[Originally published on the Australian Government Public Sector Innovation Network under a Creative Commons 3.0 BY AU licence]
As part of Innovation Month 2015 we held an event “The Computer Says Yes: what might be the future of artificial intelligence in the public service?“
We have now published the recordings of the speakers for the event on our YouTube channel, and you can also find an article from The Mandarin about the event here.
The below includes some of my thoughts from the event about the opportunities and challenges offered by artificial intelligence (AI):
- The public service is having to deal with more information (structured and unstructured) with fewer resources
- AI is likely to offer us the support needed to enable more sophisticated decision making based on better insights
- AI may be able to help us counter our biases in decision making. ‘Computational red teaming’, whereby AI challenges human decisions to help ensure that the decisions are based on sufficient evidence or reasoning, could help improve processes
- AI technology and operating environments are still emerging, and it will take time to understand what differing AI platforms offer and what organisational capabilities are required to use AI effectively
- AI, plus big data and experiments, will enable a lot more tailoring and personalisation of government services and information, including personalised medicine
- Algorithms may be able to learn to treat people as individuals – but do it at scale, and thus at low cost and great efficiency
- AI poses challenges as well as opportunities. For instance, accountability becomes more complicated with automation – who has/takes responsibility for decisions and outcomes? How will that fit with our understanding of democratic processes?
- The tools we use shape how we work. If we formalise some of our assumptions and beliefs in the processes integrated into AI, it may reinforce patterns of behaviour that are no longer necessary or appropriate – for instance, that certain things have to be done in certain ways, when new tools offer entirely new ways of doing things. Transformation is about changing the model, not just adding to it with new tools
- Work will need to be done to think about human/machine interaction, so that they can work together and support each other
- IP Australia is one example of a government organisation already experimenting with AI, and there are likely to be many more to come as government services move online, as information processing and analytical demands increase, and as citizens expects services to be tailored to their individual needs.
While it may still often be the case that the computer says no, AI offers a potential future where the computer does indeed say yes. The event highlighted that there’s a lot to consider before then, and that there is some great work being done already.
A big thanks to the event speakers, Robert Bollard, Pia Waugh, Rami Mukhtar and Professor Hussein Abbass for sharing their thoughts on this important topic.