that physics was a century ago – on the cusp of civilisation-altering discoveries.
Finding the how and why
“There’s a thing called ‘deep learning’, which is basically an architecture for setting up how a machine should learn,” explains Lucey. “It’s like ingredients of how to make a cake. I need eggs, I need milk and if I put them together the right way, I’ll get a cake. But I’ll have no idea about the chemistry – why does the cake rise?”
The great problem to be solved in robotics is the question of how to turn machine memorisation into true learning. At the moment, for example, driverless cars must be exposed to every scenario they might encounter – they can’t generalise beyond these experiences, and therefore can’t navigate environments they haven’t encountered before.
Once machines can work things out for themselves, they can take over mundane tasks and free up humans for more creative and complex work. Not only are they not job killers, but, Lucey says, each AI ‘knowledge job’ typically creates another four or five in its wake.
And their applications will be life-altering. “AI’s going to be a big help in climate mitigation,” says Lucey, ticking off a list of possibilities: monitor the environment, to spot small fires before they become big ones. Develop viable vaccines, faster. Do more manufacturing locally. Create autonomous vehicles. The list is endless.
“Australia is probably the country that’s going to benefit the most from automation, because we have a huge land mass with a low population.”
Far from bringing about a world of robot overlords, the work being done in Adelaide offers a better possibility – intelligent machines that can help solve intractable problems, and bring Australia closer to the world.
A future Lucey would like to make a reality.
Follow The Australian Institute for Machine Learning on Twitter | Website
Article by Felicity Carter
Photo Credit: Photo supplied