Facing up to AI

Just imagine. A photorealistic digital human underpinned by generative artificial intelligence (AI) communicating with an actual human in real time. Sooner than you think, you won’t have to imagine. This science fiction idea is fast becoming part of our AI future – but what would it look like?

Think about existing digital assistants like Siri or Alexa. Instead of being disembodied voices parroting pre-packaged phrases, they are realistic digital people on a screen who literally converse with you, answering questions, flagging problems, giving suggestions, and making jokes. All done with appropriate facial expressions and tones of voice. You might even think they have emotions, warmth. It could easily feel that way. Humans are hardwired to engage with faces, even in random objects like clouds. 

These digital humans could have many digital uses. You might find one in the home of a housebound person who needs a regular medical check-up; in a financial advice setting; as the knowledgeable face of an online ordering process that guides you, in your purchase and delivery. All this would happen much more efficiently for both the customer and the service provider, with less chance of human error. 

There is little doubt this technology will be socially and economically transformative – there have been predictions that broader AI technology will add $15 trillion to the world economy by 2030 – but the technology is still evolving. One of the people driving the evolution is Dr Mike Seymour, who works out of the University of Sydney Business School. 

“It might seem like the Information Technology school would be our natural home,” says Seymour, who studied pure mathematics at the University with a sideline in fine arts just for pure fun. “But business is so closely entwined with technology these days, and the Business School has really embraced us.” 

Most of the work happens in what’s called the Motus Lab, which also partners with local and international industry for advanced research projects. 

Seymour co-founded the lab six years ago with Kai Riemer, Professor of Information Technology and Organisation at the School. His expertise encompasses technology adoption and the philosophy of technology – both useful in the AI environment. 

Where other labs might have beakers and 3D printers, Seymour’s lab looks more like a hi-tech TV studio. As he talks online to SAM, he is surrounded by six cameras, including motion capture cameras that make the digital characters in films like Avatar possible. 

It’s worth noting here that people like Seymour and his research partners rarely use the term AI in their work environments. It is too non-specific. The term most often used is ‘machine learning’, where a machine processes information in a more human way that recognises patterns in vast amounts of data, extrapolates plausible new ideas, and predicts possible outcomes. 

This is the basis of one of Seymour’s most important tools, neural rendering, which can fill in the gaps in 2D photographs and turn them into fully realised and dynamic 3D images that are currently transforming movie-making, video games, and yes, the creation of digital humans. 

British science fiction writer and inventor Arthur C Clarke once said: “Any sufficiently advanced technology is indistinguishable from magic,” and the world has certainly been awe-struck and even disquieted by the AI it has seen so far. It’s still early days, as Seymour points out.

The ethical side of the AI revolution is a key part of the research being done by Seymour and the University, as the media eagerly reports the damage that AI might do, like millions of jobs lost across areas as diverse as teaching, journalism and finance (apparently plumbers are pretty safe), and social and democratic instability through the easy production of text and video disinformation. 

“Everyone here is taking this very seriously. We’re onboarding new PhD students, specifically to research these issues,” says Seymour. “But the number one protection we have from people who might misuse AI is an informed public. So we’ve done TEDx talks, two Vivid Sydney events, podcasts, TV and radio interviews and public forums. 

“I’d also suggest that maybe decades of science fiction movies about dystopian futures have trained people to be suspicious of AI technology.” In fact, Seymour himself helped make some of those movies during his previous career as a movie visual effects expert, with an Emmy nomination under his belt, an AFI win, and time spent running his own post-production company. Although, he seems particularly proud to have worked on the cult comedy Red Dwarf, in the United Kingdom. 

When asked how he managed to tear himself away from Hollywood to become an academic, his answer is simple: “When I started giving talks to other people in the industry, I realised I loved teaching.” 

A new reality

While Seymour acknowledges that AI will dramatically change the employment landscape, he can see that other jobs will emerge because of it.

An example comes from the 2020 Polish movie The Champion. It needed dubbing into English for the international version. Dubbing is always clunky, so Seymour helped to design a complex new technical pipeline to digitally change the mouths of all the actors so that it seemed they were actually speaking the English words.

“It worked really well,” says Seymour. “And it’s now spun out into a successful Australian startup with dozens of highly skilled professionals using machine learning to create export income for Australia through the international film industry. I think it’s something the University can be proud of.”

This kind of skills creation is exactly what Australia needs if it is to get full benefit from the AI revolution, with a report recently released by the federal industry and science minister noting that Australia is “relatively weak” in emerging AI technology. To fully engage with it, we’ll need more skilled workers and more computing power. 

“In the last federal budget, there were three key strategic areas for the economy. One of them was advanced technology,” says Seymour. “We’re working with people across the University to see how this technology can work for them – health, education, psychology, even agriculture.