Ian Oppermann PhD is the Chief Data Scientist for the New South Wales state government in Australia and an Industry Professor at the University of Technology Sydney. Ian chairs the NSW AI Advisory Committee.
At an extended family get together over the Christmas break, someone asked what I do for a living?
The simple answer I gave was that I work with NSW Government in data. The eyes of the person asking the question quickly glazed over and the conversation was about to change direction until I said “and AI … ”.
The expression of the person asking suddenly changed showing real interest. I went on to talk about the Revenue NSW vulnerability prediction tool, eHealth’s Sepsis prediction tool and some highlights of other AI projects within NSW. By this time, everyone was listening, not an easy feat at a family gathering.
The questions started to come thick and fast and increasingly headed in the direction of “how do we stop the robots taking over the world”? This perspective is one which comes up regularly.
My answer was that use of AI is inevitable, and the examples I gave show that it is already happening in government. Also, there is a very important difference between AI that informs, like a personal assistant on your phone or a smart alarm, and AI that does things like an autonomous train or factory robot.
Acknowledging the inevitability of use of AI, what we need to do is move forward with AI in a deliberate and cautious manner, piloting before we scale, and thinking ahead about potential harms from use of AI including those which are not immediately obvious. This was a great way of introducing NSW’s new AI Assurance Framework.
In 2021, NSW set up an AI Advisory Committee to look at government AI projects and give advice about their design and use. Part of the Committee’s work was to develop a formal framework to “assure” use of AI. Over the course of the year, the framework was tested against ongoing or developing AI projects. The framework itself is a series of questions - highlighting the need to identify benefits, risks, potential harms - and a set of resources to help people planning projects to better understand how to address issues of data, bias and governance. The framework connects to international standards where they exist, and best practice where they do not.
After endorsement by the AI Committee, wide consultation within Government, and finally approval by the NSW Cabinet (DaPCo), the AI Assurance Framework will come into force early in 2022.
We now need to build the framework into an actual tool, which interestingly will have AI in it, which means it needs to be reviewed by our AI Committee. Bringing that tool to life and making sure that we have a useful yet generic assurance framework to apply to all AI projects across government will be the challenge.
Back to the family gathering, I managed to get to the end of my story before the conversation turned to other matters. Whilst the next topics included recent Qld floods, dealing with COVID and the imminent start of the school year, I was quietly confident that AI would increasingly play a role in how we address all of these issues, and that in NSW, the Assurance Framework will get a workout in the coming year. Another helping anyone?
More information, check out the AI Assurance Framework.