House debates

Wednesday, 30 March 2022


Data Availability and Transparency Bill 2020, Data Availability and Transparency (Consequential Amendments) Bill 2020; Second Reading

5:49 pm

Photo of Ed HusicEd Husic (Chifley, Australian Labor Party, Shadow Minister for Industry and Innovation) Share this | Hansard source

In rising to speak on the Data Availability and Transparency Bill 2020, I wanted to start from a different viewpoint, and that is to reflect on a few events that have occurred in the past 12 months, and it's in a way that you wouldn't necessarily assume would be attached to this bill. In the past 12 months, for the first time, artificial intelligence beat some of the best game players in the world at a video game called Gran Turismo—sport. The first time they attempted this was in July, and the human players won easily. But by October the tables had been turned and increasingly the human players found it very difficult to beat the AI. They noticed more and more that the artificial intelligence found ways to play that people hadn't thought about. This is not the first time this has happened. In fact, AI is starting to challenge the way we approach problems as humans and is starting to challenge the whole notion of reason and logic, the way we apply them, the way we get to answers and the way we find solutions. It's not the first time it's happened in games.

But it was also interesting that when AI was used to consider the development of new antibiotics it also came up with antibiotics that had been discovered in a way that most human researchers wouldn't have necessarily thought would be the way that you would develop that antibiotic and, in addition, simply crunched through the data, way quicker than any humans would have been able to do. That's where artificial intelligence is at, at the moment. But it is not only about the way people think about its ability to crunch data, particularly large datasets. But the way it's doing it, and the solutions it's coming up with—it's doing it in a way that we didn't think about.

There are a number of things that come out of that, most notably—and I will come to this later—the framework we put around the operation of AI. In the case of the antibiotics, there was a specific requirement that whatever the AI came up with could not be toxic for humans. So, frameworks are important. As AI starts to develop, it will increasingly be used—and it is being used. Everyone thinks it's a new development. It's taken about 50 years or so to get to this point. And computing power, neural networks—all the developments that have occurred in the past 10 to 15 years—have meant that AI is getting better and better and better.

Increasingly, governments are going to find that they will rely on artificial intelligence more and more. As Henry Kissinger points out—and over the break I read this book, which he wrote with Eric Schmidt, formerly of Google—as AI gets better, governments increasingly, when they decide to make decisions outside of an AI framework, will be asked—parliamentarians will find themselves being asked—'Well, why did you use AI instead of humans to make those decisions?' That likely is a scenario, plausibly, that will occur.

The thing about AI is that to get better and better and better it needs this thing called data. It needs to be able to possess that in increasing volumes and to be able to crunch it on a continual basis and to be able to apply it in particular ways. So, for us as parliamentarians, it is really important that we engage in this debate and think through not just to the darkest edges of the debate, which the mind will tend to wander to—and I understand that that's the way we will get to that point. But the fact of the matter is that if we don't use this and we decide to shut our minds to the notion of the use of AI and the availability of data to make that happen, then, sure, we can do that, but other countries will be using it, and other countries will apply it in ways to make their economies work better, to make their governments work smarter and to make their communities operate better.

So, that is going to be an important consideration for us—whether or not we do that now or leave it for others to do, while we get left behind. I don't think most Australians would want that. In a country where we have regularly been pointed out as being fast adopters of technology, I think in the broader public mind they get that technology does make our life easier and that they want to use it in ways that improve quality of life.

But I tell you what, it's taken a hell of a long time to get to this bill to look at the frameworks around how that data is used and shared. The Productivity Commission in 2016 brought down its report that talked about how we use data. There's a whole movement around open data and gov data. There have been a lot of people, like Pia Waugh, who've been here in times past when gov hacks have occurred. The movement aims to get access to data and to think about how to do things differently and improve the quality of living, not to just be attracted to the darker edges of this debate but to actually think about, with the data that has been collected, how we can apply it in a way that makes people's lives better. That movement has existed here for quite some time.

The Productivity Commission tried to think, 'How would we use data, and what frameworks could we put in place?' This came out ages ago. The first act that we got out of the coalition was not to think about the open data requirements and the way in which government could use this much more beneficially; their first thought was to come up with a consumer data right, to make a buck out of it, and provide a framework for banking that could take data that was currently being used in finance to then be used in that respect. I don't have a problem with that, but it's interesting that that was the priority of this government. I'll come to the CDR thing in the moment.

That happened in 2016. The first thing they did was CDR. The coalition promised before the last election that they would have this framework in place. We then got this bill in December 2020. And then the only point at which this gets debated is now.

If we pass this today, which I imagine we will—and I want to commend the member for Maribyrnong because he has applied deep thought to this. He has worked constructively, he has consulted, and he has secured improvements to this bill. But all that effort that the member for Maribyrnong undertook has gone to waste because this government has failed to get this bill debated in this House and then considered by the Senate. Guess what? The Senate, tomorrow: estimates. It's out. So, by the time this bill gets considered, it is highly unlikely it will pass. It will be two terms since we've had some sort of framework in place to manage this, which is just crazy, because this is stuff we do need to do. We do need to think, as the member for Maribyrnong has been pointing out, about those ethical considerations about how data's used. As much as we need to ensure that that data's available, that it is crunched through AI and that it is applied in beneficial ways, we do need to—as he has rightly pointed out—not think in a completely utopic way but just get the balance right. I think the best thing in this debate is not to be completely in a utopian frame of mind or a dystopian frame of mind but to just get the balance right.

Ellen Broad, in a really good book—and I encourage people to read this book—Made by Humans, makes the point that AI is not some magical thing that just applies on its own. It is shaped by people, and the decisions on its application are shaped by people. She also points out government's tendency. You can't always assume that government will use technology in a way that is always beneficial, because more often than not it will use technology in a way that sometimes runs counter to people's interests. We should always bear that in mind as well. That's why I think the member for Maribyrnong's interventions are important to inject that type of thinking. We do need to make sure that we don't have other robodebts, but we also need to make sure that, if we can speed up decision-making in automated ways and we've got the frameworks right, we do so and we have that data available.

For the people who suggest that this bill will open up all sorts of terrible consequences in the way that things will be used—I haven't heard a lot of those voices. I have spoken in times past about, for example, the way in which we set up the metadata framework many years ago and the way that was pushed through. I have been on this case about the way that data has been used, accessed or applied for quite some time. So it is interesting to see. We heard from one of the speakers just a few moments ago, who I don't think ever spoke up before. Maybe I'm wrong; I'm happy to stand corrected. They never spoke up on the way that metadata was used or the countless pieces of national security legislation that come through and access citizens' data all the time. I never heard those protections being championed by those opposite at all. Let's not forget the irony of some of those on the far Right who rail against data and the way it's being used while they mine data themselves for the sake of building their own political careers, movements or whatever.

I come back to the point that this government talks a big game on data, talks a big game on digital. I note the AIIA had recommended that the government set up AI ethics boards to create the guidelines and best practice for the use of artificial intelligence in government, to create ethical guidelines for the development and use of AI, to ensure that members across the APS and members from industry are involved in that setting up of standards and practice, and that an AI register of all current and future solutions that use AI or machine learning be established to allow for citizen transparency. These are all good things. The industry is thinking about it; the government is slow to act on it. I commend the AIIA for putting that stuff on the public record to get governments to think about the use of data because it will be important longer term.

For the government, it is always a shiny new bauble when it comes to them. It is the latest thing that they might get a great media headline out of and make themselves look like they are future-leaning but they never seem to deliver. It seems to be a common thread with this government, with the coalition to be always there for the announcement but never there for the delivery.

I referenced earlier the Consumer Data Right. I have had a number of complaints from industry about the way the consumer Data Right has been set up for finance—bear in mind it will go across to energy and telecoms at some point—and about the mismanagement of this process. I had one person in particular point out that the Consumer Data Right is progressing at lightning speed after a number of years of instability and iterative development. A person said to me that there is a sense amongst the market that the first sector has not been finished and now attention and resources have been redeployed to other areas. They emphasised banking is not finished yet. There has been no straight-through consent mechanism, no consumer dashboard, no monitoring of the health of the API ecosystem. The role of intermediaries, an afterthought, has not been tested or finalised. there is no unique identifier between different sectors to allow for cross-sectoral data sharing—that is, banking has a customer ID while energy has a premises identifier. There is literally no way for one sector to register an amendment of withdrawal of consent with another sector, which is a big issue, a massive issue. That is about making sure that consent is allowed and, when it is not, how things are managed post that.

The refusal to consider business data and personal data differently from a risk standing means 2.4 million SMEs will have their operations disrupted through zero MYOB and be no longer able to share data with apps that exist on their marketplaces. This point, which I think is relevant to this debate, is that the Data Availability and Transparency Bill and the CDR are literally incompatible in their current formats. There is no conceivable way that public data sets can mingle with CDR data which, to this person's view, is a tragedy due to the potential for used cases and global best practice. These rules are being developed as a priority over standards which is backwards. The most damning thing is that, weirdly, the Treasury is leading this, not the technical people nor the ministerial office—lots of duck shoving going on! What a surprise.

We on this side suggested that those opposite be careful and think this stuff through in the last term. We wanted to have the Senate involved through that whole process as the house of review but we had the Treasurer working with Senator Jane Hume to shut down any work on this and shut down the thinking around it. We need data availability to make sure that it is applied in a way that will work for the people. We need to ensure there is confidence in the way that occurs. What we don't need is a situation where the government is only out there for the sake of the headline, only there for the sake of the announcement, trying to gather a great photo opportunity and then not following through. If CDR is any example or sends us any signal about how this will operate under a coalition then we should be very concerned. It is important to get the frameworks right, as has been attempted by this bill. I wonder whether or not it will get through. But getting the final landing, making sure this works is even more important because, if we lose the confidence of people, we will ensure that we can't get our act together at a broader level.


No comments