Hi Anton, please tell us about your journey with AI Research and how you came to help found the Australian Institute for Machine Learning as their Founding Director.
I started a Ph.D. in Computer Vision when it was largely about projective geometry. My particular interest was in Statistical methods, particularly parameter estimation. Computer Vision is concerned with trying to make sense of a complex world through images, which are an incredibly complex and powerful way to represent information. Forty percent of the human brain is taken up with visually interpreting, which reflects the value of visual information.
I believe Computer Vision has led the modern AI revolution. Deep Learning had its first real impact in Computer Vision – and now, Deep Learning can outperform humans at simple recognition tasks, which is about as close to AI as we’ve gotten thus far. I’m lucky to have been working in a very interesting field that has made incredible progress over the last 20 years.
I started deliberately growing my research group from the beginning of my time as an academic because I knew that I could tackle more interesting problems with a bigger team. This led to some success and a significant amount of pure, industry and Defence funding – and before long I found myself leading a team of 15 people. At that point, I decided to try to grow the group to become one of the best in the world. It took a while, but we got there. We’re not done yet, of course, we’re still growing – and still doing world-leading research!
With over 130 researchers, the Australian Institute for Machine Learning studies Machine Learning and AI across a wide variety of sectors and applications. In what sectors do you see AI disruption as having the biggest potential impact?
It’s not often the first thing people think of, but Agriculture will be particularly impacted by Machine Learning. We’re doing a lot of work in Agriculture at AIML. Many farmers have an incredible wealth of data available to them – so much so that it can be overwhelming.
We’re working with farmers who have access to everything from satellite images and drone videos. Using AI, we can take that information and make it actionable so farmers can make smarter, more informed decisions. We’re also working on projects involving hardware, like robots that pick tomatoes and pollinate flowers.
The impact on Medicine will be similarly enormous. Humans are extremely complex systems, far more complex than any human can understand. Reductionist science has made incredible progress in understanding the various parts of the human system, but I believe that it won’t be possible to understand its full complexity without using Machine Learning.
We’re already using Machine Learning to find new drugs, fight cancer, and tackle the social causes of diabetes – but this is just the beginning. The next steps will be based on looking at the whole human system rather than breaking it into separate parts. Machine Learning may even enable doctors, and patients, to make informed decisions about the total impact of proposed treatments – and, ML happens to be the only technology that can solve the coming health budget crisis.
What is something you know as a leading researcher on AI and its impact on industries that most people don’t know about?
I don’t think many people realize that we’ve made very little actual progress on AI in the last 20 years. There has been progress, of course, but there hasn’t been a revolution. We’ve made huge progress in Machine Learning and Computer Vision specifically, but we’re really not much closer to real AI than we were 20 years ago.
What we are quite good at now is making agents with specific capabilities. We can make an agent that understands what song you want to play, or that can find the best path across town – but these are really specific capabilities. The agent that navigates you across town can’t play you a song, and the device that interprets your song requests can’t navigate.
Machine Learning has enabled humans to come up with artificial agents that do a small set of very specific things, but it hasn’t made much progress towards achieving the kind of intelligence your pets have. Your pet cat, for example, would be quite capable of surviving by itself if you stopped looking after it. Your cat could find its own food, find shelter, dispose of its own waste, and even procreate – all without your help – in an environment it has never seen before, without any instruction. We’re probably over 20 years from AI being able to do anything similar.
AIML has only been officially established since 2018 but you are already ranked #3 in the world in terms of papers in high-quality computer vision conferences, just behind Carnegie Melon University and the entirety of the Chinese Academy of Sciences. What exactly does that mean, and how did you drive AIML to hit that milestone so quickly?
We’re ranked number 3 globally in terms of the volume of high-quality Computer Vision publications over the last decade. I know it sounds a bit convoluted – but really, it means that in the publishing field, AIML is publishing high-quality work at the same rate as the top research institutions in the world. The Chinese Academy of Sciences is 100-times our size, and yet we only ranked two spaces behind them.
The people behind the Australian Institute for Machine Learning have been working for 20 years to get to this level. The Institute brings together an amazing group of people, many of whom have been working in the field for a very long time. Computer vision has revolutionized what people call AI over the last decade and at that time, we have seen the number of people in the field increase more than ten-fold. The fact that we have been able to maintain, and even improve our international position in the face of this incredible competition is a testament to the amazing researchers on the team.
Can you briefly describe the Art Intelligence program and MURMUR?
The Art Intelligence program is a collaboration between the Australian Institute for Machine Learning and the Sia Furler Institute at the University of Adelaide. The program brings AIML’s leading researchers and engineers together with amazing artists through its artist-in-residence program, starting with Laurie Anderson. MURMUR is a gallery at the intersection of Art and AI. It’s not for art about AI, but for art enabled by AI. It’s an opportunity for artists to say something they couldn’t say beforehand, to express ideas they couldn’t express without the help of Machine Learning.
Enabling artists to exploit and explore Machine Learning as an artistic medium is a big part of what we’re trying to achieve. Almost all art thus far has been fixed in time – an artist makes their art, and it remains static and unresponsive to its audience and its environment. One potential impact of ML on art might be that art could respond to the viewer to offer a personalized, interactive experience.
I think there’s great potential for the art coming out of this program to improve upon existing social commentary on AI. I’ve seen this narrative about a Terminator-like entity coming to take everyone’s jobs, which is not how AI will impact our economy. I’m hoping that by engaging artists in Machine Learning, we might enable a more informed debate about its real impact.
How exactly will the Art Intelligence program be working to further the impact of AI on the Arts, and vice versa?
Art is an essential component of human expression – it’s a primary way that our society communicates with each other to further our mutual understanding. AI ultimately needs to learn from human data, and it’s so important to make sure that AI is being developed in a way that includes a broad range of identities, perspectives, and thought processes. This program will help art, which captures so much of humanity, inform AI to make it more complex – and it will help AI inform art with the potential for groundbreaking new works with unprecedented access to the Australian Institute for Machine Learning’s incredible minds.
The program’s artist-in-residence program has just started, but already Laurie Anderson spent a week in Adelaide working with AIML. This will see a new artwork produced and it already had an impact on the quality of the art being generated by the AIML/Art-community collaboration.
We’re also developing an Art Narrative machine that analyzes photos of an artists’ work and provides an evaluation of the piece. As AI tries to tell us why a piece of art is important, we can get a fresh perspective on how we ascribe meaning to art.
We’ve seen art and design have a major impact on other tech sectors, with graphic design and UI/UX design now being essential practices within the industry. Is the goal of the Art Intelligence program to foster a similar effect of arts on AI?
The ultimate goal of the Art Intelligence program is to explore the intersection of art and AI, wherever it may be. I think it’s fair to say that we might see a similar impact of art on AI that UX design has had on consumer technology.
The relationship between Art and AI hasn’t been explored as much as AI’s potential impact on other practices like Healthcare or Finance, so there’s still so much to explore and discover. I’m really excited to see how the program develops!
What does your sixth sense say about the future of mankind, amidst growing concerns of global warming, and COVID-19?
I think that humans tend to overestimate the short-term impact of these things and underestimate the long-term impact. The fact that we’re still having a debate about global warming, instead of taking radical action, is an indictment on us all. My impression is that things will have to get quite a lot worse before people realize that it’s in their own self-interest to enforce radical action.
There’s a lot to learn from the COVID-19 process. We were warned that it was likely to happen, and we did nothing about it. There are a bunch of other things going wrong, and we’re doing nothing about them either. We seem to be unable to make any significant changes to the way we live except in response to a crisis. We’ll need to do a lot more proactive work if we want to address global warming or other concerns in a meaningful way.
Thank you, Anton! That was fun and hope to see you back on AiThority soon.
Anton van den Hengel is the Director of Applied Science at Amazon, leading a Machine Learning research group, working on Visual Question Answering, Conversational Agents, and general Computer Vision.
He is the founding Director of The Australian Institute for Machine Learning, Australia’s largest machine learning research group, and on many measures, it is most successful. He is also a Chief Investigator of the Australian Centre of Excellence in Robotic Vision, and a Professor of Computer Science at the University of Adelaide. Professor van den Hengel has been a CI on over $60m in research funding from sources including Google, Facebook, Canon, BHP Billiton, and the ARC.
The Australian Institute for Machine Learning is an internationally recognized machine learning research group. It carries out world-class research in machine learning and is both backed and funded by, Facebook, Google, and Microsoft, among others. AIML is setting the pace of machine learning technology development that is fueling the current enthusiasm for artificial intelligence around the world.