Insights from AI Event Keynote Speaker Anton van den Hengel

Keynote speaker Professor Anton van den Hengel is a well-known and published international thought leader in deep learning, vision and language problems, image-based modelling, and weakly supervised learning.

​He is the founder of the Australian Institute for Machine Learning (AIML), the former Director of Applied Science at Amazon, and a regular keynote speaker on AI topics and a vocal supporter of developing world class, sovereign AI capabilities in Australia.

Tickets

---

One of the reasons we decided to ask Anton to speak at "Unlocking Australia's Future: Exploring the Power of AI", is that he has been consistently calling for action from federal, local and business policy makers to invest in sovereign AI capability, research and commercialisation. We took some quotes that Anton posted or shared in his LinkedIn feed to give you an overview of the state of things in Australia. 

• Adopting AI in a business is actually very simple [..] The challenge we have in Australia is that there aren't enough board members who understand it.
• Canada's government has just announced a C$2.4 billion (A$2.7 billion) investment in its sovereign AI programs [..] In contrast, Australia's current federal AI programs experienced an approx $30 million cut from the previous government's already modest $124.1 million AI Action Plan announced in 2021.
• It has recently been reported that widespread adoption of AI will add 150,000 jobs to the Australian economy. It seems that some have the idea that these are 150,000 jobs in AI. Nothing could be further from the truth. [If a] business makes toasters then the extra jobs are in toaster-making, not AI.
• It's amazing to see how much has changed at the intersection of AI and Health, and how much hasn't. Originally the arguments I faced were - that AI would never make decisions, it would just be a tool, - that people would never accept an agent as a doctor, and - that AI had to be explainable to be used in medicine. It seems we've got past all of that, which is wonderful. What hasn't changed is that AI is still not being applied in Australia at the scale required to divert the health system from its current perilous trajectory. We're still doing investigations, explorations and reports instead of getting on with the real work.
• Saudi Arabia makes a move to become a global leader in AI, with a rumoured $40 billion investment pledge to make it happen — phenomenal. For local context: Australia's current AI public investment sits at $75.7 million; an approx $30 million cut from the previous govt's semi-delivered $124 million AI Action Plan.
• Singapore's Deputy Prime Minister Lawrence Wong just announced an investment of about 1 billion Singapore dollars (over 1.1 billion AU dollars) over the next five years into boosting AI capabilities. Wong points out that this investment will also be used to "secure access to the advanced chips that are so crucial to AI development and deployment." This will likely have led to the size of the investment. Talking about AI and building AI carry very different price tags... He draws, right there, the demarcation line between followers and adopters of AI, and sovereign developers and builders of AI. 
• Great that Australia Post are using AI to improve productivity. It's a shame they decided to send their business, and our data, to an offshore multinational rather than help build Australian capability. This is exactly what the Productivity Commission recommends of course.
• Despite current anxieties about AI's impact, history suggests that innovation not only challenges existing employment but also seeds the growth of new sectors and job roles. The real challenge lies in managing the transition, ensuring that the benefits of technological progress are widely distributed and that displaced workers find pathways to new opportunities.
• Sovereign investments in AI are crucial. The White House recently highlighted the U.S. government's initiative to allocate $11 billion towards semiconductor research and development. This announcement was part of a broader strategy under the Chips and Science Act passed by Congress in August 2022, which includes a total of $52.7 billion in funding, with $39 billion dedicated to semiconductor production subsidies and a 25% investment tax credit for constructing chip plants, valued at around $24 billion.
• The easy argument about why every country needs its own sovereign AI is that they need an LLM that embodies their values, language, and facts. This is true, but the real reason is that sovereign AI capability is critical if a nation is to make the transition from exporting goods to exporting higher-value commodities, like information. Uber doesn't export cars, it exports the capacity to connect drivers and passengers (information). This export is entirely virtual, globally scalable, and incredibly valuable. You can't develop this kind of business by downloading AI from overseas, because if you could then someone else would already have done it. The Australian Productivity Commission thinks we should just download AI though, and thus stick to digging stuff up and shipping it overseas.
• The executive in charge of Commonwealth Bank’s sprawling technology operations says early adoption of artificial intelligence has made its expensive software engineers 30 per cent more efficient. The bank also plans to increase customer interactions with AI.
• The Productivity Commission has just released a report saying it thinks that Australia should just use AI from overseas. Australia sits between Uganda and Pakistan at number 93 in the world on the economic complexity index and we're slipping down the rankings every year. AI is one of the best chances we have to increase our economic complexity because it is industry agnostic. The idea that we should just download other countries' AI is how we got here, and is the very embodiment of the problem.
• Good to see that "demand for AI is booming" in the Australian job market, although it's hard to see how we're going to fill the need. What's more important is that the Australian demand is mainly for engineering rather than researchers. It's indicative of the fact that we're largely just making small improvements to existing processes rather than inventing new global business opportunities.
• The pace at which generative models have moved from research idea to essential business capability has been really impressive. It's a pretty good antidote to the 'fast follower' approach whereby some companies and nations have decided to just wait until someone else develops the tech and download it from them. Much of the work that companies need to do to keep up with AI is really mundane and can be achieved by downloading a package. The high-value opportunities, however, arise from a deep understanding of the technology, what it can do and where it is going.
• "...the UK government is spending around twenty times on AI what Australia is, despite the UK economy and population being just three times bigger."