AIML-sponsored and hosted conferences end 2025 on a high note

In December, AIML co-hosted and presented two high-profile conferences, the 26th International Conference on Digital Image Computing: Techniques and Applications (DICTA 2025) and the Next Generation Responsible AI Symposium, ending 2025 on an incredible note.

From 1-2 December, the Responsible AI Research (RAIR) Centre and its partners, the Government of South Australia and CSIRO’s Data61, hosted the Next Generation Responsible AI Symposium held at Tech Central in the Lot Fourteen Innovation District. The symposium was a seminal two-day event created to develop and discuss ways to ensure that artificial intelligence (AI) is ethical, inclusive, and equitable.

The event attracted over 170 attendees and kicked off with a keynote presentation and panel discussion hosted by Dr Sue Keay, Director of the UNSW AI Institute and founder of Robotics Australia Group.

During her keynote speech, Dr Keay noted that “8 years of relentless under investment” in AI has caused Australia to fall behind other nations in having conversations and developing strategies around AI.

"Australia leads the world in fear of AI," she said. She stated that the nation’s lack of AI strategy and its reliance on technology from other countries further hindered the country’s development in the AI space and recommended that Australia become “less reliant” on AI models coming out of other nations.

Panel discussion

After her keynote, Dr Keay introduced the members of a panel discussion to answer, “What is Responsible AI?” among other topics.

Panel member and AIML Director, Professor Simon Lucey, responded that responsible AI is “the type of AI that we want.”

“Responsible AI is the North Star, allowing us to achieve the type of AI that we want” and to address specific challenges, he continued. “I’m so excited about these new frontiers.” Professor Lucey noted that a critical component of all AI development is the creation of systems that allow humans and machines to work with and complement each other, stating that this was an area ripe for early career researchers (ECRs) to move into. 

Jon Whittle, AI and innovation executive, said he believed that responsible AI will provide developers and users with the ability to “maximise the positive and minimise the negative” as it relates to AI’s development.

“Responsible AI is less about the application of AI and more about the implementation,” he said. “What are the unintended consequences of AI?” Dr Whittle suggested that proper governance and inclusive AI-focused consultations will address these issues and “bring society along” in utilising AI systems that are ethical and responsible.

Dr Andrew Dunbar, Executive Director, Industry, Innovation and Small Business for the Government of South Australia, stated that responsible AI “ensures that AI aligns with human advances as it evolves.” Dr Dunbar agreed with Dr Keay’s assertion that Australia needs to be a leader in AI development and not rely on models from other nations.

Palawa woman, ANU Professor, and Director of Old Ways, New, Professor Angie Abdilla, questioned the use of the phrase “sovereign” when seeking to develop Australian AI capability that is independent of other nations. She stated it was important to ask how the ‘North Star’ and other features of responsible AI are being defined.

Professor Abdilla noted four terms that she considers when thinking about how AI systems can be used fairly and equitably.

“When I think of responsible AI, I think of AI that involves reciprocity, regulation, regeneration, and relationality,” she said. “All automation systems should be working towards [supporting] those with the most need first” which would ensure robustness and make systems more dynamic. Professor Abdilla observed that Australia’s unique environment requires technology that can utilise the experiences of its rich and diverse structure. 

Day 1 of the symposium ended with a networking event at Stone & Chalk, located at Lot Fourteen. At the event, attendees mingled and viewed posters that had been put together by responsible AI researchers in attendance at the conference.  

Breakout sessions and awards

The Next Generation Responsible AI Symposium also featured breakout sessions and workshops across the two-day event. The sessions, delivered by a cohort of over 40 presenters – more than half of which were women - covered all things responsible AI, including ways to address AI bias, the importance of diversity and inclusion in generative AI, and understanding who AI impacts most and how.

Day 2 of the Symposium brought attendees together for workshops to discuss four key themes: Responsible AI (RAI) for Societal Equity; RAI for Sustainability; RAI for Organisational Impact; and RAIR for Health and Wellbeing. The workshops were hosted by AIML Deputy Director Dr Melissa McCradden and researcher Dr Erdun Gao; as well as CSIRO’s Dr Muneera Bano, Dr Petra Kuhnert, Dr Kavi Katuwandeniya, Dr Sunny Lee, and Dr Sarah Hartman; and Dr Rifat Ara Shams from Monash University.

Day 2 also featured an important communications session that provided researchers with vital information on how to engage with the general public and the media through outlets such as The Scientist and Science Meets Parliament. The presenters, Caitlin Curtis and James Hereward from the University of Queensland, provided helpful tips for researchers to ensure that their message gets to those they are most trying to reach.

Awards were also presented during the symposium. AIML PhD students Lana Tikhomirov and Luke Smith won Best Oral Presentation for their talk, “One AI to (Not) Rule them All: Re-analysing an AI reader study in radiology using techniques from cognitive science.” Other winners include "Making AI Work Visible with the GenAI Arcade" by Dr. Aaron Snoswell from the ADM+S Centre at Queensland University of Technology; "A Case Study in Human Label Variation: Developing Fair Models Utilising a Diversity of Opinions" by Dr Kemal Kurniawan from the University of Melbourne; and "The devil is in the details: Alignment failures in Adversarial Machine Learning" by Dr Andrew Cullen, also from The University of Melbourne.

The event concluded with a combined Next Gen Interlink networking event at the University’s Elder Hall for attendees of the Next Gen RAIR Symposium to network with those attending the DICTA event starting the next day.

“Many users of this game and other tools rely on ‘vibes-based guessing,’ said Professor Hays, noting that the locations in the game move fast but require little actual detective work. Professor Hays discussed the importance of creating well-developed reasoning chains when working in the geo-location space, and observed that based on the number of hallucinations, vision language models (VLMs) likely ‘vibe guessed’ as well.

Dr Angela Yao, Associate Professor in the School of Computing at the National University of Singapore, conducted a “VLM fitness check” on Day 2 of the conference where she presented her work collecting 4000 videos of people putting toys together over a 15-hour period in order to train models to better identify users and relevant objects.  

Professor Yao, who also presented at AIML’s 2024 Research Showcase, noted that while AI excels at perception, it still struggles with situational context.

“Situational and user context and awareness is missing,” she said during her presentation. “Contextual AI is the next frontier for embodied intelligence.”

Speakers say AGI is no threat, discuss importance of closing moral loopholes

Day 3 of DICTA saw AIML Chief Scientist, Professor Anton van den Hengel, expand on previous discussions of AI as a ‘stochastic parrot’ parroting our knowledge back to us, and that models need far more data than humans do to learn.

“You have to have memory to be able to do things,” said Professor van den Hengel. “Robots have to have tasks defined because robots don’t have memory.”  Professor van den Hengel believes that this lack of memory is what will keep us from ever achieving artificial general intelligence, or when a machine possesses the ability to understand or learn an intellectual task that a human being can.

“AGI doesn’t exist,” he said. “We are in no danger of singularity,” the point where AI surpasses human intelligence and can improve itself better than humans can. “We don’t need ‘do no harm’ rules because robots aren’t doing anything without human intervention.”

Alex London, Professor of Ethics and Computational Technologies at Carnegie Mellon University, wrapped up the keynote speeches with an energising talk on AI ethics and the effective development of responsible AI.

“Moral loopholes create incoherence,” said Professor London. “If loopholes exist, actors can act with impunity or influence a process or outcome without accountability.” He provided several examples of how companies’ use of AI created moral loopholes that led to incoherent policies and processes, lawsuits, and other negative outcomes. He suggested that the best ways to close moral loopholes was to increase public awareness, hold senior leaders accountable, encourage the company’s Board of Directors to implement check and balances, and utilise regulation and litigation when needed.

Additional events

In addition to sensational speakers, DICTA 2025 was also home to over 30 informative breakout sessions, talks, and workshops across a range of topics.  Presenters included academics, industry practitioners, and early career researchers (ECRs) just beginning their academic careers.

Awards were also presented. The Australian Pattern Recognition Society (APRS) presented the 2025 Early Career Researcher Awards 2025 to Feng Liu of the University of Melbourne, and Jing Zhang of The Australian National University. 

The APRS/IAPR Best Paper Award was presented to Qi Tan, Rong Wei, Zhiyu Xi, and Jingqing Yang for their paper, 'Consistent 3D: Diffusion-Driven Sparse View Completion and Reconstruction with Geometric Priors.'  

AIML Associate Professor Johan Verjans and his colleague Nadhir Hassen were awarded the Best Contribution to the Science of Pattern Recognition Award at the conference. The award for their paper, 'Flow-selectivity SSM: A Generative State-space Model with History-Aware GFlow Net Policies" was sponsored by the Defence Science and Technology Group. 

Organisers and members of Leading Coding, an Adelaide-based education centre that offers university-level programming courses for kids and youth, were also recognised in a special event. Congratulations and well done to all. 

DICTA Attendees were also treated to a fun run/walk along the Torrens River that was great fun for those who participated. And among the smash events of the conference was the gala dinner held at the Convention Centre that featured Michael Brown, MP, Assistant Minister for AI, the Digital Economy, Defence and Space Industries; a live band; and lots of dancing computer scientists.

We send our sincerest gratitude and congratulations to the organisers, attendees, presenters, and volunteers of both of these events. Your involvement and support made them the phenomenal successes that they were.

Please click here to sign up for the AIML Mailing List and be provided with information on all of our upcoming events.