����ɫ��

“AI, as we know it – like ChatGPT – wouldn’t really exist, or wouldn't really be intelligent without us,” he said. “So, in some ways, when I think about AI, I don't think about AI as an artificial intelligence. I really think it as a reflective intelligence. It's reflecting ourselves back [to us].”

“Everything is AI, but it's really ours to seize,” Professor Lucey continued. “Because if we can come up with innovations and a way of challenging some of these fundamental problems, we can really push the envelope of what's possible.”

EMCR presentations

Several of AIML’s world-class early and mid-career researchers (EMCRs) spoke before the crowd, providing comprehensive overviews on their areas of work and research.

PhD student Fengyi (Frances) Yang presented on her work, ‘Robust Fitting on a Gate Quantum Computer,’ done in conjunction with AIML Professor Tat-Jun Chin.  Their work was selected as an oral and best paper candidate at last year’s (ECCV).

She described her research as “an exciting emerging shift in how quantum methods are being integrated into mainstream computer vision research” noting that it outperforms existing quantum processes.

Next, PhD student Duc Duy Nguyen detailed his work on ‘Class-Agnostic Repetitive Action Counting Using Wearable Devices,’ which takes information that wearable devices such as smart watches provide (i.e. number of steps the wearer is taking) and expands it even further.

“We want the user to count anything that they like,” he said. “[Right now] it's nearly impossible for the user to actually define what they want to count on the screen.”

“CaRa Count, which stands for Class Agnostic Repetitive Action Counting, allows users [to] specify what they want to count by providing a short example sequence and later use the information to count [the] number of repetitions in the query.” CaRa Count, which was accepted to the (AAAI) conference last year, outperformed other similar methods.

Last up was post-doctoral researcher, Dr Nikhil Kurian’s presentation on “How Bias is Encoded in Medical Image Analysis Models.”

“Biases are encoded throughout the [medical] decision pipeline,” said Dr Kurian. “The consequence is that the subgroups that are already suffering from lower sensitivity [i.e. Black and younger patients] are also those for which the model is most certain in making a mistake, which compounds the clinical risk.”

“[These biases] reinforce structural disadvantages,” which may lead to some of the very medical concerns being treated.  

Nguyen was awarded the EMCR Oral Presentations Award for best presentation given out later in the afternoon by Professor Lucey. 

Career progression session

A highlight for many in attendance was the Career Progression session where AIML researchers with slightly less traditional career paths spoke on how the work and opportunities they’ve received during their time at AIML has impacted their careers.

PhD student Cameron Gordon spoke glowingly of a computer vision internship he had with Apple several years ago which came about because of his studies at AIML.

“Researchers at AIML and Apple at par,” he said. “Every single individual who is coming through the doors of AIML has the capacity to work at these large tech organisations.”

When asked by AIML Institute Manager, Dr Kathy Nicholson, what are some of the things that he thinks those currently studying at AIML should be putting their hand up for, Gordon encouraged AIML members to “seek collaboration with other organisations.”

“Visiting and working with other organisations globally [is] one of the things I think is most valuable,” he said. “Taking the opportunity to work with other institutes [and] having the opportunity to explore adds the most value.”

Dr Mohsen Dorraki, an AIML lecturer, shared that the skills he’s learned through his career as well as the mentorship and support he’s received from AIML members has provided him with a long list of valuable and useful skills, including how to communicate with non-academic partners, project management, and translational thinking.

AIML Researcher, Dr Zhibin Liao, humoured attendees with an honest discussion around his decision to go into lecturing. 

“It is time consuming to be a good teacher,” he continued. “Being a lecturer means that you have to know everything you say through the lecture. If there's a question, you need to be able to explain it well.”

When asked what surprised him most about lecturing, Dr Liao replied “how things can go wrong so fast.”

“There are so many great bits about teaching, and some bits I’d like to do better,” he said. “When I'm drowning in the teaching mud, I…realise it’s an honour to be able to present… the researchers before you [and] their ideas. Hopefully that gets through to students and inspires their future career.” 

Poster presentations and more

AIML Chief Scientist, , wound things down with a discussion of the current AI landscape that was both grounded and philosophical.

“We are still at the beginning of this [AI] process,” said Professor van den Hengel. He noted that “we’re asking the wrong questions” around artificial general intelligence (AGI), an emerging AI tool that aims to replicate the broad intellectual capabilities of humans. 

“People are saying that AGI will provide a living wage because it's going to take away most of the jobs, and that it’s going to take over,” he said. “AGI is not coming anytime soon. It's not going to cause mass unemployment. It's not going to do most of the jobs of humans. None of that, I'm afraid, is going to come true.”  

After his presentation, Professor van den Hengel accepted the along with Professor Javen Shi for their 2015 publication, ‘Image-Based Recommendations on Styles and Substitutes.’ The Test of Time Award recognises papers or research that have had a lasting impact on their field years after their original publication.

The event concluded with the traditional poster presentations from AIML EMCRs highlighting their areas of research and work as well as conferences where they’ve presented.