From being able to snap the perfect selfie to self-driving cars, artificial intelligence (AI) has entered our lives in many ways and it is now being applied to solve complex problems in most scientific fields.
What is artificial intelligence?
The field of AI refers to developing computer systems that use big sets of data to perform “intelligent” tasks: think visual perception, understanding natural language, reasoning and decision making. Machine learning is one way of building such systems based on providing the computer with examples of what it should do, and let it figure out (learn) how to do it.
AI entered the scene of computer science in the early 1950s when computers started learning checkers strategies and speaking English.
Since then the field has made remarkable progress. With faster and more powerful computers, access to larger amounts of data and algorithm improvements, AI systems have taken off.
Below are some exciting examples of how Australian researchers and engineers are using machine learning methods in areas like observation and protection of endangered species, medical diagnosis and enhancement of the healthcare system.
How AI is improving medical diagnoses
In diseases like cancer, early and quick diagnosis is the most important step towards a positive outcome for patients.
One common diagnostic method is magnetic resonance imaging (MRI). Specialised medical doctors detect cancer using MRI, however, there is a lack of this expertise in Australia – particularly in most remote areas of the country. This can delay the diagnosis of patients for whom time is crucial.
With a degree in Electrical Engineering, a PhD in Biomedical Engineering and MRI expertise, Dr Elliot Smith has had the intuition to combine machine learning and medical imaging for a faster and accurate diagnosis of cancer.
TAKE OUR QUIZ: What’s your perfect Artificial Intelligence job?
In 2016 he founded Maxwell Plus, a Brisbane-based AI startup. Elliot and his team of young engineers are training machine learning systems to look through medical images, genetic data and blood data and find the link between these data and the appearance of the disease.
“Today we have trained our system on over a million medical images to be able to find signs of prostate, breast and lung cancer at levels approaching human performance. This has been a big achievement for us but more than that, seeing how clinicians embrace this new technology is a great sign that the future is one filled with AI supported doctors.” Elliot says.
How AI is enhancing healthcare
In the effort to enhance the Australian healthcare system and patient outcomes, researchers are looking for methods to predict mortality, and reduce readmission and length of stay of patients in hospitals.
This involves handling huge amount of data such as patients symptoms, treatments, past history, age, etc., collected from hospitals and medical practices.
Dr Oscar Perez Concha, a research fellow from the Centre for Big Data Research in Health, UNSW, is using machine learning to extract the most important bits of information from big sets of data.
He is also training machine learning on these datasets to be able to make predictions about the possible outcome for a patient who gets admitted into the hospital with a certain condition, personal and medical history. This will greatly help doctors in making choices in shorter time about the best course of action for every patient.
“I don’t think AI is the future, I think it’s happening now,” says Oscar.
How AI is saving the dugongs
Marine biologist Dr Amanda Hodgson from Murdoch University has spent the past 20 years studying Australian dugong populations.
These gentle sea cows are spread over areas that extend for several hundred kilometres and spend most of their time grazing underwater. Finding out how many they are and where they are is a difficult task.
Traditionally researchers would use small aircrafts, fly over extended areas of ocean to spot the dugongs and record the information. This was time consuming, expensive and even dangerous. In 2010, Amanda and her colleagues started to use drones to capture overlapping photos along the survey area.
This was a cheaper, safer and more accurate method and allowed to explore areas inaccessible by planes. “However, one survey resulted in tens of thousands of images, which then took hundreds of hours to manually review to find the dugongs. So my vision was to automate the image processing – to save time, money, and, quite frankly, our sanity.” says Amanda.
In 2014, she began to collaborate with Dr Frederic Maire from the Queensland University of Technology to use machine learning for image analysis. They have now processed over 37,000 images via a machine learning system that has found over 70% of dugongs in the images.
The machine learning system took 18 hours to complete the task while researchers had taken 377 hours to trawl through all the images.
What is the future of AI?
No matter of what science path we take, artificial intelligence is “an amazing new tool to have in our toolkit. It’s going to allow us to do things like process huge amounts of data quickly and get the answers to our research questions much faster,” says Amanda.
“Machine learning will provide tools to help rapidly accelerate new discoveries across many areas of research,” says Elliot, “Because we can scale computation power much faster than human brain power, this allows discoveries to happen at an increasing pace.”
– Manuela Callari
Author: STEM Contributor
This article was written by a STEM Contributor for Careers with STEM. To learn more, please visit our contact page.