Machine learning is driving the computing revolution now sweeping through the world’s economies, and an Adelaide-based institute is ranking high among leading research centres dedicated to this sophisticated form of data analysis. A type of artificial intelligence that enables a computer to improve its operations with experience, machine learning has already up-ended global transport, communications, and retail industries.
It has spawned global behemoths such as Uber, Snapchat and TikTok, and brought vast computing strengths to ordinary pursuits like supermarket shopping, says Professor Anton van den Hengel, a senior academic and centre director at the University of Adelaide’s Australian Institute for Machine Learning, based at Lot Fourteen.
Consistently ranking among the top five in the world for computer vision research, a type of artificial intelligence that enables computers to understand meaningful information from images and video, the Institute is attracting significant talent from Australia and abroad.
“We are competing not only against the best universities in the world, we’re competing against the biggest companies in the world, we’re competing with Google, and Apple and so on,” van den Hengel says.
“It’s incredible, really, that we’ve got this great research strength, in such an economically valuable area, here in Adelaide.
“The reason that there’s all this enthusiasm for machine learning at the moment is it’s transforming every industry. We’re living through it.”
Machine learning applications are becoming increasingly critical in agriculture, where, for instance, the difficulties of finding sufficient labour to pick and sort fruit became apparent during the pandemic.
The next generations of fruit-picking robots will use computer vision to identify a ripe fruit or vegetable of the correct size, identify a path to harvest it, and execute the harvest with an implement without banging the tree or other branches.
Driven by machine learning, the new smart Amazon supermarket, for instance, will use special cameras with machine-learning computer vision to work out the exact contents of a customer’s shopping basket, no matter how many times the items have been swapped and replaced, and deduct the charges from a bank later that day – eliminating the need for exit price checks and immediate payment.
In far loftier spheres, computer vision will also be essential in the effort to explore space, van den Hengel says. It takes eight minutes for a message signal from Earth to reach Mars, so robots deployed on the red planet must be able to act autonomously; to see a problem, decide on a multi-part solution and then take action: this sequence is enabled by computer vision.
“We’re developing the technology that will enable the next generation of these applications,” he adds, noting the Institute has a specific innovation fund to commercialise research into the sorts of machine learning applications that will have a global impact.
“Australia needs to change the way we relate to this technology. This is the electricity of our future,” van den Hengel says
Employing 166 researchers and computer scientists, the Institute assists Australian industry and governments with a vast range of machine learning applications. For BHP it was improving mine safety; for Ausminerals, it was finding more profitable mineral resources; for the South Australian government it was in the fields of agriculture, minerals exploration and health; for Canon, it was developing a touchscreen interface.
For South Australian special effects company Rising Sun Pictures, the Institute helped with ways of using machine learning to seamlessly replace the faces of stunt performers with the faces of key actors in the combat scenes of a recent Marvel blockbuster.
Despite these successes, to date there hasn’t been an Australian data-based company participating globally, such as Uber, Google or Amazon.
The skills are there, the ability is there, but Australian entrepreneurs have yet to make an entrance on the world stage of data-driven enterprise.
“The question is whether Australia is going to participate in this new, global, machine learning-enabled economy, or are we just going to be a place where Californians come to make money,” van den Hengel says.
Future enterprises will spring from sustained and well-funded research, he adds.
In the vanguard is the Institute’s recently launched Centre for Augmented Reasoning, a $20m federally funded initiative which will develop high-calibre machine learning in a field which combines an advanced ability to learn from data (using traditional machine learning) with the ability to reason.
“Research capability in machine learning is the source from which a whole lot of value springs,” adds van den Hengel.
“It’s the seed, it’s the core by which you create this opportunity. This is what the Centre for Augmented Reasoning is trying to achieve. We’re trying to create a core of world-class research capability that will provide the seed for Australia’s participation in this incredible economic opportunity.”
The synergy of Lot Fourteen, fed by the sheer proximity of cutting-edge researchers working in hi-tech, world-class, research-driven enterprises will be a boost, he believes.
Just the other day, van den Hengel found himself in the coffee shop line discussing the overlap between machine learning and quantum physics with the man who has made the world’s most accurate clock.
“That feedback is essential,” van den Hengel says.
“It means, not least, that there’s a job around the corner doing something interesting on Lot Fourteen. There are so many hi-tech jobs. So many opportunities to do interesting things with interesting companies.”