How should WA prepare for a productivity revolution driven by data science and machine learning?
Autonomous vehicles will process about one gigabyte of data per second, according to US tech company Intel, roughly equivalent to an hour of standard definition television.
The Square Kilometre Array radio telescope planned for the Murchison region will create many orders of magnitude more information.
That 132,000-antenna operation will produce 157 terabytes of data per second, the CSIRO has estimated.
The numbers give a glimpse into just how big the theme of ‘big data’ can be.
(click here to view a PDF version of thiss 12-page liftout)
As the information accumulates, computing power is also improving to process it all.
Along with the data and processing power comes machine learning, in the field of artificial intelligence, where computer algorithms can be improved automatically by running prodigious iterations.
Business News spoke to a number of organisations to explore how they are using the technology to solve problems.
From detecting heart disease (see page 10) to predictive analytics in resources (page 7), machine learning could unleash a wave of productivity and higher living standards.
It could sharpen-up provision of public services, make local industry more competitive, and earn export revenues.
But with every story of disruption there’s a need to ensure all of society can make the transition.
It’s easy to see why there’s some anxiety on this front.
In 2018, McKinsey Global Institute estimated that up to 375 million workers worldwide may need to change occupations by 2030 because of artificial intelligence.
Meanwhile, a 2015 paper by Australia’s Office of the Chief Economist predicted up to 44 per cent of jobs could have tasks at risk of automation in coming decades.
McKinsey nonetheless believed full employment could be maintained to 2030 in most scenarios, as demand for other jobs grew, including creatives, tech professionals, builders, teachers and care providers.
There’s a further thread to consider: Western Australia will face growing pressure as competing regions embrace technology, because automation used in Pilbara iron ore mines to cut down costs could equally find application in Guinea or Brazil.
James Telders (left), Sherief Khorshid, and Mike Clark are directors at Three Springs Technology. Photo: Gabriel Oliveira
Leading the charge
It’s often said that WA’s Pilbara region has more autonomous vehicles in operation than Silicon Valley, with trucks, trains and drill rigs on iron ore mines and through the supply chain.
While the scale attracts headlines, there are pockets of innovation across the economy.
Big state-owned utilities Water Corporation and Western Power have used machine learning.
Picking out rooftop solar panels from satellite images and predicting potential pole top fires has helped better target investment, maintenance and upgrade strategies for the power network.
Three Springs Technology is a smaller, private consulting business developing its own products.
Co-founder Sherief Khorshid worked in quantitative finance in Hong Kong before returning to Perth in 2016.
His first move was to develop an automated currency trading system, and Three Springs then moved into computer vision applications, including in the medical and mining sectors.
Computer vision trains algorithms to detect content in images, such as a face or an object.
Three Springs consulted to ASX-listed Resonance Health, automating some of its medical products.
Mr Khorshid said Three Springs shared its time between consulting work and developing its own products to scale-up.
One such development was a comment moderation product initially created by Mike Clark, who joined the business in a merger in November 2019.
The software is now being trialled with a major media organisation.
It’s in a branch of machine learning called natural language processing, where computers are trained to understand language.
Another computer vision use case was working with a conveyor belt manufacturer, analysing video footage to detect wear.
It could shave significant time off the maintenance process, Mr Khorshid said, with analysts otherwise needing to scroll through many hours of footage looking for defects and then finding the right place on the belt to make a repair.
“It’s all about saving time and making it easier for them to monitor the conveyor belt,” Mr Khorshid said.
Leederville-based Frontier Automation has used sensor equipment and machine learning to improve vessel safety when docking in ports.
“When [a vessel] is moored up … the technology ensures safe mooring, safe loading, you can have excessive movement even when moored,” chief executive Jochen Franke told Business News.
If a vessel damaged a fender in a port it could cost millions, he said.
Targeting a very different sector of the economy is U Group, using big data and machine learning for retail.
The startup has a panel of consumers who submit receipts into the company’s apps, Liquormate and ReceiptJar.
Founder Tyler Spooner told Business News computer vision could detect items on the receipt, and machine learning would crunch the data for retailers to use.
The technology is intended to help brands build better products for consumers, and can target market research questions to users of a product.
Consumers participate in a rewards program and can keep an accessible record of their receipts electronically.
U Group partnered with data giant Nielsen earlier this year and has worked to onboard some big retail brands.
The startup built its own technology to read receipts and extract data, Mr Spooner said, with about 2 million receipts in the system and more than 250,000 coming in each month.
Mr Spooner said while businesses had huge amounts of potential data, it was the quality that was important.
ASX-listed Imdex has used machine learning for geoscience applications.
Speaking to Business News, chief executive Paul House highlighted the importance of data quality.
That’s because a machine can only learn through the data fed into it.
Some major investments in computing power are in the pipeline in WA to support the huge requirements in data processing.
DUG Technology listed on the ASX this year in a $26 million initial public offering, with the company flagging as much as $100 million of new capacity in computer rooms and machines in the years ahead.
In October, the Pawsey Supercomputing Centre awarded a contract to Hewlett Packard for a new supercomputer, which will have 30 times the processing power of its existing assets.
That $48 million deal will prepare the centre for data from the Square Kilometre Array.
Change prep
In coming weeks, the state government is expected to release the WA Data Science Ecosystem Report, which will guide development of the industry in the state.
That has been developed by the WA Data Science Innovation Hub, based out of Curtin University.
Hub director Liz Dallimore said developing an adequately skilled workforce and sparking collaboration around data science would be two key priorities for the industry.
The themes were echoed among many organisations that spoke to Business News.
Ms Dallimore said WA had a comparative advantage in data analysis, but upskilling workers to understand and derive meaning from data was critical.
Universities were playing a role developing local talent, she said, but businesses had often needed to search other jurisdictions to find suitably skilled employees.
“It’s still difficult to attract good people here,” Ms Dallimore said.
“That’s the perpetual challenge in WA … to get people to realise there’s opportunities.
“It’s about promotion, telling those stories about things happening in WA.
“I was at a conference speaking to a data scientist based in Sydney … he said he had no idea what was happening in WA.”
Competing with east coast businesses for employees was a big challenge, according to a number of industry figures.
“The talent pool is not that deep but there’s a lot of interest in the field,” Three Springs’ Mr Khorshid said.
“It’s difficult to get people who are good at AI and can code.
“To be commercially useful you need to [do both].
“The key is you need to have a supply of graduates that are studying the right things, that can already code really well.”
A spokesperson for health insurer HBF, which is using machine learning to prevent fraud and predict health trends, said the data science capacity in WA was improving but needed nurturing.
Organisations would need to invest in talent, the spokesperson said, which HBF was doing with universities.
“The other key is developing diverse teams of different capabilities to help deliver value … there needs to be a strong understanding of the business, and highly developed visualisation and storytelling skills to translate findings,” she said.
On the collaboration theme, WA Data Science Innovation Hub’s Ms Dallimore said knowledge transfer across industries would be a priority, particularly to smaller businesses.
An example of an organisation hoping to drive that collaboration is Ministry of Data, which has organised hackathons for startups and entrepreneurs to solve problems for state government agencies.
Founder Tim Sondalini said Ministry planned to move beyond hackathons to broker relationships between tech businesses and government agencies.
Mr Sondalini said COVID-19 had put some of WA’s development on ice for the past few months.
“The feeling I’ve got when I talk to people … we’ve gone a little bit on pause,” he said.
“[But] there’s a lot of energy.
“There’s a real energy to start getting back to where we were.
“I’m optimistic, I believe in technology.”
Mr Sondalini highlighted two strengths in WA for data science, including the human capital attracted and built by the resources industry.
“Having the big mining companies in WA means we have brilliant minds, brilliant skills, we can adapt quickly to problems state agencies have,”Mr Sondalini said.
The second was the open-data policy.
“An open-data policy exists in the state government, it’s very good,” he said.
“We should celebrate it more.”
Other jurisdictions are lobbying for an open-data approach, which gives citizens increased access to government information.
Examples of accessible data in WA include aerial imagery, heritage building plans, transport routes, car parks and utilities infrastructure.
Public sector agencies needed to take the lead and articulate problems, Mr Sondalini said, with NSW a good model.
Three Springs’ Mr Khorshid said open-data policies had supported tech development and highlighted medical research as an area with significant potential.
WA had world-leading data sets in medicine, he said, but enabling access was an extensive process.
The other collaboration priority was around business culture.
The Three Springs team and U Group’s Mr Spooner agreed larger businesses needed a cultural shift to become more willing to work with startups.
Bigger companies might be more comfortable using bigger organisations, but that would not necessarily deliver the most value, industry members said.
However, there was a growing number of large businesses embracing startups, they said, a trend it was hoped would accelerate.
Creative destruction?
When asked about the concerns in some segments of society about job losses from automation, many in the industry acknowledged the challenge but were optimistic.
Ministry of Data’s Mr Sondalini said he was optimistic about the potential of technology.
But he acknowledged the need for empathy and support through the transition.
“There’s going to be a bit of friction out there,” Mr Sondalini said.
“You can’t be naive; not everyone is going to learn how to be a Java/Python developer.
“With support, people can find new ways to adapt their skills.
“People tend to be quite resilient in finding new ways to adapt.”
WA Data Science Innovation Hub’s Ms Dallimore said AI could take some jobs and create new ones.
Disruption of jobs was already constant, she said.
“What I say to kids is that, ultimately jobs will be taken, but if you’ve got an ability to use data, think critically, the things a machine can’t do, they’re the aptitudes you want,” she said.
Rachel Cardell-Oliver is shaping the next generation of computer science experts. Photo Gabriel Oliveira
Optimism
Data science’s potential has attracted interest from the generation that will be entering the workforce during a period of accelerating change.
At the University of Western Australia, the Data Science club has grown to have 250 members in just 18 months.
President Claire Heffernan had previously completed degrees in arts and in marketing, and said she started studying data science to upskill, particularly analysing consumer behaviour and target markets.
“It opens a lot of doors … it can be implemented in any industry,” Ms Heffernan told Business News.
“That’s why data science has really blown up in the last 10 years. There’s a lot of data lying around, people may not know how to use it.”
She was confident students from an arts background would have potential in the data science space, because understanding and communicating the data were important skills.
UWA Young Engineers chairperson Josephine Liantono also has an interest in data science, and interned in cognitive consulting with a professional services business.
She, too, was optimistic about the potential impact of machine learning.
“It allows us to do more exciting work,” Ms Liantono said.
“The machines do more menial tasks.
“It frees your time up to make more of an impact in the organisation and with the people around you.”
Ms Liantono highlighted potential positives for workers, particularly around safety and working conditions.
Workers might want to move into something less intensive, she said, rather than operating heavy equipment on site in a high-pressure environment.
“They can control the machines from elsewhere,” Ms Liantono said.
Rachel Cardell-Oliver will play a particularly critical role in setting up WA’s data science and tech future by preparing the workforce.
As UWA head of department of computer science and software engineering, Ms Cardell-Oliver said the university intended to give students an understanding of the fundamentals behind algorithms and programming.
That was more important than specialising in any particular programming language because it taught skills that would be applicable in decades ahead.
“When training computer science graduates, we always have the possibility that the future does not look like the present,” Ms Cardell-Oliver said.
“It’s really important students know [the fundamentals].
“We give them experience in lots of different industries; it’s a very applied field.
“Our graduates from 30 or 40 years ago worked on autonomous rail in the Pilbara.”
Computer science students had to adapt their knowledge as the needs of society changed, she said.
“I’ve learned a new programming language every year or two since I graduated,” Ms Cardell-Oliver said.
She highlighted two skills considered particularly important for tech students: the ability to work with others and ethics.
“[Coding] is a team activity,” Ms Cardell-Oliver said.
“Most systems are too big for one person to understand.
“It’s much more collaborative than people think.
“There’s an image of a young man or woman in their bedroom … but most are teams.”
Team collaboration was part of the curriculum, while student clubs and mentoring programs supported students to develop those skills.
A degree specialising in AI at the university would include a minor in ethics, she said.
“Technology is not value neutral,” Ms Cardell-Oliver told Business News.
“You don’t build things in a box and give up responsibility for them.”
However, she said creators of new tech would need to think through consequences.
“There can be a danger of not doing so, you see that in some of the face recognition stuff,” Ms Cardell-Oliver said.
“It’s got some unacceptable uses.
“Education isn’t just about filling people with facts, it’s also about values.”
