Closing Panel

Peter Lewis

Who's got cognitive overload? One of the things with conferences like this is there is so much information thrown at us. It's everything all at once, whatever that movie was. And the risk is that we leave — and it's been an intense experience — and we go back to our day-to-day work. So the purpose of this final session is really to not give you any more information, but just to settle with what's happened over the last couple of days and say what has stood out to all of us and what we can do with it.

The people on the panel should be known to most of you who have had a key role in the conference, but Greg, Johanna, Tiberio and Kim. They're going to help us go through this, and it's really simple. We've got about 45 minutes and these two questions: what are your key takeaways, and what do we do next? We'll start with the panel. Then I'm just going to go around the room with the mic and encourage as many of you as want to feed into this. For the guys on the stage — and the other thing you can do with this tool is upvote — and it will be interesting the things that people have said: “Yeah, that works for me too” as they'll be the things that are at the top of the screen when we get to the end of this.

So, Greg, what would you put in there, or what did you put in there if you were working on your phone?

Greg Sadler

For me, the biggest takeaway has been this point about power concentration. 18 months ago when we had the last forum, it was, I think, shortly after the election of Donald Trump, and power concentration was not at the centre of the conversation. It was in the conversation, maybe, but it wasn't a key idea. We voted it as the top issue in the Slido, and we also heard that it was the top issue on the same Slido in New Zealand during the weekend. So it's kind of extraordinary that this topic has gone from in the mix to the number one issue in this 18-month period.

And the policy diagnosis of this is complicated. How do we exercise our sovereignty? How do we avert power concentration? How do we create a multipolar world? It's a very complicated and difficult question, and quite different to technical AI safety. No one's writing the line of code that's going to magically solve power concentration.

Peter Lewis

It's really interesting. I wasn't able to come yesterday, but I've been here today, and I was expecting this to be more like Revenge of the Nerds — lots of people talking about their control of the technology. But it was actually more about the political economy, which was really interesting. What about you, Johanna?

Johanna Weaver

Look, I actually think my key take-out is the power of the room — so many of the conversations that we've been having about the power of the networks that are being built here. How many people here are not gatekeepers of knowledge, which is one of my pet peeves. How many people are looking at how can we bring more people into this conversation? And so I really applaud everyone for that approach. But the key take-out for me is actually one that's come from those conversations in the margins.

And that, for me, is existential dread. And I want to address this, because for me, I think it's quite important for everyone in this room to actually acknowledge and call this out. We're working in a field, as we've upvoted here, to say that it is impossible to keep pace with change, but it is becoming a mainstream conversation that more people need to know about it. And in large part, the expertise in order to respond to this is sitting in the room. So the speed at which it's moving, the urgency of which it is.

I don't know about you, but a lot of days I feel a huge amount of pressure and unable to be able to keep up, even when you're having big wins and what you're doing. So for me, I think a key take-out is what are we doing to take care of the people in this room so that we don't end up in a situation where we have that cognitive overload and we're not actually able to make the contributions that we need to make in the community.

Peter Lewis

I think it's interesting, the second part of that key takeaway. It's not just the pace of change, it's the connectivity, that notion of connective tissue between different silos. And it's also interesting — I don't think you can do that with a technical solution. You do it with the human solution, which is why everyone that's here today is part of that solution. Tiberio.

Tiberio Caetano

I guess one of the key takeaways for me is that it's not about AI at all. What we are talking about, it is about us. It's about agency, it's about humans, and it's about us looking inside — not only outside, but also inside — and closing our eyes for a moment and reflecting with ourselves and asking ourselves, how and from where do I want to respond to what is happening? I like to tell people that I'm the most optimistic AI safety person I know, and indeed I am.

And really, much of that comes from this conviction that we as humans, we kind of forgot, in the sense that much of the point of being a human is to be in presence with other humans, with ourselves, with our communities, with our creations, with our kids, with our parents, with our friends, even with our adversaries in football, when they beat us. I'm sorry, this has punctuated my week. And I use ‘punctuated’ by I could have said, but how do you, what's the word, is ‘punctured’ my week as well.

Johanna Weaver

Don’t go down that rabbit hole.

Tiberio Caetano

No, no, no, I won't talk about Brazil losing, don't worry. Well, I did, but my point here is that to me that's the big takeaway. The biggest takeaway is, what fight are we willing to fight? And is it really a fight, or is it a fight against whom? Does it make sense for us to fight against each other? Does it make sense for us to fight with AI? Is AI the target? What is the target of the fight? Is it our delusion? Is it our greed?

Is it our ignorance? Is it our lack of willingness to listen and to truly be together? And I think this is much the spirit of this forum. And my hope lies in aggregations and communities like this that we managed to put together. I think it's only by listening deeply and being together and thinking carefully that we can move on and actually live the best possible life that is feasible, and to take the best and most wholesome and mindful actions we can possibly summon from our deepest interior.

That's my observation.

Peter Lewis

Perfect. Kim, you were here at the previous conference as well. How has this felt different to the last time that you gathered? I think the room's bigger, the crowd is bigger, more diverse.

Kimberlee Weatherall

All of those things.

Peter Lewis

It's less academic and more practical.

Kimberlee Weatherall

Not so much academic and practical, but the things we were talking about in a relatively hypothetical or speculative kind of way even a year and a half ago feel much more real today. It's funny that you frame it in those terms, because this was going to be my response also. We did this 18 months ago. It did feel a little bit Revenge of the Nerds at times in that event — not wholly, but certainly at times.

Johanna Weaver

There's nothing wrong with the nerds.

Kimberlee Weatherall

I know, we’re nerds too. I claim my nerdom. But you know what? What I have loved about this time is feeling like there's more people on our side. In a sense, sometimes you do feel like this feels like a hopeless fight, or that there is that sense of existential dread. And I walk into this room and I see all these amazing people with lots of energy doing really important things on a whole lot of different parts of this problem. So I feel less alone. It's really nice.

The circles of this discussion have been widening. The reality of the challenges is hitting more people. And I take a lot of heart from that, because we need that in order to be in this fight, as we might say. The other thing that I have loved is — again, we did have this at the previous forum, but I've loved it again — I've been reminded very much as the conversations between all sorts of different levels and across disciplines. I've sat in these rooms and heard really technical people say the most important thing here is human.

It's fabulous. Because it is those conversations that we need to be having.

Peter Lewis

Is there anyone on the floor that wants to contribute on this? I noticed that two of the things that are going up the list there is importance of data centres for Australia and compute as statecraft, which I think some of the people that we're in the political economy one earlier today might be picking up on. But there are humans, there's hardware as well. I don't know if you want to talk to that or something else, but feel free.

Audience member

Kia ora koutou katoa. Thanks again to the wonderful people and speakers from New Zealand. And for a lot of the New Zealand delegation, hands up if you're still watching from New Zealand, just me? Cool. I've got another one back there. Okay. Thank you so much for giving us this event. It has been incredible in terms of not just where Australians are sitting, but also, as a country, from New Zealand's perspective, where we can sit in strategically and help support, in the very ANZAC way, in this big challenge of AI safety.

So thank you.

Peter Lewis

Johanna, any thoughts on that question of the fight? In your head, who are we fighting?

Johanna Weaver

Can I answer it slightly differently? I think in part what we're fighting is accepting the world as it is. I've heard a lot of people talking about: we need to slow down. The frontier labs are not slowing down. Trump is entirely unpredictable, and that's a very uncomfortable situation for Australia to be in, to not be able to rely on an ally. And China is becoming more and more proficient, and their models, as open source models, are proliferating around the world. So I think for me, actually being really real about that environment of the world — that's the biggest fight for me.

But I'm also optimistic, because we have done this national assessment of where Australia's capabilities lie, and Australia has a lot of capabilities. And a big part of that comes down to the people in this room who are focusing on, well, what is it to have safe and trustworthy AI? And I think when you look at those geopolitical circumstances and those geopolitical forces, that is a competitive advantage for Australia. And that's not a sound bite. If we can be building AI that countries and other people trust and can rely on, that will be the making of Australia as a nation.

So I feel the pressure. But I also think this is an incredible opportunity. And Anna Goldsworthy speaks about this in her essay — what an incredible moment that we are living in, and what an incredible opportunity to be able to shape that. But also, how do we take care of ourselves so we don't all burn out in the process of doing that?

Peter Lewis

I've got a man who's no stranger to conferences about AI. What have you taken out of the last two days, Toby?

Toby Walsh

Thank you, Peter. I'm heartened and disheartened at the same time. I'm heartened there's twice the number of people here this year. I'm heartened that most of the conversations now seem to be ones that are political flavour, which I think is where we're going to find the solutions to these problems. But on the other hand, I'm disheartened that the number of problems is just ever bigger. There was that list yesterday of the 2024 safety report and all the issues that we were worried about. Then 2026, all these new issues, but none of those 24 issues

have been solved. And I'm also disheartened that there are lots of people who aren't in this room. For example, there's almost no one from the tech companies that are causing these harms. I don't think we're going to solve it until we actually force them to come up and be in the same room with us.

Peter Lewis

There's a question up here I might throw to you, Greg. If you were to ask your local MP to do something around AI safety, what would you ask them to do right now?

Greg Sadler

I guess I'm often focused on what's the practical policy recommendation. What's the thing that's shovel ready now that will actually make a difference? I think at a panel yesterday, there was a — add a zero to the end of the Safety Institute's budget. I think that's a pretty common-sense shout, a bit of a low-hanging fruit. I think there's just this idea that — you talk about the inertia and momentum, and what is it that we're fighting, the acceleration of AI and the need to take action.

And I think something maybe we all agree on is that the default path that we're heading on is not a good default path, and the good paths are slightly off, or quite a long way off, from the path that we're on. So I think the practical policy recommendation was where we're addressing, identifying these concrete AI harms that are coming off the research conveyor belt. We need to start actually solving these things as they're happening, to agree with what Toby Walsh was saying. So I know that there were a number of sessions on biosecurity that I just missed.

I think there's practical actions that we can take today on those kinds of risks. And this is not a short list of practical things we can do. So let's start crossing some of these things off the list, start doing things and start building that bias towards action, where we're solving problems rather than just sitting here and admiring them.

Peter Lewis

I actually organised to have a local MP in the room. Kate, thank you for being here for the whole two days as well, not just turning up for a speech. So what sort of questions are you getting from your constituents, and how hard is it to make change from a position as a local MP?

Audience question

That's a big one. I think at a community level, the thing that is making sense to the people is the jobs impacts and probably the child safety impact. They're the two things that people talk about when I say, what are you most concerned about? Very few people say AI. But then if I ask, are you concerned about AI? People say yes. So I think it doesn't yet have the salience as a top issue, but I do think that's changed over the last few months. Actually, it is increasing.

In terms of how hard it is to have impact — I can put things on the agenda and put pressure on the government, which I'll continue to do. If I got to answer that question, what would you ask your local MP to do? I think emphasising the urgency and the need for ambition. And seeing the AISI is a good way of — good symbol of both of those. But I think there is just a lack of a sense of urgency in government on a lot of these issues at the moment.

And that will only really come from community pressure. So do write to your MP. Do say I'm deeply concerned and excited and want to say the government to being really ambitious about being a leader on the safety front and also making the most of the opportunities. And I think that that's when we start to see some movement.

Tiberio Caetano

Because Kate mentioned two words that to me are absolutely key to what I was about to say. So thank you, Kate. Urgency and ambition. Let's actually dig a little bit into that. When I see the evidence that we've seen over the past two days here and those graphs that showed up in several talks, we conclude one thing. We conclude that AI capabilities are not only fast, they are not only accelerating, but the acceleration is accelerating. Okay, let's use those words. Since to avoid numbers. That's a fact.

That's not an opinion. It's not something that the labs are just telling us. It's something that independent parties, governments and independent parties have assessed, evaluated. This is a fact. So we all have to start from the facts. If we don't agree on the facts, it can't work. That's the first thing I know, getting to the urgency there. And the second thing we also know is that — so that's a paradigm shift. That's the first observation. This is a paradigm shift. This hasn't existed before in history — a technology having three properties: being a general-purpose technology, being a technology that's not understood by those people who built it, and third, being a technology that improves itself with the help of humans.

At the moment, think of any other technology that has all those three properties together. I can't think of any. Please find one and tell me later so that I can learn more. Okay. That is what's happening. It's a paradigm shift. Now let's talk about governance, policy, safety and the other side of the equation. I don't see a paradigm shift there. That to me is the fundamental philosophical problem we have. We have a paradigm shift happening in what we are doing to the world. And with being happening with climate change, it's now happening with AI at a rate that I believe is much more concerning in terms of timelines.

And second, we are not changing the paradigm of policy, of governance, of what it means, or what safety means, or what even care and humanity means. And that, to me, is the fundamental challenge. So it's urgency in the sense that we need to respond now. But second, we need to be, I believe, tremendously ambitious. And if you think that you have ambition, you're probably not being ambitious enough.

Peter Lewis

It's about harnessing the energy and the ambition in the right direction. Again, to the panel — and then we might go to the room. What will each of you do next to ensure that the ideas, the connections from the last two days just don't go in the experience basket, but make them active parts of your work? I'll start with you, Kim.

Kimberlee Weatherall

I've already set up a bunch of meetings with people. That's the good stuff. And I'm serious, because that is part of the point of this, which is that I've made connections with a whole bunch of people who've come up to me and talked to me about some project that we're working on, or some mentoring they wanted, or whatever it was. So my first action is to follow up on all of that, because we have to keep growing the community and we have to keep making those connections.

And there are people, each of us can reach a whole lot of people outside of this room to be part of this. So that's the first thing. And the second thing is I am going to be thinking very, very hard about my harder problems. Those questions around holding on to humanity, the questions around broader capability rising across the population, because you've got China is saying the only way that we're going to get action is if people involved, this has to be a democratic project. We are a democracy.

Then working out how to get there is the other step. So that's then my two actions coming out of this. What about you, Hannah?

Johanna Weaver

I think for me it's about how do you build and leverage the community to do what we've done in this room at scale. A shout out to our Women in AI Safety fellows and recipients who are here. Thanks to the forum for supporting that. So looking at ways that we can practically extend and continue programs like that, I think, are really important. I also think engaging with politicians. For us at the Tech Policy Design Institute, we're based in Canberra. We do a lot of engagement directly with politicians to help them be upskilled, but we need people to be applying that pressure, as Kate is saying, to be engaged out there.

And then the other thing for me is, I know it is really important that we add that extra zero to AISI, but the other part of that is it can't just be defensive. I think we really all need to be adding our weight behind advocacy for public interest compute. To me, they are the two sides of the same coin. If we want the ambition, we're not actually going to get that if we're not also funding public interest compute. So that's my top three. Great.

Greg Sadler

Maybe a small thing. One of the things that we do is help run Australians for AI Safety. That's a website that's designed to empower the community to reach out to their members to raise these suggestions. So I think there's a small task to just do a revamp of that website and maybe put in some of those calls that Tiberio was talking about, so that people can more easily and more appropriately send the letter that we're talking about. So I think that's a very humble and important action.

Peter Lewis

It's interesting. The reporting of the minister's speech yesterday morning was, his shift was that AI risk is not a hypothetical. Bad things are happening now. That changes the equation a little bit, doesn't it? Are you seeing that in your work?

Greg Sadler

It's pretty clear that there's a diverse range of AI harms that are happening now. But some of those harms that so far away — superhuman cyber attack sounded like science fiction 18 months ago. But with Fable, this feels like it's on the cusp. You've got models being pulled by presidents, and this sort of geopolitical conflict is happening. So we're sort of now a measurable way down this pipeline towards these large-scale AI risks. So I think that "these risks are hypothetical" is no longer a talking point that you can.

Peter Lewis

Seem to be living through a science fiction. Maybe it's happened. Science fiction is happening. Tiberio, the number one here which Tara program pursue technical AI safety work. Can you talk to that?

Tiberio Caetano

Well, I think on that, I think we have a lot of talent in this country. We have no idea how many brilliant people are here working either independently or actually supporting AI safety, technical safety work elsewhere in the world. So I would connect that with what Greg just mentioned, the idea of some sort of hub, virtual hub, or a web page, or work towards scaling the community aspect to meet fundamentally the problems relate to what Johanna said before, in terms of we have to address the problem of scaling this — scaling the AI safety forum.

We're not going to scale physically, so we have to use AI, we have to use our creativity, and we want to find other ways to scale this conversation so that you can connect all the talent there with you. And if any of you put any thought into what that looks like, I just wish I wasn't really at the LinkedIn blitz. That's the last place you want to have the community. But where does the community exist, or does it exist everywhere at once?

Johanna Weaver

I think there are multiple communities. There's not one community, and that's actually part of the challenge. It's connecting the different communities that exist. So there's the communities that are concerned about data centres. There's the communities that are concerned about jobs. There's the communities that are concerned about kids. There are the communities of people who aren't even thinking about these issues and can't articulate them as AI concerns, but how to connect them up. So for me, one of our big passions at TPDI is saying there's no such thing as AI policy or tech policy.

These are just public policy issues. And having people realise that you can actually have a say in shaping it. And for me, that's the key thing, because so many people, I think, feel fatalistic that this is just something that's happening to us, as opposed to that we as Australia actually hold considerable agency, immense expertise, and we can help shape what happens next.

Peter Lewis

Anyone else got ideas to throw on the table while we're getting front? We've got one here. We've got one here. And then I'll be up the back.

Audience question

Thanks everyone. I spent a couple of years working in climate tech, and I've seen a lot of parallels in the conversations that have been in that community to those that are happening in this community — a sense of urgency, a need for ambition, and the concern about concentration of decisions being made about many people without those people. It's heartening to note that the bottleneck in this field seems to be talent, whereas the bottleneck in the other field seems to be resourcing to apply such talent. So I think we're in a good position in that sense.

Another commonality that these fields have is the fact that the change could be, well, smoothed, made better, improved by a conversation by and data, as is on the board here, deliberative democracy. And that doesn't happen in our present environments. So I would encourage people to make it happen. So if you do write to your MP, that could be an ask for deliberative process around what is it going to happen in this country in the next few years. Or do it yourself. And if you need some ideas, there are some experts in the room who can help you get started.

I wish you all the best. Have courage.

Peter Lewis

It's a really good point, isn't it, that the barriers to entry need to drop for people to be active as citizens in this space? I think so. Thank you.

Audience question

Hello. My name is David. I'm a researcher in safety, and I think maybe historically the safety community has kind of isolated its attention a bit to existential risk. And I think that crowd kind of likes to talk maybe to each other a lot, and also to a very select group of people, because that issue has been misunderstood or poorly understood in the past. But I think this wider audience that we have here today is really promising for the area, not just of existential risk, but also from the really large systemic problems that we're seeing from AI.

And I think it's an imperative on the safety community, particularly given Australia doesn't have that leverage over frontier model producers, to take a forward stance on the systemic issues and on engagement with policy and other special interest groups. I think that's a real opportunity for our unique position in the world to make a big difference there, where maybe other countries and historically, just the community in general has not paid as much attention. And maybe in terms of something practical I can offer. So on the third Thursday of every month at the Chippendale Hotel, we have an AI safety social meetup.

So please come along. It's at 530 on the third Thursday of every month at the Chippendale Hotel. To just talk about existential risk over a beer. Well, we just started it off again last month. It was the first one. There is beer.

Peter Lewis

That's good. This is an excellent example of self-care.

Audience question

So we have a lot of, I guess, high-level awareness in this room. But I'm curious to hear from the panel, how do you speak to the people that have little to no technical knowledge? How do I have this conversation with my mum, with my friend who isn't computer literate? How do you have those sorts of conversations, considering the level of nuance here of we do want data centres, but we wanted as public compute and all these sorts of things?

Tiberio Caetano

This is something we've been thinking about at the Gradient Institute quite a bit. And in fact, recently, just last week, we released a guide for supporting the average person to relate to AI safely. So in terms of doing that, part of it is actually a research work. A lot of effort was going into that exercise of us talking to family members, eliciting that information, testing ideas with them, bringing that back to the lab and figuring out, okay, trying to assess and interpret and analyse the data and look at that.

So a lot of this, I think, is about going out in the field and exploring, testing ideas.

Johanna Weaver

And I think what I would add to that is I actually think there is a high appetite for people. My brother, whose name happens to be friends, which is just perfect. He's a bricklayer, and I'm having conversations with him right now that — I think often we feel like we have to dumb the conversation down. And actually people want to engage with these issues. So it's actually having, being open to having the conversation, and then the power of analogy. So Zoe and I were talking the other day about how do we explain what model adaptation, domain adaptation, and alignment is.

And we came up with the analogy of Australia imports cars. And so we're importing foreign models. Australia has higher safety standards, so the cars have to go through particular processes. When they come, its steering wheel gets moved to the other side of the car. And then if you're using that car in a particular environment — four-wheel driving — you jack it up, then you put compressors and things like that on it. And using the power of analogy really, I think, is incredibly useful to making things more accessible.

And then you can have the conversation about model and domain adaptation in a really effective way. So I think what I've noticed, though, in the past two years is the appetite to have these conversations is so enormous. And so I think it's actually incumbent on us to pony up and have the conversations.

Peter Lewis

On that, not getting the public into it. I'll just go to Lizzie, who's got to leave in a minute, from Digital Rights Watch, to give her take on how you get people active in these discussions.

Lizzie O'Shea

What I would say is, in spite of the immense power of the industry and a sort of frustrating form of inertia, I would argue, on the part of government, the public in Australia are very hostile to lots of forms of AI, because I think they're concerned about harms. And so that's a conversation that can be facilitated, I think, in a very effective manner. It's translating that into pressure for action, both on the part of government but also industry. So I think industry kind of ignores forums like this at its peril.

And government should use it as a resource to convey some of these concerns around the that the public have, that is filtered through organisations like mine, like Digital Rights Watch, but also lots of other work that's being done in academic settings and other nonprofit settings and the like. And to turn that into meaningful action, because the alternative is much worse, I think, for government. And that's what I think they need to be reminded of. There will be low take-up of the technology. So the productivity benefits that they speak of all the time will not be available — will be increasing frustration with democratic processes that will lead to alternatives that they maybe can't control.

And so we need to find ways to make that interest relevant to government. And I think this is one form, animating the community organisations, community members to do this work is partly the job we've given ourselves, the Digital Rights Watch, but we welcome others that require some form of expertise, but also translation work of that expertise, so that people can turn their concerns that are quite general into specific examples, like Kate was talking about, child safety and environmental concerns around data centres. But the other big one is scams.

Lots of people talk about how frustrated they are at scams, and that's going to be elevated and accelerated by AI. So understanding where the public is at and trying to translate some of that expertise into a dialogue with communities, I think will be a really effective way to do it. And we'd welcome anybody who wants to join us in that activity.

Peter Lewis

Thank you. We're going to bump Mark over the other side of the room.

Audience question

Thank you. I'm Henry Fraser from QUT Law and ADM+S Centre. So we've had a regulatory trajectory, in terms of the regulatory conversation, that has landed in rather suddenly on this metaphor of safety as the prevailing concern. What do you see as the — and I was really sort of heartened to hear someone in technical AI safety happy to start thinking about socio-technical and systemic issues. But what do you think are the opportunities and the limits of this metaphor that is now kind of dominating the discourse about AI governance and responsibility and so on and so forth?

Greg Sadler

I can have one quick thought on that, while I other people stew. I kind of like the aviation metaphor that the minister used, but I guess when you start thinking about that a little bit more, things sort of break out very differently. There's an aviation safety conversation where you could say, oh, well, how many pressure cycles can aluminium have before a stress fracture extends from a rivet. But it's a very different story to where should a flight path go, and what suburbs should that overfly, and who should be exposed to the noise pollution from airlines,

which is again a different conversation to should billionaires have private jets. And all of these are really valid and important conversations, but they bring together different slices of information and different sectors and groups that you want communicating. So I suppose there's a lot of juice in the safety metaphor. But maybe we need to keep chasing it down the rabbit hole and say, when we're talking about some area, is this more analogous to the flight path, or is this more analogous to the rivet? So we can kind of have the right people in the conversation.

Also, to the extent that we're talking about meeting people where they are, if we just talk about AI safety, I think it can be a little bit alienating at times. It can sound too technical. What I have heard over the last couple of days is a whole lot of discussion about this harm or this problem or this concern. And I think that's actually where you meet people, where they are. We don't need to hook it all on a language like safety. We need to talk about the issues that people are facing, which are multitudinous, and we all know about them.

Peter Lewis

We've got one more question and one quick comment.

Audience question

I'm representing the Western Australian AI Hub. So first of all, I just want to thank you for the opportunity for me to come here. So my heart is actually on fire, because it is the very first time we are — I'm here with a lot of people from different states. And I just want to, let's, we have Kate, actually. And we also have — it is a great opportunity, actually, to be able to share today in the lightning talk, and also with a lot of people, you are very generous and sharing your knowledge.

And we just want to be more involved in what you are doing for the for our Australia. So that's number one. Number two is actually following up the guy over there and also like idea. So I think one thing that I feel it can be improved is when we are here talking about AI safety, maybe we can invite the big technology company coming here and then having a conversation with them. And also, as we also care about the vulnerable people, we would like to have people's representative — maybe children or elderly people or anyone — coming here so we can really have an authentic conversation.

Thank you.

Peter Lewis

Thank you. Last one.

Audience question

Thank you so much. I've heard a few comments about people not really being here who represent tech industry and perhaps frontier AI development. As somebody who has worked in the tech industry, building a lot of the leading new products, what would you say to people who are in those organisations today?

Peter Lewis

Anyone want to respond to that before we end?

Tiberio Caetano

It really depends who you're talking to.

Hunter Jay

I also wanted to do open invite to everyone doing co-working at AI safety. You can come co-work in Sydney for free. If anyone here wants to do co-working AI safety for free, come visit AI Safety Sydney space.

Peter Lewis

That sounds like a great initiative. There was one other great initiative. What will we do next? The afterparty. The Forest Lodge is just across the road, across the footbridge. And I will be there. I know we've got to get into the wind-up now. Thank you, everyone, for your input on this. My personal ad: if you like these discussions, we run a podcast called Burning Platforms. Myself and Lizzie O'Shea, who has been the star up the back there. So please subscribe to that and share it around as well.

I think Liam wants to do the wind-up now, but before we do, can we do final thoughts on the panel? The one thing I just wanted to get everybody to reflect on before we finish the elephant in the room, I think — and even the framing around AI safety — is a sense of inevitability. And I just want you to reflect what is inevitable, and what do you think we can control coming out of this discussion. Let's go from Greg to Johanna. I'll give each of you a mic.

Greg Sadler

There's a lot of inertia in the system in various ways, whether that's technical inertia or policy inertia. But I think there has to be some radical optimism here. We can shape that inertia. It's just going to require a lot of focused energy from the people in this room to do so. So I think there is hope, which is going to require a lot of work to move things.

Johanna Weaver

It’s unpredictable. That is inevitable. But I truly believe — but it is also backed up by the national assessment that we've just done — that Australia does have more power than most countries at the moment. And so it really behoves us to seize that power and actually shape what comes next. It's not about saying the big tech companies are only bad. There are some things that we can work with big tech companies for. There are some things we need to build at home. We have immense expertise and capability in this country that we need to be supporting, and not just ceding the space and saying that this is going to happen to us.

So for me, it will continue. The only thing that's certain is that it will continue to be uncertain.

Tiberio Caetano

So deep uncertainty is inevitable. Not only uncertainty, but deep uncertainty is inevitable. Deep agency is also inevitable, which is a good thing. So we —

Johanna Weaver

I don't think it is. I think we have to make a decision to just

Tiberio Caetano

Capability

Johanna Weaver

The people in this room are the leaders in making that decision.

Kimberlee Weatherall

What we have control over is what we do — what everyone in this room does, the actions that we can take. And we started out this whole forum with, we are taking agency in this space. We are owning our expertise, owning our power that we do have, and the opportunities that we have. And what we control is what we do with it.

Johanna Weaver

And, P.S., fund the regulators.

Kimberlee Weatherall

That too.

Liam Carroll

I think that's a great place to end. So please join me in thanking Peter Lewis and the rest of our panel members from this closing session.