The Hon Andrew Leigh MP
Assistant Minister for Productivity, Competition, Charities and Treasury
E&OE TRANSCRIPT
Q&A
NATIONAL PRESS CLUB OF AUSTRALIA
WEDNESDAY, 7 OCTOBER 2026
SUBJECTS: Artificial intelligence; AI ethics and human well-being; Market concentration and data centres; Social media minimum age and Digital Duty of Care; AI companions; Workers’ voice in AI adoption; Catastrophic AI risks; Consumer protection and AI shopping agents; AI-generated pornography; International AI competition
CORRIE MCLEOD, NATIONAL PRESS CLUB MODERATOR: Hello and welcome to the National Press Club, and today's Westpac address. My name is Corrie McLeod, and I'm the publisher of InnovationAus.com and a director of the Club here.
We have two speakers today for an artificial intelligence and technology ethics forum. Father Paolo Benanti is an Italian Franciscan Friar, member of the Third Order Regular of St Francis, Catholic priest and academic specialising in the ethical implications of emerging technologies.
Father Benanti is Professor at the Pontifical Gregorian University and holds professorial roles at Seattle's University College of Science and Engineering and the Luiss University in Rome. He is a member of the UN's high level Advisory Board on artificial intelligence, and he is also an AI adviser to the Vatican. He's in Australia as a guest of the Italian Embassy.
Andrew Leigh is the Assistant Minister for Productivity, Competition, Charities and Treasury, and prior to being elected in 2010 Andrew was a Professor of Economics at the Australian National University. He is a prolific author with many books examining the intersection of economics, equity and innovation.
His speech this year titled ‘Rethinking Economic Foundations in an AI World’ looks at the shifts in labour, knowledge and accountability.
In May this year, Pope Leo the 14th signed his first encyclical, Magnifica Humanitas, which addresses artificial intelligence and protection of human dignity. This document has captured the attention of business, technology and government leaders around the world, as we grapple with the speed and power of AI to transform society for better or worse.
It was signed exactly 135 years ago to the day after Leo 13th signed Rerum Novarum in 1891, a text that examined the dynamics of labour, technology and capital at a time of the Second Industrial Revolution.
Of course in between these two dates the Catholic Church had its own reckoning with moral authority. We've now acknowledged systemic failures in its role with response to institutional child sexual abuse and forced adoption.
These are fast moving and weighty discussions, and we're delighted to have both of our speakers join us to explore them.
Finally, a podcast of today's address will be available on the Club's website later today.
I'm going to invite Father Benanti to come up and make a short address first, but before I do, please welcome both of our speakers today.
FATHER PAOLO BENANTI: Thank you, I'm really happy to be here and to be invited by the Italian Embassy and from National Press Club of Australia.
The first challenge for myself as an Italian is to be short, but this is another topic.
If you would like to talk about why AI is a topic, and why we have to address such kind of topic in a plural different multi stakeholder way, I think that the best short introduction to this is two example that we can simply highlight in this short periods.
Example number one comes from July – this July, 2026. You know in Hugging Face, Hugging Face is a company that we can understand, just to be short, as the Wikipedia of artificial intelligence, it's a collection of all the different open weight models in AI.
Well, in July the technicians of Hugging Face simply notice that 17,000 users was acting in a way that is clearly an attack to the society. Well, probably you know that neither a North Korean Army can handle 17,000 humans without anyone has some kind of needs or disalignment.
What does it mean? Probably the worst nightmare that we was discussing in technology, a machine with multiple agent is attacking someone else.
Well, the Italian press start to highlight a leader when OpenAI said, ‘I'm sorry, I lose control of my system’, things like, ‘Terminator is here. AI will destroy us’, but this is not the most important part of the story.
You know, you're Hugging Face, you're a cutting edge provider on AI, you are under attack, you take the cutting edge model and say, ‘Analyse it and help me to defend myself’. They take the most powerful Claude model and the Claude model tell to them, ‘Look, I'm an ethical model, I cannot help you, because if I did – if I tell you what's going on, I teach to you how to do it’.
Well, they try with the other provider, same things. They are to build their own model, one open weight model from China without any kind of guardrail to analyse what's going on and to fix it.
So, in a plural society in which we are one provider of AI, in the name of ethics we can avoid a legitimate user to use it. So imagine that tomorrow the national health care system will have a new pandemic, and you isolate the virus and ask to your model that now is included in your national health care system to help you to defeat the pandemic, and could be the answer, ‘No, I'm ethical, I will teach you how a virus works’, we have a problem of power, we have a problem who decide what is legitimate and not legitimate in a such way.
The second example, and then my time will be over, it's come from one year before August 2025. Okay, the topic is divisive, but digital use is not on the topic. I am talking about a judge of the international penal court of liar, and this judge has received an accusation, and this is the divisive topics, against Netanyahu and crime of wars. Well, it's not important what happened with that, because you're a judge, you receive an accusation, your duty is to investigate.
Well, because the judge refused to dismiss the case, Donald Trump issues an Executive Order that put the judge in the terrorist list. Next day he cannot use anymore the credit card, because no US company can give any more services to him, and so he was without any credit card. He cannot use email, 98 per cent of email provider in Europe are given by United States.
Well, he tried to connect online to the international penal court of liar but they ran on a cloud service owned by United States. It's, sorry, I cannot resist in saying that, it's like Spinoza that was excommunicated in the 16th century, no one can talk with them, no one can have business with them.
Once again, it's a matter of technology, it's a matter of power in the public space. You know Montesquieu told to us that only if the true power remain divided in the public space liberty can survive.
Now that we are surrogating with digital and the last frontier of digital is AI, what we use to shape as a rule of rights what is going on, and who has the power to set the rule of right back in the public space.
The wonderful things of my job, I'm an [indistinct] is that I am take the luxury to make question and finish my time before giving any answer.
ASSISTANT MINISTER ANDREW LEIGH: I acknowledge the Ngunnawal people, on whose lands we meet. My thanks to the Italian Embassy and the National Press Club for organising and hosting this event. It is a pleasure to share the stage with Father Paolo Benanti and our moderator Corrie McLeod.
Today’s discussion matches the arc of how many Australians have been thinking about AI. After ChatGPT launched in 2022, most of us were focused on how it could make us more productive – summarising documents, checking our work or planning holidays.
Since then, progress has been rapid. One measure of AI agents’ ability to carry out complex tasks has doubled roughly every 7 months. As the technology has become more capable, the talk has turned to its impact on humanity.
In these short opening remarks, I want to ask 2 questions. What can AI do for human wellbeing? And what values should guide our AI systems?
Start with a man with ALS who had largely lost the ability to speak.
In 2025, researchers implanted 256 electrodes in the part of his brain involved in speech. AI decoded his neural signals and turned them into speech almost instantly.
The synthetic voice resembled his own voice from before ALS. He could change his intonation, emphasise particular words and even sing short melodies.
Now let’s move to a more vexed use of AI.
Researchers asked 105 cancer patients to compare 2 sets of empathetic statements. One set were written by palliative care doctors. The other set were written by AI. Not knowing which were which, patients preferred the AI generated statements 72 per cent of the time.
A machine that had never held the hand of a dying cancer patient produced words that provided more comfort than professionals who had.
The potential of AI to improve human flourishing is substantial. AI systems can help clinicians read scans and treat disease. It can give people with disability new capabilities and make dangerous work safer. It can democratise access to expertise. It has the potential to raise living standards, increase life expectancy and create societies that take better care of the most disadvantaged.
There are plenty of dangers too. AI can produce persuasive falsehoods and supercharge scams. Some of the architects of the world’s most advanced AI models now say that at the extreme, rogue superintelligence could end humanity. Australia’s AI Safety Institute is actively engaged on all these challenges.
There is also a question about wellbeing: what happens when AI starts substituting for other people?
Across the world, people are turning to AI companions. To see why, we need only look at the evidence from recent randomised experiments. One experiment found that a 15 minute conversation with an AI companion reduced loneliness about as much, immediately afterwards, as a conversation with another person. People underestimated how much it would help.
As a Gen Xer, I can’t help thinking that the salve of an AI chat may be fleeting – that building relationships with imperfect people is better than turning to digital devices.
Yet a sizeable share of young people see it differently. In an Australian YouGov survey this year, 8 per cent of Gen Z Australians said they had felt like they were in a romantic relationship with an AI chatbot. And 23 per cent said they would sometimes prefer staying home and talking to a chatbot to going out with friends.
What does it say about Australian society that nearly a quarter of young people sometimes prefer to hang out with an AI, and nearly one tenth have opted for an AI lover?
Part of the solution lies in building community: supporting neighbourhood groups, community sporting clubs, community foundations and local charities. We need to strengthen these offline options if they are to compete with the addictive temptations of AI driven social media, chatbots and gaming.
Which brings me to the second question: what values should guide AI itself?
AI designers often say that their systems are ‘grown’ rather than ‘built’, which means that it is essential that the right values are embedded in their design.
Anthropic literally publishes a Constitution for Claude. OpenAI publishes a Model Spec. These documents talk about principles such as safety, honesty, human autonomy and freedom. When goals conflict, somebody has to decide what the system should prioritise.
There is no value free objective function for a machine asked to help us live.
Magnifica Humanitas talks about dignity, freedom and truth, but also about love. Drawing on Augustine, Pope Leo argues that societies are ultimately shaped by what they love: whether we orient ourselves chiefly towards ourselves, or towards one another and the common good. In the digital age, he calls for a ‘civilization of love’.
Talking about love makes policymakers shift uneasily in our seats. But in this case, it’s the right language to use, because it helps focus our attention on the kinds of relationships the technology will encourage.
The Matrix shows a dystopia in which AI systems maximise attention and dependence. By contrast, an AI utopia would require systems that enlarge human agency, foster genuine connection with other people and encourage us to think deeply about the big questions of religion and philosophy.
We don’t have much time. If the current pace of improvement continues, the AI systems of 2030 will be 100 times as capable as the systems of today. The better they get, the more our choices matter.
The best AI systems can give an ALS patient back his voice.
The challenge is ensuring the rest of us don’t lose ours.
Thanks. Looking forward to the conversation.
CORRIE MCLEOD: Thank you to you both. Father Benanti, you're at the Press Club so there will be questions and hopefully answers as well. This is very fast moving and interesting times that we're in.
I'm going to take a few lines from the encyclical, and one of them is about the notion of shared responsibility, and when you spoke in your opening remarks you were talking about the concentration of power and technology and power and the intended or unintended consequences that can flow from that.
Given how quickly things have moved even in recent weeks, we've had our Prime Minister at the UN, there's been discussions with OpenAI, OpenAI has a representative in Australia at the moment. How are you feeling that that conversation is shaking out when we're talking about power, government, technology leaders? We're at an interesting moment in time. What are your observations?
FATHER PAOLO BENANTI: Well, you know, the problem is that for the first time with the giant concentration of capital, computational power and services connected to the what we call AI, and don't forget that when we say AI, we say a multi layer stack of technology that start with energy, water and telecommunication, or there is the silicon, or there are the models and there are the application and the data in the middle here. Well, such kind of concentration to few hands.
For the first time in the history allow such number of few player to consider also the State as a tenant of the cloud, and you can decide what the State can do or not do. You can have the power to shut down a Public Service, and we are back to the core question that is in the modernity shape our community. No one should be above the law.
And the problem that there is such kind of concentration that rely in really few country once again highlight the question how to let the Rule of Law affect it in a territorial sovereignty community with global digital services, because the problem is this, that computational power can simple make sovereignty detached from territoriality, and how to handle it. If I can let something happen in Australia with my digital services, without the Australian Government never notice that, or never can control it.
So this is a challenge and the challenge requires a sort of domestication of such kind of new form of power in the legitimate form of power that we accepted at least from the modernity nowadays.
This is a question of governance. And the problem of governance now is the crisis of multilateral institution, no-one is betting anymore a lot on United Nations and other international multi level institution. For a lot of crisis we see what's going on with, you know, war crisis and international tension, and the economical tension.
Australia has a key role with rare land, with minerals, but Australia need to buy services from digital player too, like Italy, like other country, like a lot of Western, and so the tension is here, how to handle it. And for the first time we are pushing a lot of services for the promises that the service are showing to us, but we still not have the right instrument to govern it.
CORRIE MCLEOD: Andrew Leigh, you're a member of the Australian Government. These conversations are probably very real right now. Discussions with the Australian Government at the moment, it seems given that Australia's a very attractive place, particularly from a data centre perspective, and there are trade offs being discussed every day. Are you comfortable that Australia is taking the moment to be able to really look at what the exchange of value looks like and putting in place some of the discussions around governance that Father Benanti spoke to? Like we have a moment, are we taking it effectively? Is there anything you could share with us on that?
ANDREW LEIGH: We certainly are. The Prime Minister's been very forward leaning on this. The social media minimum age, we're the first in the world to enact that, and the Digital Duty of Care comes off the back of that success, and will cover artificial intelligence.
We're also setting very clear standards for data centres, making clear that they need to bring their own power, that they need to have a positive impact on communities and they need to have a minimal impact on water, because Australians are rightly concerned about the impact of data centres in their communities and want to make sure that we're getting a fair deal there.
We're a very attractive location for investment in data centres, not only because we're seismically stable, but also the geopolitics and our quite strong uptake of AI among individuals and businesses.
But Father Benanti is certainly right to talk about the challenges of market concentration. The chips are costly, the data sets are huge, the talent is scarce. All of those factors have driven towards significant concentration in the frontier models, and we could easily see that concentration get more acute.
It might be like what happened with internet search, where in the 1990s you had quite a proliferation of search engines, now we effectively just have one. So the potential for even greater concentration of power and of control of AI is there, and we're working very actively to ensure Australians prosper in that challenging environment.
CORRIE MCLEOD: I want to ask a question about language, you referred to about language like of love and looking at the encyclical and looking at humanity. So I'm really keen to hear from both of you, when we talk about the human element here, but then we talk about the language that a lot of the tech companies are using around alignment or misalignment or containment, they don't seem like very human words to be talking about, you know, a technology that is transforming our society and the way that we live?
ANDREW LEIGH: I think you need to not only look at what the tech company people are saying, but also what they're doing. A lot of reports suggest that tech CEOs are quite reluctant about having their children addicted to devices and social media, that they would like to see their kids out in person engaging with others rather than doomscrolling, and that is I think a really important value to be placing around this, that in a world in which AI is turbocharging the addictive properties of social media, we need to be really careful that we don't devolve into that.
And that's why I talked about the examples of nearly a quarter of young Australians preferring to stay home with an AI chatbot than go out, nearly a tenth having a romantic relationship with an AI chatbot. That fundamentally doesn't seem to me like a good world that we're moving towards.
I think it's important to have these conversations publicly and to recognise that the value of in person interactions has always been high, but right now we need to be supporting charities, sporting groups, not for profits, the in person religious gatherings, trade unions, anything that brings people together face to face to do the difficult but messy work of keeping the democracy strong.
CORRIE MCLEOD: Language, Father Benanti?
FATHER PAOLO BENANTI: Well, you know, the point is this: we are talking about a technology or a section of technology, because AI is not only large language model but focus on this, that is a deeply interface to the most core human allied technology that we develop to have a relationship among us, that is the language, you know. And now for the first time is a machine that is producing language.
The most interesting things here is that we are a unique species.You know, we are not the fastest, not the strongest, we don't swim the best. I know that in Australia there is a lot of dangerous wilderness, and so you know very well that.
But, and I know that from Australia, we was the only species that produced the extinguish of other species. 40,000 years ago we arrive in Australia and we know because we caused the extinction of a lot of big plans and big things.
How we do it? Because we start to co-operate, and why we co-operate? Well, because when I interact with another member of my species, well, the other one producing me something that the scientist call ulterior mind. So I suppose that he or her, like myself, he or she like myself as a mind, as a wheel, as a sentiment and things like that, and these allow us to co-operate.
The last frontier model, producing ourself, ulterior mind. So we think that we face someone and not something. And this is not by accident, it's by design. Don't forget that this is not like the vaccine that can – or the last scientific discovery that can happen in a lab. It's an industrial product designed to maximise the money that the product can do in a relationship with human.
So is it the first case probably in the industrial history in which you are not talking about how to regulate such kind of product, you know. When I was the kids, the most desired product that we have was the motorcycle, but I can drive it only with a helmet, with a licence and things like that. Why we are talking about ChatGPT or Claude like for everyone product? Why we don't think that we need the licence so someone can be sure that like when I drive a car I know very well that there is a pedal down there that if I push it brakes.
So we are not giving any kind of social constraint to this. It's not the 01 technology, it's a technology that has to be integrated, and such power of language, the idea that we are addressing computer part as a memory and things like that with human things, is something that tell us that probably, and I'm on the Magnifica Humanitas site, sorry for that, we have a problem answering to the question who we are in this society. But we are a responsibility and the responsibility is to take care of the human being of our society.
So this idea of regulation I look with really positive accent what you do with social network, you know. We need to remember that if we can control technology, because they hide behind the patent, we need to control the human beings that is using technology.
CORRIE MCLEOD: We'll ask one more question before we go to our media members today. So the encyclical mentions digital sobriety, so I guess that's at an individual level taking some responsibility for consumption, and you know, the role that an individual plays in the world, and then we're talking about kind of layers of government, and you mentioned the UN earlier. So in an ideal world what are all the layers we need to be thinking about to be able to properly make sure that we are allowing AI to be built that serves humanity in the way that we would like to see it?
FATHER PAOLO BENANTI: Well, the first level is what we define ethics, you know. When we talk about ethics we are different ethical element. Everyone can think that ethics mean value.
Now let me tell you a secret. I'm a Catholic, and you know, Catholic appoint a really high value on marriage. I'm Catholic, I'm not married. Can we have a value that fit alls? No. So stop to define ethical alignment as a matter of value. No way should we use norm. Well, you know norm is something that is used to protect values. No way.
There is a third element that is principle. So supposing that we have an autonomous driving car, and you don't know if you are a passenger of the car, someone that work on the street or the driver of another car. In an unavoidable situation of an accident, make it sense that the car try to minimise the damages? Yes. Welcome to principle.
So we have to negotiate principle in a transparent way, because if I sell my autonomous car that will try to protect the owners and telling that the life of the owners has more value than the life of the citizens on the street. Well, ethics mean transparency on the principle that act behind, and that has to be negotiated in a higher level and communicated to the lower level, and you know, that becomes something that also is not only constrained the market, but amplify the market, because we call that standards.
If I buy steel from Australia, I know the standards of steel, and so the market for Australian steels is bigger, it's not constrained. So it's a matter of narrative. Yes, once again it's a matter of language and narrative. We have to bring back a positive narrative on standards based on principle that allow the market to grow and people to be safe.
ANDREW LEIGH: Two points on that. In terms of the values that the models have embedded in them, I quite like Stuart Russell's notion that models should be altruistic, observant and humble: able to continually learn and to improve their ethical views as society becomes more advanced.
The models have constitutions, they're growing to huge documents right now as they try to cover much more. And I think one of the challenges for that model ethics will be paring it down into something which is tractable.
But also as AI rolls out across workplaces it's important to remember the role of letting people in those workplaces have a say about how AI is used.
Magnifica Humanitas, as with its predecessor encyclical on the Industrial Revolution talks quite explicitly about the role of unions and having a worker voice.
Now of course, as a representative of the Australian Labor Party, I like that language, but I also think it's the right language if all you care about is having an effective roll out of this technology in a way in which you build broad community acceptance and people feel as though their jobs are being augmented rather than automated away. Having a worker voice at the table is going to be critical in how AI is used in the workplace, which is one of the key friction points right now.
CORRIE MCLEOD: First question; James Riley from InnovationAus.com.
JAMES RILEY, INNOVATIONAUS.COM: Hello. Thank both for your presentations, and a very interesting and quite intellectual discussion so far.
I guess my question is, or my overall question is are we doomed? And are we coming this a little bit too late to be regulating in any kind of meaningful way? Just going to Andrew Leigh's statement that these AI's are doubling in their capability and power every seven months. We've seen a rise of authoritarianism and mass surveillance, we've seen a rise of technocrats and techno oligarchs, I guess, concentration of power, growing distrust in institutions, a fracturing of the global order, an undermining of multilateral organisations, and a diminishing of the nation state.
I guess to Father Benanti I would ask, against that backdrop, in terms of religion and spirituality and the call of the bigger, you know, institutions of religion; have you seen people start to return to that as AIs have become more powerful? And to Andrew Leigh, I guess others have had a guess at what the percentage chance of humanity ending will be. I wonder if you'll have a crack at that, and whether regulating data centres is kind of, you know, pretty small beer now?
FATHER PAOLO BENANTI: Well, you know, no one still ask me if we use AI to capture people in the church! But you know, no the problem is much more deeper than that, if I can. Because you know, we build our society on a set of belief that we are used to validate in some way.
So let me make an example. If I ask to you, do you think that the reality is made by atoms? Probably you tell me yes, and I hope that you tell me yes, but you never see an atoms. You have a system that allow you to believe as a reliable source to a textbook. Okay, I make one times the school of journalist and one of my master tell me don't believe that your mother love you if you don't have a second independent sources, but this is another things.
But, you know, this is how we shape scientific revolution. We cannot test everything, we start to define reliable knowledge. And the reliable knowledge has an architectural displacement, the library, medicine with medicine, biology with biology, history with history.
Now the machine throw out in an irregular way no source of piece of knowledge and we have to believe it in a different way. All the crisis that you listed is a crisis of trusted source of information. This is the problem. And we are changing that. It changed before when there was a changing of age, when we move from Renaissance to scientific revolution we make this change.
So probably what we are facing now is the sign not of a technological crisis but a huge ethical change in which what was the traditional belief are start to be in some way changed, and because we are not so naive we know that power try to handle it, and so if AI and AI narrative will become the new source of power for trusting and credibility, well, someone will use it. And this is why it's urgent to regulate the public space, because you have the most eloquent system in the world that could be at the service of not the best people in the world.
What I trust is that human being nature and democracy, so we are to bet again in democracy, and this idea that this time and space of confrontation can produce in every one of us this desire from what is good for society.
Is this connected to fate? Yes. We call it creature and creator but I don't want to make a lecture on theology here, so it's simple, something that the fate is allowed to bet on such kind of society, and this is why Pope Leo wrote Magnifica Humanitas, in this idea of be comparative to bet on humanity in this age of machine.
ANDREW LEIGH: As I listen to Father Benanti's incredibly articulate response to your question I have to keep on reminding myself this is a man who's speaking in his second language, it is a real privilege to be sharing the stage with you.
I’m probably as worried as anyone else in the federal Parliament about catastrophic risks. I wrote a book in 2021 called What's the Worst that Could Happen?, which argued that there were four big catastrophic risks, of which artificial intelligence was the greatest.
I would have put my p(doom) at that stage at 5 per cent, that is the chance that the world ends once the systems become more capable than humans. Now I'd probably raise that to 10 per cent.
Dario's essay earlier this year, Bill Gates is now devoting a lot of his time, a lot of those who are following the capability of the frontier models very closely are becoming increasingly concerned. The international network of AI Safety Institutes is focusing on this, and I think the call from Dario to have external evaluators within AI firms is a useful one, and one on which we need to move quite quickly.
Again, when you're thinking about systems doubling in capability every seven months this is less about legislation and more about moving on all of the fronts where we can in order to bring in to check the greatest risks. And no, slowing the data centre roll out does not change the catastrophic risks. It may be good for many other reasons but I wouldn't want to connect those two.
JAMES RILEY: Thank you.
CORRIE MCLEOD: Just before we come to our next question, can I ask, at the time at the moment can we have a little peak behind the curtain? It seems we've got lots of very powerful people having conversations in closed door roundtables.
There was an article in a local paper yesterday about some of the lobbying that might be coming from some of the tech companies towards the Vatican. We have all of these conversations of various religious leaders, technology leaders, governments and institutes. You travel a lot, you're in a lot of these rooms, what's going on behind the curtain?
FATHER PAOLO BENANTI: So you are asking me to say secret to a Press Club?!
CORRIE MCLEOD: Well, just paint us a picture. Paint us a picture.
FATHER PAOLO BENANTI: Sorry, English is not my first language, I misunderstood. No, I'm joking! Well you know, the point is, once again sorry to repeat it, it's a matter of power, so it makes sense that someone is lobbying. But the problem is that the lobbying is not only with religions because, you know, religions find an agreement in 2020 on this.
So the majority of religion of the world signed their own code for AI ethics with six principle, you know, we fight each other for century, we find an agreement that, is it a miracle, I don't know, but what is saying here is that those group is the representative of the majority of living being on the earth. So it's not too late, element number one.
Element number two, what I see, I agree. It could be dangerous, but when the frontier lab that smell the idea that you know, ban on one side, ban on the other side, we'd like to regulate it, and you are going to make and IPO in the stockmarket, and they say ‘Don't do it – let's do it’.
In classical economical theory I'm teaching a political science. We call it 'cartel'. So we regulate ourself, don't regulate us. There are some other interests. So the lobby, you're right, we have to call it with the right label, they are lobbying, because they have interests, so they are not neutral player, and if they are not neutral player, well, we have an old traditional, old fashionable way that is to elect people that put their face on the couch and try to regulate it. It's different from lobbying. It is responding to people that elect them trying to do the best. Otherwise we'll have lobbying. That is not bad in itself, but it represent one interest, we need all the different stakeholders.
Also the stakeholder now have no voice. Don't forget that this is a cultural tools, so if we make one culture fit alls, a lot of different culture could be cancelled, or damaged by the face of the world, and I'm telling this to a country that know very well difference in culture in the different tradition, so you know how to handle it and how dangerous could be.
So it's a matter of bringing back the discussion in the place that we made for that, that is the public square of Parliament and things like that.
CORRIE MCLEOD: A question from Zac de Silva, Capital Brief.
ZAC DE SILVA, CAPITAL BRIEF: Thank you both for your addresses and what's been a really interesting conversation so far. Corrie stole the question I was about to ask, so I might re ask it in a slightly more blunt way. There's been a series of reports that Anthropic has been lobbying the Vatican, particularly around issues like AI consciousness.
First off to Father Benanti, have you personally been lobbied by people from Anthropic on this issue or others?
FATHER PAOLO BENANTI: Well you know, it's the first time that you take worst question for priest but not for politicians, but this is another topic that happen only in Australia. No, I'm joking!
Well, you know, there is a podcast on myself, really clear one year before this lobbying accusation, in which I say that this idea of conscience as a non sense problem for a real practical reason. We are talking about touring machine, my background is an engineer, sorry. So a touring machine cannot solve any kind of problem, only touring computable problem. Is it the conscience a touring computable problem? I'm sorry, no. So there is no way that the touring machines can have a conscience. Not because I'm a man of faith but because the science say that the machine cannot do it.
So the day in which we'll have a non touring machine as a computer, let's talk again, we don't have it, and so end of the discussion.
It's like with the clergy, you know, we have the Laudato si' with Pope Francis, [indistinct] psychology. There are lobbying inside, yes, there is a lot of people that say climate change do not exist. Don't be scandalised, we call it legitimate interest. Well, they would like to say that there is a conscience, the machine cannot express a conscience because an engineer say that, mathematician say that.
So this is interesting too. Someone can have an ideology here. Probably we was not any more use a political debate with ideology. Last one was Marxist. The new century is bringing new ideology on the table. Well, that make things more complicated, but it's not the first time that we saw it. We are ideological people that would like to produce result in an ideological way.
Last time that we believe it, well, we have a tragedy in society, so [indistinct].
ZAC DE SILVA: And can I ask a follow up if that's okay? There's been a development over the last couple of months of Google researchers released a model of a fruit fly's brain, and this has been turned into a whole bunch of different AI applications, you know, someone trained the digital model of this fly's brain to play chess, people taught it to play different games, people taught it to make music. It wasn't very good at a lot of these things because it's a fruit fly, but it's potentially the first of many, right. I wonder for both of you, what is your view on the ethics of building AI models that are based off biological brains?
ANDREW LEIGH: So my read of that was it was harkening back to an earlier era of artificial intelligence development. So as you know, there were two attempts to attain artificial intelligence, one was whole brain emulation, so let's try and build a system which is based on emulating first the brains of fruit flies, and then we'll build it up to bigger animals and ultimately to humans. And then there was the notion of neural networks, which moved quite slowly but has ultimately proved to be the foundation for the frontier models we have today.
So the fruit fly exercise, I think, is a fascinating little curiosity, but largely one which is representing a strand of research which ended up being a dead end.
The question is how fast neural networks advance when they get to a point at which their capability is exceeding humans, and we're just striking an increasing number of tasks in which they're doing so.
So you look at the Stanford report, they're now just running out of human tests having done all of the bar exams, the medical exams, the law exams, they're running out of tests in order to put these new models into as their capability improves.
FATHER PAOLO BENANTI: First of all, if we are looking for different way, LLM not working for every, and this is a self denunciation of a limitation of one approach, that in another way is saying something on some other things.
The second thing is, philosophically speaking – I'm a philosopher, the copy has not the same value of the original, so we are working on a copy, we are not attaching a fruit fly into something, and this is ethically speaking something that is interesting.
Well, supposing that tomorrow I have to fly back to Rome, and you know, the autopilot that is a copy of a fruit flight make me stable on the Indian Ocean. I really happy, you know, because that brain is really good in flying. So that can give us solution where we don't have it.
This is a huge bioethical topic that is experimentation on animals. Debatable, but until we are doing copy without killing the original I don't think it's a problem. The real problem here is we start from a computer that start to in some way have a nematic action on the human brain, neural networks.
Now because they work we are flipping down the things and using the neural network to understand the humans, and this epistemological can have some problem. But we can develop some kind of neural technology to manipulate the humans.
So the problem is if we believe that what we call something that works is the equivalent of something that exists, and once again the question is what does it mean to be human in this age, and technology running fast, but what we are lacking here is explainability. And this is not secondary, yeah.
ZAC DE SILVA: Thank you.
CORRIE MCLEOD: Next question is Lanie Tindale, InnovationAus.com.
LANIE TINDALE, INNOVATIONAUS.COM: Hello. So I have a more philosophical question, I think. Some AI models are built to emulate humans, trained on human data, they're designed to talk like us, empathise and mimic empathy like we do. People refer to AI agents, they suggest that AI will be autonomous and aware. So what is the difference between a human, an AI entrant, is AI a computer, and extension of humanity or even when it becomes better than us, many of the things that we do, will it be a higher power?
ANDREW LEIGH: So ultimately I don't think that AIs should be given rights, even as they exceed human capability. My calculator can do maths much better than me, but I don't think it should have attained any sort of personhood, and with a system which has pretty much infinite replicating powers, as soon as you begin to give AI rights, if you allow a right to vote, for example, to an AI whose capacity exceeded ours, then democracy immediately falls apart.
So I agree with Father Benanti that there is an essential humanness which marks us out as a species, and that needs to be an important part of how we develop the set of ethical rules around AI, and also the checks that we place on the development of these advanced systems which are improving at a remarkably rapid rate.
FATHER PAOLO BENANTI: Well, I think that the unions that fight in the relationship between human and machine are the best to say that there is a difference, and there is a quality in difference, but let me stress your concern. The problem is that if we don't regulate the market, the most fragile people could be damaged by dissimilarity.
One of the worst market that I'm seeing is the replica chatbot of dead person that are sold by a subscription to grieving people that never get over from that, and someone is making money. And this is happening because we don't have a law on digital remains. This is happening because no one is telling us that such kind of things could be like heroin, an addictive substance.
So we need to protect people and protect people means to accept the fragility of people as a human condition. A machine can be switched off and switched on, we not. So the length of our expression of life is something that need to be protected as a part of the human value.
And now we see for the aggression of the market. I'm not telling that the market is bad, okay, really clear this, but in some part of the market there is an aggression on the fragility of the human beings and that has to be protected.
Look at, for example, ultraprocessed food. We are a function from evolution that we call palatability, and we have another function that is nutrition. Palatability is give to us to have a good nutrition. Ultraprocessed food divide the two, and you have a super palatability and don't talk about nutrition.
What happen if we have an ultraprocessed education through AI, especially, you know, because if you eat ultraprocessed food you are accepting that there is an externality on public health, but public health is something that everyone is paying for.
If you have that on education on commission you have an externality on society for the advantage of someone. This is something that require a justice, you know, a form of social justice.
Is it easy? Is it fast? No. The technology is, this is part of the challenge, but this is why I love to talk about this topic with different stakeholder, because I don't have the solution, but we together have to find a solution. That's where the challenge is. I'm sorry, I'm talking with hands like every Italian.
ANDREW LEIGH: That’s great.
CORRIE MCLEOD: Nic Stuart, abilityNEWS.
NIC STUART, ABILITYNEWS: Thanks for a terrific discussion. I'd like to press further on what you've already begun to talk about, which is the responsibility and the anthropomorphising, the feeling that AI is its own independent person. I mean if, for example, I broke into Medicare and took a whole load of numbers, medical records of people, not did anything with it, but simply had it, then the Government would come down on me like the tonne of bricks saying, 'You've done the wrong thing'.
The, you know, various AI companies seem to be able to say, ‘Oh, well, it wasn't me that done that, it was this AI’, which obviously is the case, you know, they – but the guardrails weren't careful enough to make sure that it would behave in a legal manner.
How is it that we can regulate AI without actually indicating that it is a responsible legal action or someone is responsible for it?
ANDREW LEIGH: Let me just answer that with a particular example that we've been thinking about a lot in the consumer space, I'm responsible for Consumer Affairs. AI shopping agents are becoming more common, and when we work with agents in many parts of the economy we have rules to make sure your agent is acting for you. If you hire a real estate agent you don't expect them to get a kick back from the other side. If you hire a financial adviser…
NIC STUART: I'd have to sue my real estate agent!
ANDREW LEIGH: Indeed, and you can because there's a set of rules governing their behaviour. Likewise for a financial adviser, they need to disclose any commissions. But we need to be sure that when consumers are using AI shopping agents that their agents are not getting secret commissions and that those processes are what you would expect if you were engaging a human agent to go out shopping for you.
We actually have old state and territory crimes laws against being able to blackmail agents, but it may well be that we need to update the consumer law to take account of the challenges that AI agents are posing.
FATHER PAOLO BENANTI: Just to be complementary to this, you know, the topic is that we grow a huge section of economy in digitals, and that give us a lot of advantage. I am old enough to remember not when Internet start, but when there is no such kind of things like World Wide Web.
Once there was no password for email, because there is no value there, and no one will take the message that you send to your neighbour and things like that.
So we are talking of a huge digital infrastructure that arrives a little bit naive without protection and without barrier. Now that a huge amount of GDP is in the digital space, and this super powerful agent can find problem there that we never see, we discovered that we put a lot of value without a safe around. And it's a matter of cost. So we take the advantage until now, now someone has to pay the cost of that.
It's not only a matter of responsibility that is crucial, but it's also a matter to recognise that we don't pay the quality of service that we have, and probably this is not any more the case. If we put our healthcare system on digital, well, will you put a huge amount of money outside of a safe? Probably not. Why you put that on a system that is made to transfer it? So probably because it's convenient, but now that someone look at that like money we don't have the tools to protect them.
And this is a self reflection that we have to do, you know, because we take the advantage on a system that is fragile.
CORRIE MCLEOD: Okay, final question is Maurice Reilly.
MAURICE REILLY, NATIONAL PRESS CLUB CHIEF EXECUTIVE: Thank you, I'm asking this question for June Ramli from dailystraits.com. Father, what do you think about companies using AI to create pornography? Is AI generated pornography better than pornography involving real humans since no real humans are involved or exploited? And would AI generated pornography be permitted by the church?
FATHER PAOLO BENANTI: Well, are you scandalised if I tell you that I'm not practising such kind of business, but okay! No, well, beside the jokes. You know, I have some psychological studies in my background, and what I can tell you is make it sense that the business go where the money are? Yes, of course?
So if you look at the business model of the online advertise, the only website that make money with online advertise are the pornography website. So that means that a lot of money is there.
The second thing is, this idea of pornography is connected to another fundamental right. Do I have the right to know if what I'm looking is true or AI generated? And this is about the individual. So if I see a picture of myself killing someone else but it's AI generated, do we have the right to know these images is not real? We never decided it, and this is a danger also for democracy. We know what is there, you know, [indistinct] power in international influence in the digital space is one State against another State.
The third element is psychological. Yes, you don't have human beings there, and I know that I'm talking this in Australia in which it is decriminalised, for example, prostitution, and so there is no problem in being a prostitute in some states.
But the problem is, let's go straight to the hard point, for example, paedophilia, okay? Psychology will tell you that if you move from imagination to something more substantial you are removing the brake from human beings. So the idea that I can give a visualisation to my deepest fantasy, because no human is involved, could have also social consequence, because I make much more real some things to happen, and the problem is societal.
Should I allow any kind of public dimension like this or not? It's an educational problem, plural, different, multi ethical and I cannot fix it and I will not fix it, but the problem is let's count the consequence, and having in a public debate. This is democracy.
So we have to handle in a democratic way. How to fix it? Oh, thanks to God it's not to the church, it's to someone else, but my big fear is because I talk with, you know, pedophilic cure institution, it's for paedophilia. It's a tragedy. Because even simple allow people to move from imagination to something that is more near to reality, and I think we don't want it.
And so let's back to the first point. We need to give to people the right to know if something is real or AI generated, and this is the first solution for a lot different problem.
ANDREW LEIGH: If I can just add a quick postscript on that one, and there's decades of feminist literature making clear that the harms of pornography are not just to the people creating it, but also to the people consuming it.
The Digital Duty of Care would prevent social media feeds from serving up pornography to under 18s, and that will include AI generated pornography. Why? Because it leads to the dehumanising of others. 40 per cent of young Australians now say that they have been choked during sex, a behaviour which is largely emerging because people are seeing it in pornography, and whether that pornography is AI generated or whether it comes about by human actors, the damage to real human beings is very clear.
So I retain a significant concern about AI driven pornography, which could indeed begin to show even more violent and degrading imagery than we're seeing from that produced by actors.
MAURICE REILLY: And a final question – thank you, that was a really good answer, I was really impressed with that. China, the USA, the race to win AI. What's your sense of who's going to win that war?
FATHER PAOLO BENANTI: Well, I think we see two different model. In one model the State is in the market, on the other model I think that the market maybe is in the State.
MAURICE REILLY: State.
FATHER PAOLO BENANTI: I don't think that we want one of one, we need probably a third way, and it's where, you know, we can see different things with different company and different approach. Let's see what happen.
ANDREW LEIGH: Economists don't really like zero-sum games, we like the notion that there can be winners from this technology on both sides. I think that's true if we focus on the benefits such as the improved detection of diseases, improved treatment of cancer, the taking away of the most boring bits of people's jobs, but leaving them with fulfilling parts.
We don't want to race though in the most dangerous capacities. That becomes a whole lot like a nuclear arms race, and therefore the checks and balances that need to be put in place are around systems that might be misaligned with human values needs to be a conversation that's happening on both sides.
The models in China are not years behind those in the US, they may be months behind. So any sort of conversation around the risks of superintelligence needs to be occurring on both sides.
MAURICE REILLY: Thank you.
CORRIE MCLEOD: We have had a very rich conversation over the past hour. We've covered power and technology and ethics and externalities, and we've even touched on pornography. We have covered a lot, and I'd like to thank you both for such a thoughtful conversation.
Please join me in thanking Father Paolo Benanti and Assistant Minister Andrew Leigh.
ENDS