Written by Elanor Huntington, Executive Director - Digital, National Facilities & Collections, CSIRO and Stefan Hajkowicz, Chief Research Consultant, Analytics and Decision Sciences, CSIRO
Generative AI – as exemplified by ChatGPT – burst into our collective imagination in 2023, triggering a flurry of GenAI theatre as well as genuine engagement. It’s part of a seismic shift in the nature and reach of AI. But not because I have an app that will remove the drudgery of writing emails, fail to write good Haiku and give me a recipe for a Mojito.
It’s what lies underneath – “foundation models”. These vast, powerful, fine-tunable and reusable models are what powers ChatGPT and its cousins. They’ve emerged as a new class of infrastructure. And we already depend upon them.
Australia’s geographic sprawl has long underpinned our national appetite for infrastructure – largely to connect us through roads, train lines, ports and across the skies. We have an emotional attachment to this infrastructure because it brings us home to our families after work, it takes us across borders to loved ones at special times through the year, and it underpins the daily human interactions we took for granted until the recent pandemic, when we realised just how vital those connections are.
Perhaps less often, we think of the iconic research infrastructure that has underpinned each leap forward in our societies and each inch forward in each of our scientific careers. Everyday Australians may not know about infrastructure like the Australian Centre for Disease Preparedness in Geelong where we study pathogens ranging from COVID-19 to avian influenza and beyond, but many of them will know Murriyang, our iconic Parkes radio telescope, courtesy of the film ‘The Dish’ and our role in landing the first human on the Moon.
Personally, we each might have a piece of research infrastructure that shaped our careers. For Elanor, working on building and using a piece of major research infrastructure as part of the global research effort needed to figure out if we could build gravitational wave detectors was a transformative experience. It has shaped her whole career and worldview on infrastructure. Research infrastructure is about the convening power of a big problem and a big solution; it forces us to think 20, 30... 50 years into the future. So, what infrastructure should we be thinking about now to set us up for the next 20, 30, 50 years?