The future of AI: Professor Svetha Venkatesh

Considered one of the top artificial intelligence institutes in Australia, Deakin University’s Applied Artificial Intelligence Institute (A²I²) is co-directed by Professor Svetha Venkatesh, an Australian Laureate Fellow with more than 30 years of experience in the field of pattern recognition and big data.

A²I², which launched in April 2019, is focused on developing cutting-edge solutions to problems of societal importance, including critical areas such as health, security, manufacturing and aged care.

To find out more, we asked Professor Venkatesh a little bit about her background, A²I²’s achievements so far – and what the future of artificial intelligence (AI) might look like.

How did you first become interested in AI?

I’ve been in the area of pattern recognition for a very long time – almost 30 years. Devising ways to understand how patterns work and how they can be interrogated for the broadest number of possible purposes is my passion. Over the years, I have loved to identify unsolved problems and construct rigorous computational models, theory and algorithms to solve them.

The field of pattern recognition, which is what it was called then, slowly grew into more specialised machine learning and now advances are being made in the field of AI. So I have grown with this field.

What was it like recently being ranked in the top 15 women working in AI, according to UK innovation firm Nesta?

Of course it’s a great honour, and I’m very thankful that I have worked with a wonderful team that has allowed us to do really innovative work, which is recognised through some of these kinds of awards.

Your research has led to two start-ups, including iCetana, a software company that provides video analytics solutions and was listed on the ASX in December 2019. Can you tell us how iCetana came about?

iCetana began as a PhD project of one of my students, Dr Saha Budhaditya, Senior Data Scientist at Medius Health and, Sydney. It was a very interesting journey.

The route to commercialisation is a very long one, but I met this wonderful venture capitalist, Matthew Macfarlane, at that time. We formed a really unique partnership and we worked closely together to commercialise the underpinning technology.

iCetana is now 10 years old, which is a long time for a start-up. Now it’s a major supplier of security solutions to companies like Boeing and Facebook, so we’re very happy with how it has turned out.

These commercialisation projects are a really hard slog and it really requires a good partnership between an inventor and a venture capitalist. I have to thank Matt (who is now the CEO of iCetana) for the unique friendship and partnership.

What’s the overall vision of A²I²?

We want to produce impactful, cutting-edge machine learning and AI solutions. We are a unique institute in that we choose to take on problems that not only help solve important societal problems, but will also push the boundaries of machine learning.

We have reached many milestones in this space. We have made important contributions in automated security – iCetana and the Virtual Observer are commercial examples. Together, these two technologies were the first to provide solutions for automated static and mobile security in large-scale, urban spaces.

Another example is that we addressed the problem of long waiting times for therapy for children who have just been diagnosed with autism. In a unique partnership with therapists, we built an early intervention system for children with autism called TOBY Playpad, which has been used by over 8,000 children worldwide.

We’ve constructed important risk prognosis algorithms for health – for example, improving the accuracy of predicting readmission for a number of chronic health conditions.

More recently, we have embarked on accelerating the very process of experimental design itself, allowing faster design of novel materials and processes.  This includes the design of new aluminum alloys, and alloys with special properties.

Our Institute is comprised of over 30 fundamental machine learning and AI specialists, 27 PhD students, and over 30 software engineers who can then translate these ideas into prototypes. It’s this unique combination of skills that makes the Institute a wonderful place to be in today.

Does A²I² focus on any particular sector?

Our priorities are security and health.

More recently, a pivotal pillar of the Institute is accelerating the process of innovation itself. This is being achieved through advances in machine learning, which allows us to speed up the way in which materials or processes are designed and constructed by doing the minimum number of experiments.

We have diverse partners in this space. We have collaborated with Professor Steve Allender, Director of Deakin’s Institute for Health Transformation’s Global Obesity Centre to design Machine Learning driven controlled trials to identify the best strategy to increase the number of exercise-scripts issued by GPs in as little as seven weeks (this would have normally taken more than an year). This method shows huge promise in accelerating medical trials.

We have also collaborated with Professor Matt Barnett, Director of Deakin’s Institute for Frontier Materials (IFM) to design the first alloy with the broadest composition with targeted strength – this could be a flexible way to recycle many elements.

What stage do you believe AI is at?

AI, I would say, is very much in its infancy.

The part of AI that we have solved is the initial part, where you can take a signal and extract patterns. What we haven’t done yet is to build useful reasoning or planning systems and embed them in context. In fact, the best algorithms we have can’t even learn as well as a small child, so we have a long way to go.

Going forward, I think the vision for A²I² is to continue to produce these cutting-edge machine learning and AI technologies, translate them into products that have societal impact and, most importantly, engage with the community so that we produce research that fills a real need in society.

One of the areas that is very important to us as we develop AI is the ability to provide assurance to humans regarding what is inside the AI. We need to be able to explain our algorithms to the people who deploy them. To help address this, we have invented a new area called algorithmic assurance, where you have algorithms watching other algorithms to see if the algorithms are functioning as they should.

AI has opened up an important frontier where the  community has to decide how and when to share autonomy with machines. A²I² wants to be part of this conversation; what kind of society do we want to have, and what is the role that AI will play in it?  We don’t want it to be a runaway technology, we want it to be the Australian AI version – a version that we’re happy to have in our communities and that will advance our futures.

Written by Larissa Ham