Generative AI (genAI) is already having an impact in Australian science research, and it has the potential to even playing fields and advance discovery at an unprecedented rate. 

“It’s a tool for empowerment and democratisation of research and also acceleration of research,” says Professor Michael Small, Director of the University of Western Australia (UWA) Data Institute.

Small, who himself predominantly uses traditional machine learning in his own research at UWA, uses his role at the Data Institute to foster a community of practice within the university that harnesses the opportunities genAI offers to scientific research – lowering the bar in terms of access, as he puts it.

He also says there’s no question about the place of AI in universities: “we need to look at using these tools … in the right way, in the ethical way, in an appropriate way. And to do that, we need to empower the researchers to understand what is the right way.” 

The right way has increasing definition. In April, the NHMRC and ARC firmed their policies on using genAI in grant applications: as a preparation aid for applicants, and effectively as a proofing agent for assessors. Encompassing these allowances are also the principles of transparency – declaring its use – and outlining consequences for breaches.

TEQSA has also ringfenced responsible genAI use for researchers, preparing resources emphasising accountability and disclosure principles, protecting privacy – encouraging academics to acknowledge its limitations while pursuing the broad opportunities it offers.

Life-saving discoveries

Professor David Ascher is the Deputy Associate Dean in the Faculty of Science at The University of Queensland. He leads a group that harnesses AI to investigate genetic variation and its impact on proteins.

“Even though AI is not new, the new models and approaches have opened up a lot of opportunities,” Ascher says. He points to joint research with the Victorian Tuberculosis Program addressing drug-resistant tuberculosis.

Ascher says it’s challenging to perform genetic testing, but one of his PhD students used AI to show how mutations affected protein structure. They were able to “accurately identify patients that were likely to be resistant to [the TB treatment] pyrazinamide … That moved through very quickly to practical use,” Ascher explains.

In conjunction with the Baker Institute, Ascher’s team was also able to identify patients at risk of heart failure. “By using the measurements that are coming out of a patient’s ECG, we were able to pick up really minute signals using an AI algorithm that would be able to separate patients that were likely to undergo heart failure.” That discovery has now progressed to trial.

Fostering ingenuity to simulate the universe

Over at the Swinburne Centre for Astrophysics and Supercomputing, Professor Darren Croton has harnessed genAI to build upon his Theoretical Astrophysical Observatory, or ‘TAO’. It’s an open-source galaxy formation simulator originally built from large cosmological and galaxy formation simulation models. Now Croton and his associates use genAI to code and deploy new modules to expand TAO’s capabilities. 

“My team and I are using AI to accelerate and leverage the development of those,” Croton says.

Yet he’s most excited by what his students are doing with AI, pointing to one who used large language models (LLMs) to build upon a preexisting visual browser for the TAO.

“It’s about 400 million times better, it’s more powerful, more flexible, more customisable,” Croton says. “He’s not a hardcore programmer; he leveraged LLMs a lot, but he still directed it. He had the vision and the purpose and was able to describe what he wanted.”

Croton too acknowledges the pitfalls of unethical AI use. “My personal thing is: am I professionally comfortable with doing something?” he says, but he’s an optimist about what the technology will offer science.

“I’m getting old now,” Croton says, “but it’s the students, PhD students and the postdocs, the early career researchers coming up. My hope is they’ll learn how to use these tools effectively. It won’t replace their thinking or their understanding, or their ability to do science. It will enhance it.”

 – Written by Matthew Ward Agius