Finding a new career in AI
From bushfires to coronavirus, the ever-growing thirst for answers has led to a continuing global shortage of AI skills. Darren Yates explains how you can skill up.
Back in 2016, IBM came out with the statement that 90% of the world’s data was created in the last two years. If that’s kept up, and there’s no reason to suggest it hasn’t, it’s a simple insight into just how much data we all generate as we do daily life.
But increasingly, researchers are finding golden nuggets of knowledge hidden in that data, whether its cures for diseases or untapped profits for business. What appears to be in short supply, however, are experts with skills in data to find these nuggets. As machine-learning continues its march into new areas, demand for AI professionals is ramping up.
IS AUSTRALIA FALLING BEHIND?
Right now, there’s a global AI arms race being run, but concerns are growing that Australia is falling behind. One reason is a lack of skilled AI professionals. Former Google Australia R&D boss and CEO of Cicada Innovations, Sally-Ann Williams, speaking at the recent ‘Australian Financial Review Future Briefing’ in February, reported that, based on her experience, local companies were likely over-estimating their AI expert head-count (tinyurl.com/tf2t2kd).
Data61, the CSIRO’s data-science arm, produced a report in November 2019 that identified Australia as having 6,600 AI workers, up from 650 in 2014. However, that number will reportedly have to rise to at least 32,000 by 2030, with area expertise required in computer vision, robotics and human language technologies, among others.
Another problem Australia faces is the lack of collaboration between industry and research. According to the 2015 ‘Innovate and Prosper’ report (atn.edu.au/siteassets/publications/ atninnovateprosper.pdf), Australia ranked a miserable 29th out of 30 OECD countries for business university collaboration. What’s more, it appears to still be an issue today. At the same AFR Future Briefing, Australian Institute for Machine Learning director, Professor Anton van den Hengel, countered a common view that universities weren’t interested in collaboration, suggesting local businesses needed to do more reaching-out to universities.
THE GROWING OPPORTUNITIES
It’s a little sad, for there are some seriously high-impact opportunities for AI at the moment. The recent bushfires that tore through large areas of south-eastern Australia left many devastated communities. This and other natural disasters are the focus of the Bushfire and Natural Hazards Co-operative Research Centre (bnhcrc. com.au).
It’s part of a broad ‘co-operative research centre’ (CRC) program funded by the Australian government, under the Department of Industry, Science, Energy and Resources, bringing together governments, research institutions and industry on big-focus projects. Other CRCs include areas such as iMove (transportation), Food Agility (sustainable food) and Digital Health (tinyurl.com/wqp6mbk). However, the one factor many CRCs have in common is the desire to utilise data technology to boost innovation.
POSITIONS VACANT
At time of writing, employment website Seek listed 641 positions around Australia under the title of ‘Data Scientist’, which also encompassed roles such as ‘machine learning engineer’, ‘data engineer’ and ‘applied statistician’. As an exercise, I analysed the first 100 jobs, looking for any mention of education requirements. These were categorised into four groups of increasing relevant degree-education level – none, Bachelors, Masters and PhD. Job listings mentioning more than one degree level were listed under the minimum required unless a preferred option was suggested, for example, ‘Bachelor’s degree (Masters preferred)’ was classed as ‘Masters’.
Of those 100 ads, 43 listed no formal education requirements – but before you celebrate too hard, each role required considerable experience in a range of coding, statistics and business intelligence skills. A further 31 roles
required a Bachelor’s degree in a quantitative area, such as maths, computing etc, 13 positions required a relevant Master’s degree and the remaining 13 required a relevant PhD. The job advertisers ranged from an international fast-food brand to local universities.
WHAT YOU NEED TO LEARN
As those job ads showed, AI is a pretty broad church, with different roles to suit different interests and strengths. However, virtually all of those roles are technical and good coding skills are expected. Just about every ‘top 10 programming languages’ list over the last few years places Python and Java in the top-five (and top-three on many occasions). Of those two, Python is quickly becoming the ‘go-to’ language for machine learning, thanks to its easy-to-understand syntax and ever-growing library of add-ons.
A further analysis of the first 20 of those ‘data scientist’ jobs ads backed this up, with 17 (85%) specifically requiring Python skills, the most common language listed.
Following close behind and receiving specific mention in 16 of those 20 ads was SQL. This might surprise but it shouldn’t – much data exists in databases and clearly, having SQL skills is well-regarded by employers. The third-most common language mentioned was R, a statistical programming language well-known in data-science and machine-learning circles. All three languages were mentioned in 12 of the 20 ads.
However, coding is a foundation skill and even in junior positions, you’ll typically need a good working knowledge of data science principles, such as ‘deep learning’ and predictive modelling methods, such as RandomForest.
The other key to remember is that machine-learning and data science are a means to an end, not usually an end in itself. Unless you’re in specific research, the amount of time you spend dealing with actual machinelearning algorithms could be as little as 10% of the whole project. Other skills, from data collection and engineering to presentation skills, are also highly valued.
WHERE TO LEARN IT
With 43 out of 100 job ads not mentioning tertiary education requirements, there are plenty of roles where experience rates highly. However, that experience is very often based on having high-level foundation skills and knowing how to use them.
As for where to get those skills, there’s no shortage of options – whether you’re after something informal just to dip your toe in the waters, or you’re ready for the rigours of more formal education. For starters, try Kaggle (www.kaggle.com), one of the largest data science communities online. There are lots of good textbooks as well – Data Mining Concepts and Techniques by Han, Kamber and Pei is a good overview.
If you’re ready to get more serious, there are low-cost online courses from Coursera and edX (edx.org/course/ subject/data-science). Many Australian universities now offer formal degree courses in data science, including the University of Adelaide, University of Technology, Sydney (UTS) and Western Sydney University, just for starters.
THE REWARD
Machine-learning and data science are the cutting edge of computer science – and will likely remain so for the foreseeable future. Learning it isn’t an overnight task, but the breadth of areas it can be applied to is mindboggling. If your current career no longer does it for you, the future is looking bright for AI, machine learning and data science.