Aaron Quigley of CSIRO spoke about the ways AI can be used to improve workflows and connect people to their own form of personalised healthcare.
While the recent AI hype has focused on large language models like ChatGPT, there are other ways artificial intelligence systems can benefit humanity and improve how we process data.
One sector that has the potential to be significantly boosted is healthcare, with a number of AI-based projects taking place worldwide. Recently, Northern Ireland’s Sonrai Analytics entered a partnership to deploy its AI tech in the United Arab Emirates to improve cancer patient outcomes.
Last year, health-tech start-up xWave Technologies teamed up with CeADAR, Ireland’s centre for applied AI research, to make medical testing more predictive and personalised using AI.
Aaron Quigley is the science director and deputy director of CSIRO’s Data 61, the digital specialist arm of Australia’s national science agency. Some of his work focuses on using machine learning to analyse large datasets to find ways to improve healthcare, along with ways for people to monitor their own health.
Quigley told SiliconRepublic.com that the one of the Data 61’s missions is to “use AI to accelerate the scientific discovery process”.
He also spoke of the advances AI has undergone over the years and said big tech companies in this field are benefiting from the work worldwide academia has undertaken “all the way since the 1940s and 50s”.
“We’ve had this real spike in the last couple of decades. It’s a human capital solution as much as a research, innovation solution, there’s just a lot of people who have the skills now to do this kind of advanced AI research.
“But not in every area, there’s plenty of areas where it’s kind of languished or really hasn’t moved as fast as we might have expected”.
When discussing AI’s role within healthcare, Quigley said one element is “augmenting the workflow that people already have” rather than an outright replacement of roles.
One example of this he gave was Data 61’s work to enhance work displays, by using computer vision techniques to analyse where a person’s eyes were focusing when looking at a monitor.
With these techniques, Quigley explained that it’s possible to tell if an employee missed an important pieces of information while looking at a screen.
“Sometimes your eyes aren’t looking at the screen where changes have actually been made,” Quigley said. “We can detect that using simple computer vision techniques, using very simple sort of AI methods. And then we know you haven’t actually seen an important notification.
“So when your eyes come back to that screen, we know you are looking at that screen, we can basically give you sort of a visual update to say this has actually changed on the screen since you’re looking at it. Because our studies found that without this people were missing all of these very important notifications.”
Last month, Microsoft announced it’s working to bring the capabilities of GPT-4 into healthcare in order to “automate clinical documentation at scale”. The tech giant said the purpose of this new product is to reduce administrative burdens and help clinicians spend “less time on paperwork”.
Quigley said he was “not surprised” by the way Microsoft plans to bring AI to healthcare because “what we’re all trying to do with AI is to augment the human experience, not to replace it”.
“What will happen is the very basic techniques, the very basic set of steps that a lot of people have to kind of go through to bring together information, that will become much more automated,” Quigley said.
“But then we will focus on the higher value types of activities, the work that kind of requires more thought and thinking, and those simpler elements will be taken over or be supplemented by AI.”
At the start of the year, one of the predictions around AI’s development this year was the growth of ‘augmented intelligence’, which is described as the human-centred partnership with AI and humans working together to retain a ‘human perspective’.
Quigley also highlighted the importance of AI being used in a “responsible” and “traceable” manner, as he feels a “day of reckoning” will come in the future when regulations arrive and the legal system catches up with AI creations.
“The companies that kind of just jump ahead and do this blindly, without thinking about those later implications, they’ll suffer down the tracks,” Quigley said. “Companies probably need to pause and do this in a responsible way to make sure that it is actually something that they can defend.”
“Because what you don’t want is the scenario where the computer is making decisions that nobody can explain. That’s pretty much the worst case scenario.”
Using AI to fill the gap
In his work at CSIRO, Quigley said a “one health approach” is being taken to find ways to improve people’s health. This involves analysing a combination of human health, animal health and environmental health to produce an optimal “health and wellbeing environment”.
The real-time data being analysed to create the combination requires thousands of data points, with a device placed on the ears of cows to monitor livestock and detect unusual activity, for example.
On the human side, Quigley explained that available healthcare data from CSIRO projects such as the Total Wellbeing Diet allowed its researchers to make their own AI system, Hope, which can monitor people’s health and improve the work of dietitians.
“So a limited number of dietitians could actually support hundreds of thousands of people actually using AI techniques to give prompts and suggestions and guidance to kind of fill the gaps,” Quigley said.
Australia is also working on a project called AquaWatch, which aims to create a sort of “weather service for water quality” to safeguard freshwater and coastal resources. Quigley said that this system aims to predict “what’s happening with every dam, every river, all the entire water systems around Australia”.
Due to the sheer volume of data that needs to be analysed and predicted, Quigley explains the importance AI plays in these systems.
“How would a set of humans do that?” Quigley said. “How would you be able to collect hundreds of millions of images, having all of these different sensor systems…All of that data has to be analysed and managed. And then you’re trying to make predictions.”
10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.