NSW gov takes cautious approach with generative AI

The NSW government will be taking a "deliberate but cautious" approach when implementing new artificial intelligence technology in line with citizen trust around data and AI usage.

NSW government chief data scientist and industry professor at UTS Dr Ian Oppermann told an Infosys and Trans-Tasman Business Circle event that trust is a “very big issue”, with his work on the NSW AI assurance framework and a multitude of other government policies.

“Ultimately, there are a whole lot of other elements around trust and demonstration of trustworthiness," Oppermann said.

“We need to explore what happens when things go wrong.

"We need to be very clear about what we will not do with data in order to help build confidence, that we're behaving appropriately and demonstrating trustworthiness.

“Even that isn't enough because, of course, we can clearly identify what looks to be okay now or what looks to be really not okay now, but we need to find frameworks to think into that space in the middle."

Oppermann said that appropriate and responsible use of data, digital and AI would remain a live issue.

“We also have ... a watchword which is ‘deliberate but cautious’," he said.

"We are going to use AI; we are going to use data. We will do it deliberately, but we'll be really clear about what and when and how … we'll do it cautiously, but we are going to move forward.”

Oppermann said that artificial intelligence is, at its core a tool that needs to be understood.

“AI is a tool, a tool driven by data and as exciting as all the possible use cases are … we always need to come back to tool driven by data and think about the issues we need to understand," he said.

The surge in popularity of generative AI has the state government "excited about the potential for AI” with a “number of town hall meetings” conducted about how to approach the technology best.

Oppermann said the biggest difference between the latest launch of conversational chatbots such as ChatGPT, “is just the power of them”, going beyond standard digital assistants.

“We did have to remind people that in NSW, we have an AI strategy and AI ethics policy and also an AI assurance framework," Oppermann said.

“What we've been doing is helping people remember what we've already put in place around the less impressive uses of AI, but also give a little bit of guidance as to what we really must do from a data privacy, data security, and cyber security perspective.

“It's just really a matter of re-grounding people and providing a little bit of context for these amazing new tools that are coming out and I imagine we'll have to do that again when the next generation and the next generation of tools come out.

“But the principles and policies we've got work well, and the adaptation to take in some of the new things that we've suddenly seen amplified through these new tools is what we constantly need to be doing.”

Another element Oppermann said the state government was conscious of was the impact of too much personalisation.

"There is a limit to just how much customisation people want from government," he said.

“There are limits around just how personalised the service delivery from government needs to be or is accepted to be”.

As an example, he said while people appreciated data usage for customer identification purposes, it may be considered “too far” when over-personalising a penalty or fine.

“The NSW government is certainly thinking about customer-centricity and even the language of the customer as part of that," Oppermann said.

"There are limits to just how personalised that service might be for different circumstances, but it requires us to think about every single use case differently and rather than do a one-size-fits-all.”