Professor Taylor, who is computational platforms research group leader at CSIRO's Data61, has considerable experience of high-performance computing in Australia and the US, including its use for simulation.
The idea of building a Virtual Australia in conjunction with many other organisations is to provide a safe environment for testing autonomous devices. Nvidia this week announced Drive Constellation for the virtual testing of autonomous vehicles, but Prof Taylor wants to generalise the idea to other devices.
"Driving in Australia is very different," he said, pointing to the significant stretches of dirt roads and the presence of kangaroos (which have proved challenging for object detection systems).
It would also be useful for those working on the smart cities concept. Just in terms of transport modelling, it could be used to investigate what would happen if current public and private passenger road transport was largely replaced by roaming 'pods' providing point-to-point service.
Using 'foreign' models might result in products that do not perform well in Australian conditions. Perhaps delivery drones need to be trained to cope with the afternoon wind changes that are a feature of some Australian cities, and watercraft need to be aware that "our conditions are slightly different." In the jargon of the AI community, Australia's uniqueness presents a number of corner cases.
In addition, Virtual Australia could be made available to school students so they could develop their ideas, he suggested.
And the experience gained in building Virtual Australia could be applied to constructing models of other parts of the world to help local companies create exportable autonomous products - robotics is an opportunity for Australian industries, he said.
Either way, we will need virtual models to check that AIs work correctly and safely under local conditions before they are released into the real world, said Prof Taylor, so it is important that we build those models and create the tools for people to use them.
Physical testing will still be needed, "but it will almost certainly start in the real virtual world" due to the amount of testing needed.
Prof Taylor was a speaker at this week's GPU Technology Conference in the US.
Disclosure: The writer attended Nvidia's GPU Technology Conference as a guest of the company.