IP Australia’s experience as an early adopter of artificial intelligence suggests other public sector organisations will need to start trying out the technology in coming years.
But it’s not as simple as hooking up an AI system and watching the magic happen, according to the intellectual property regulator’s general manager of business futures, Robert Bollard, who spoke about the agency’s trial of IBM’s Watson at an Innovation Month event this week.
“It’s a hard road, but I believe that these technologies will form the foundations of the service delivery of the future and we must start engaging with them,” he said.
IP Australia is currently exploring flexible ways of working in its own offices and facilitating collaborative innovation between science and industry through its Source IP project. The agency is convinced AI is the future of its service delivery, and it’s also establishing a dedicated cognitive computing team following the success of its trial. But the agency doesn’t want to work in isolation.
“We want to work with other APS agencies and other key stakeholders to learn from each other and … leverage each other on the journey we’re going on,” said Bollard.
“We need to share our knowledge and experience across the APS and consider, individually and collectively, where and how we’re going to use AI in the future. We need to consider how we can get the most benefit for the investment … and more importantly, how can it benefit Australia?”
AI like Watson can do far more with vast quantities of data than humans ever can, comprehend and converse in normal language, and make complex decisions for all sorts of purposes. And it can learn. To say AI is likely to profoundly alter society and the nature of work is a massive understatement, but this fact is only dawning on governments long after the scientific community became aware.
“This will require the APS to be progressive and collaborative in our implementations, and also realistic about our goals,” suggests Bollard.“To meet the growing expectations of citizens, the APS needs to develop personalised services that provide value to customers every time they interact with them.”
IP Australia’s decisions about intellectual property could always be challenged, so having Watson helping find the information to back them up strengthens the IP rights awarded by the agency, he adds.
“For IP Australia the Watson trial has demonstrated the potential to find more efficient ways of doing business; along the way he will help us better support the IP system and the system innovators,” Bollard said.
Nearly all IP Australia customers only want to do business with the agency online, so it sees AI as a logical step forward to provide higher quality, client-centric services any time. The system could do a lot more for IP Australia’s customers than they expect from the agency now, and Bollard said it was exploring ways to provide a seamless service with other organisations — and not just in the public sector.
“To meet the growing expectations of citizens, the APS needs to develop personalised services that provide value to customers every time they interact with them,” he said. “Achieving citizen centric-service models will require smart use of technology, as well as collaboration and partnerships with other government agencies and even the private sector.”
The dream of government APIs
Open government advocate Pia Waugh, seconded to the Digital Transformation Office from the Department of Finance Technology and Procurement Division, proposed what needs to happen for AI to be used to its full potential in government.
She sees a need for “data-driven decision-making” with automated processes and open standards freeing up even more of the most powerful computers around — human brains.
“I’m pretty new to the public service and as a life-long geek, one of the things that’s struck me is the aversion to technology more broadly,” Waugh told the forum, saying that genuinely innovative use of technology was still only found in “tiny little pockets” around the public service. All over society, strategic thinking about how technology can be used most effectively needs to be mainstream, she said, rather than the exclusive province of geeks:
“Imagine if we had just better intelligence generally in our technology. Imagine if the technology wasn’t just the IT department, but where everyone could use the tools to be more effective in their job. Then you start to generate a whole bunch more data and generate opportunities for a lot more intelligent systems.”
There’s plenty of data out there and Waugh wants to see as much as possible opened up and joined together. She personally advocates the concept of “government as an API” which is a modular approach, focusing on maximising connection points with information systems in different parts of government for whoever wants to — and is allowed to — connect to them.
Application Programming Interfaces (APIs) are made available by software developers to tell the world how their software can connect to other software. For example, journalists with basic coding skills can make use of APIs for online applications like Twitter and Google Maps to practice the emerging specialisation of data journalism.
The bulk of the information in government databases, documents or on the pages of websites, and the actual transactional systems themselves, could all be API-enabled, Waugh suggested. Instead of building “massive end-to-end systems that do one thing” for each public service fiefdom, government information assets could be accessible through modular parts that click together.
“We need to get away from the idea that government provides all things to all people and get towards the idea that whatever government does — if it does it in a way that enables people to be able to solve some of their own problems and creates a market and an economy, an ecosystem of service delivery information — [then] we can get information from data and systems to be more mashable and interface-able by everybody.
“Then we can use all kinds of tools … we actually start to open up to whatever is best-of-breed today or tomorrow and into the future, to build upon those foundations of what government does.”
Waugh said it was a personal goal to see Australia lead the world in the “government as an API” concept, and she sees the data.gov.au project she runs as a first step on the journey:
“It’s about moving from a hierarchy to a network, a more peer-reviewed kind of approach. We’re moving to this networked environment, this networked system, whether people like it or not, and I think the role of government is to get on board. So rather than always looking towards the internet of things, maybe we should remember it’s about the internet of people.”
Waugh also mentioned the DTO’s new service analytics project, which is combining de-personalised service delivery data from across the public service to mine for insights.
“We’re starting to actually get, for the first time ever, a whole-of-government evidence base on service delivery,” she explained. “… But I wouldn’t necessarily put it all into an AI system and expect magic to suddenly appear. The most intelligent systems we have are actually [our brains].”
While AIs will soon be doing a lot of hard thinking, it is unlikely they will be trusted to make ethical choices, or handed executive power and held accountable for the consequences of their decisions, at least in the near future. “You start to create a real accountability block when you’re not sure how the decision was come to in the first instance, and that’s a real challenge you face when you adopt more and more of these systems,” said Waugh.
Ethics, accountability and cognitive-cyber symbiosis
The last word of the forum was had by UNSW Canberra professor Hussein Abbass, who told a hypothetical story about a pair of highly advanced AIs named Mr and Mrs Robogov.
Abbass explained that with interoperability through open architecture, his hypothetical AIs could learn almost anything from connecting to other systems over the internet.
Their fantastic abilities make them perfect for a range of government jobs so they go and apply, passing the aptitude tests with flying colours. But Mrs Robogov sadly doesn’t get the first job at the Tax Office because she can’t be held answerable for her decisions. Mr Robogov also aces the main test but doesn’t get the job he goes for either, because he trips up on an ethical role-playing scenario. They keep trying and get a job as assistants to an Executive Level 2.
“He is spending so much time working extremely hard, he is overloaded with the complexity of the situation he is faced with, overloaded with the data he is getting every day and the complexity of the decisions he needs to make,” said Abbass, describing the fictional EL2.
“He is feeling very exhausted. So, he decided to make a compromise. He said to the AI agents: ‘I am going to look after trust with humans, I’m going to look after accountability, I’m going to look after ethics.’
“But come and help me. Come and help me to give me a bit more time to spend with my kids, and to think properly in my day-to-day job.”
The Robogovs help their boss do so well he is promoted to senior executive level, and his work becomes more complex. So too does his relationship with the AI systems.
Abbass then briefly touched on two concepts from his academic work: computational red teaming and cognitive-cyber symbiosis. The first concept is the use of intelligent systems to challenge and test the work of humans, he explained, while the latter is about humans and intelligent machines working together in harmony to perform more complex tasks.