Submission 82
UQ AI Research Network and Enterprise AI Centre
Submitter group: Research institutes.
An AI model identified the submitter group and wrote the summary, key points and recommendation summaries. Counts do not measure importance or merit.
Summary
Read the original submission for the full text.
Submission from the Steering Group of the UQ AI Research Network, not an institutional submission from The University of Queensland and not the University's corporate position. Argues Australia's domain-specific AI opportunity is a translation problem, not only an infrastructure one, covering terms of reference c, e, f and g: governed data access, translational funding, interdisciplinary programs and dual-competency researchers.
Key points
Quotes are the submitter’s own words. Text taken from documents, including scanned pages read by text-recognition software, can differ from the PDF. Check the original. Listed in the same order as on the submission’s page in the interactive guide.
Argues that compute is necessary but insufficient; the key constraint is translational capability combining domain and technology expertise.
compute is necessary but not sufficient. The key constraint is translational capability which requires people, data and institutions fluent in both the domain and the technology.
Says each research team currently negotiates access to Australian health, agricultural or environmental data separately, project by project and jurisdiction by jurisdiction.
A model trained on Australian health, agricultural or environmental data requires lawful, ethical, auditable access to that data at sufficient scale. Today each research team negotiates this separately, project by project, jurisdiction by jurisdiction,
Says governed data holdings must follow Indigenous Data Sovereignty and CARE principles.
We endorse the position that governed data holdings must follow Indigenous Data Sovereignty and CARE principles.
Argues that sovereign capability is achieved when a model is in service, not when it is trained.
Sovereign capability is not achieved when a model is trained. It is achieved when the model is in service.
Argues that research training and funding systems poorly support people competent in both a domain and AI.
The system produces excellent AI researchers and excellent agronomists and leaves the space between them thinly populated by people who succeeded despite the incentives rather than because of them.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Deliver AI through missions requiring cross-disciplinary leadership and place-based centres uniting industry, government and research consortia; adapt funding machinery to AI's transdisciplinary nature.
AI should be delivered through mission-oriented, programs that require genuine cross-disciplinary leadership, and place-based centres of excellence that bring together strong consortiums across industry, government and research4. Australia’s funding machinery needs to better accommodate the transdisciplinary nature of AI innovations.
Establish nationally interoperable governed access to sectoral data, with standardised, auditable pathways and trusted research environments, reducing duplicative project-by-project negotiation while retaining ethical, legal, community and Indigenous data-governance requirements.
Establish nationally interoperable governed-access for sectoral data, with standardised, auditable and ethically governed access pathways and trusted research environments. The goal should be to reduce duplicative project-by-project and jurisdiction-by- jurisdiction negotiation while retaining appropriate ethical, legal, community and Indigenous data-governance requirements.
Extend translational funding for domain-specific AI beyond the research phase to cover real-world validation, regulatory evidence generation, software and workflow integration, procurement and adoption, deployment, and post-deployment monitoring, evaluation and governance.
Extend translational funding for domain-specific AI beyond the research phase to cover real-world validation, regulatory evidence generation, software and workflow integration, procurement and adoption, deployment, and post-deployment monitoring, evaluation and governance, recognising that a trained model is not yet a sovereign capability.
Design research training for dual competency with better aligned PhD Placement programs.
It requires research training programs designed for dual competency with better aligned PhD Placement programs (see recommendations from a national study6),
Treat dual-competency researchers, fluent in both a domain and in AI, as a distinct and strategically scarce workforce category.
Treat dual-competency researchers, fluent in both a domain and in AI, as a distinct and strategically scarce workforce category.
Fund cross-disciplinary doctoral and postdoctoral programs.
Fund cross-disciplinary doctoral and postdoctoral programs,
Require joint appointments in nationally funded AI missions.
require joint appointments in nationally funded AI missions,
Direct research funders to constitute assessment panels capable of evaluating cross-disciplinary AI research on its technological and sectoral merit.
direct research funders to constitute assessment panels capable of evaluating cross-disciplinary AI research on its technological and sectoral merit.