WEBVTT - Apocalypse or a four-day week? What AI might mean for you 0:00:04.840 --> 0:00:06.560 The Australian Financial Review. 0:00:07.360 --> 0:00:11.280 I use AI in my assignments information sheets. I can 0:00:11.320 --> 0:00:14.600 pay directly into this thing called turboland dot A and 0:00:14.640 --> 0:00:19.279 it uses AI to generate me study notes, flash cards, 0:00:19.360 --> 0:00:21.000 quizzes and podcast The. 0:00:21.000 --> 0:00:23.040 Other day I asked it to create me a running 0:00:23.079 --> 0:00:27.120 program to get fitter for on my sport, and it 0:00:27.480 --> 0:00:31.280 created a five week running plan with long runs into 0:00:31.280 --> 0:00:31.920 full training. 0:00:32.680 --> 0:00:37.040 Teenagers are already using AI as part of their everyday lives. 0:00:39.240 --> 0:00:42.360 Is I make it summarize my notes and turn them 0:00:42.360 --> 0:00:45.199 into a podcast and I listened to it like on 0:00:45.240 --> 0:00:48.239 the day of exams. I'll put the whole grading rupric 0:00:48.400 --> 0:00:51.600 into chat GPT and then put my assessment and ask 0:00:51.680 --> 0:00:52.760 for harsh feedback. 0:00:53.280 --> 0:00:55.800 This was a big focus at the Financial Reviews AI 0:00:55.880 --> 0:01:00.960 summit last week. Tech leaders, chief executives and signed we're 0:01:01.000 --> 0:01:03.880 all asked what they were doing to prepare their kids 0:01:03.880 --> 0:01:07.880 for the future. The answers varied from encouraging them to 0:01:07.920 --> 0:01:12.040 become AI ninjas to doubling down on the human experience 0:01:12.200 --> 0:01:17.039 by studying landscaping or philosophy. But everyone agrees that we 0:01:17.080 --> 0:01:21.200 are hurtling toward an AI disrupted future and our kids 0:01:21.480 --> 0:01:25.119 will be the ones dealing with the fallout, and while 0:01:25.160 --> 0:01:29.399 there's optimism about scientific and medical breakthroughs and a much 0:01:29.480 --> 0:01:34.679 needed boosting productivity, there's also concern about job losses and 0:01:34.800 --> 0:01:39.120 power concentrated among a handful of mercurial tech billionaires. 0:01:39.520 --> 0:01:41.399 The thing that I find a bit frightening at the 0:01:41.400 --> 0:01:44.600 moment is it's so unclear what the next generation going 0:01:44.600 --> 0:01:47.039 through high school, how they're going to plot this path 0:01:47.080 --> 0:01:49.760 over the next few years, because they're finishing school and 0:01:49.760 --> 0:01:53.520 going into workforce when nobody in the workforce even knows 0:01:53.600 --> 0:01:55.160 what they're going to need in the next few years. 0:01:55.800 --> 0:01:59.800 Welcome to the Finn I'm Lisa Murray. This week technology 0:02:00.320 --> 0:02:03.560 Paul Smith on the AI future and whether it's more 0:02:03.680 --> 0:02:24.040 likely to be utopia or dystopia. It's Thursday, June twelfth. Hi, Paul, 0:02:24.080 --> 0:02:25.520 thanks for coming on the podcast. 0:02:25.720 --> 0:02:26.280 Thanks for having me. 0:02:26.360 --> 0:02:26.600 Lisa. 0:02:27.040 --> 0:02:30.560 You hosted a few panels at the AI Summer last 0:02:30.560 --> 0:02:33.519 week and spent the day working the room. There was 0:02:33.560 --> 0:02:37.359 a general feeling that it's all happening much faster than 0:02:37.440 --> 0:02:42.520 we expected, and companies, investors, economies and governments aren't ready 0:02:42.560 --> 0:02:45.359 for it. What were your main takeaways? 0:02:45.800 --> 0:02:48.320 Well, yeah, you're right, there was definitely a sense of 0:02:48.960 --> 0:02:53.760 foreboding the Australias in danger of falling behind. But overall 0:02:53.800 --> 0:02:56.160 I came away with a sense that there was a 0:02:56.200 --> 0:02:59.920 really clear need to have some frank conversations about both 0:03:00.080 --> 0:03:03.000 benefits and dangers of a shift that's coming that's going 0:03:03.080 --> 0:03:07.640 to redefine national productivity, employment and in society itself for 0:03:07.680 --> 0:03:11.800 decades to come. We had a couple of Australia's biggest 0:03:11.919 --> 0:03:15.520 chief executives, conwalth Banks mac Common and Tawsha's Vicki Brady 0:03:15.960 --> 0:03:18.480 appearing at the summit as well, and they've both spent 0:03:18.520 --> 0:03:21.000 a lot of time over in Silicon Valley recently and 0:03:21.120 --> 0:03:24.440 really are positioning themselves as forward thinking in the AI 0:03:24.600 --> 0:03:27.400 realm and keen to make sure that their organizations aren't 0:03:27.880 --> 0:03:31.080 left behind. And they both admitted that things are moving 0:03:31.080 --> 0:03:33.960 at a faster rate than they previously expected and had 0:03:34.000 --> 0:03:36.120 some open questions about how it's going to affect their 0:03:36.120 --> 0:03:40.320 companies now. One of the most interesting speakers there was 0:03:40.400 --> 0:03:44.880 Liezel Yearsley, the founder of Australian artificial intelligence company Akin. 0:03:45.280 --> 0:03:48.160 She's a real veteran of the AI scene in Australian 0:03:48.240 --> 0:03:52.400 has built successful chatbots way before they were a thing 0:03:52.440 --> 0:03:59.440 that everyone was chasing. She said, Australia's grossly underestimating the 0:03:59.480 --> 0:04:01.520 scale of change AI is going to bring, and put 0:04:01.520 --> 0:04:03.120 it in a historical context. 0:04:03.440 --> 0:04:06.040 I think the magnitude of what's coming at us is 0:04:06.080 --> 0:04:09.480 nothing less than the transition we saw between the eighteen 0:04:09.520 --> 0:04:10.960 hundreds and the nineteen hundreds. 0:04:11.120 --> 0:04:13.400 She pointed out that in the eighteen hundreds we were 0:04:13.440 --> 0:04:15.200 all involved in manual labor. 0:04:15.560 --> 0:04:19.359 The eighteen hundreds, ninety five percent of us were pushing 0:04:19.360 --> 0:04:22.160 a plow through the dirt, and that's what we did 0:04:22.400 --> 0:04:24.480 with our time and our effort. We were dominated by. 0:04:24.440 --> 0:04:27.960 Muscle, and then we moved on from that with muscle 0:04:28.000 --> 0:04:30.719 to machines in the Industrial revolution, and. 0:04:30.600 --> 0:04:34.560 What we did with industrial revolutions machines replaced muscle, completely 0:04:34.600 --> 0:04:36.640 transformed our own entire planet. 0:04:36.800 --> 0:04:39.440 And then with machines doing most of the manual work, 0:04:39.800 --> 0:04:41.960 we now work mostly with our brains. 0:04:42.600 --> 0:04:46.040 You know, ninety five percent of labor is cognitive, and 0:04:46.080 --> 0:04:49.240 we AI is coming after absolutely everything. 0:04:52.040 --> 0:04:55.119 So she paints a worrying picture for a lot of us, really, 0:04:55.480 --> 0:04:58.800 because if ninety to ninety five percent of the work 0:04:58.839 --> 0:05:00.880 is now able to be done by box, that doesn't 0:05:00.960 --> 0:05:02.520 leave a lot for the rest of us to think about. 0:05:02.520 --> 0:05:06.280 Does it The pace of change was repeatedly emphasized at 0:05:06.279 --> 0:05:09.560 the summit, But would you say the mood overall was 0:05:09.600 --> 0:05:13.080 generally optimistic, that most people were signed up to a 0:05:13.120 --> 0:05:15.599 more utopian view of the AI future. 0:05:15.960 --> 0:05:17.479 Well, that's right. I mean, you're not going to have 0:05:17.800 --> 0:05:21.040 too many people turn up to an AFR AI summit 0:05:21.120 --> 0:05:22.960 and say that they're going to end the world, So 0:05:23.000 --> 0:05:24.680 you kind of have to leave that floating in the 0:05:24.680 --> 0:05:28.120 background a little bit and hear the case for the 0:05:28.200 --> 0:05:31.760 positive impacts. And I might sound like I'm on the 0:05:31.800 --> 0:05:34.680 side of the doomsayers, but there really are some very 0:05:34.720 --> 0:05:38.279 positive examples of how AI has already made a major 0:05:38.279 --> 0:05:42.200 difference and some positive changes that were highlighted. We had 0:05:42.480 --> 0:05:46.440 James Manyika, who's Google's head of Technology and Society, presenting 0:05:46.480 --> 0:05:48.320 early on in the summit, and he was talking about 0:05:48.560 --> 0:05:51.799 the scientific breakthroughs that AI has already helped them usher 0:05:51.839 --> 0:05:55.599 in over at Google. Google's team won a Nobel Prize 0:05:55.720 --> 0:05:58.880 for using AI to predict the shape of proteins, and 0:05:58.920 --> 0:06:03.000 he's talked about the potential for AI to improve diagnostics, 0:06:03.080 --> 0:06:07.120 detect natural disasters, and how autonomous driving is taken often 0:06:07.160 --> 0:06:09.719 people can be away from dangerous positions on mind sights 0:06:10.200 --> 0:06:13.120 and can hold online meetings in different languages in real 0:06:13.160 --> 0:06:15.800 time without a human translator. I think that was a 0:06:15.839 --> 0:06:20.440 really tangible example, and they've demonstrated that technology recently is 0:06:20.760 --> 0:06:23.240 in development. That really is the first example, one of 0:06:23.279 --> 0:06:25.680 the first examples that you see of real sci fi 0:06:25.760 --> 0:06:27.960 stuff that you'd really want. 0:06:27.520 --> 0:06:31.760 You see the efficiency gains of having that conversation. 0:06:31.440 --> 0:06:33.280 Being on holiday and going into a shop and being 0:06:33.279 --> 0:06:36.160 able to speak to someone without cracking open a phrasebook 0:06:36.160 --> 0:06:40.360 and pointing and shouting. We also had Craig Blair on 0:06:40.480 --> 0:06:42.560 Wile of the Panels. He's the founder of er Try Ventures, 0:06:42.560 --> 0:06:45.200 one of the biggest venture capital companies in Australia, and 0:06:45.279 --> 0:06:48.320 he was talking about how AI is already helping startups 0:06:48.320 --> 0:06:51.679 in his portfolio go from the idea stage to making 0:06:51.720 --> 0:06:54.840 profits in just months with a much smaller team than 0:06:54.839 --> 0:06:57.359 they would have in the past, so really fast forwarding 0:06:57.520 --> 0:07:00.600 the creation of companies. And one point that people said 0:07:00.880 --> 0:07:04.359 was that Australia really hasn't scratched the surface yet of 0:07:04.520 --> 0:07:07.760 making a fortune about our natural resources and our natural 0:07:08.080 --> 0:07:11.800 strength in terms of being a great location to host 0:07:11.920 --> 0:07:15.000 data centers and other infrastructure that's going to power the 0:07:15.120 --> 0:07:18.880 AI boom. And so we've got abundant renewable energy options 0:07:18.920 --> 0:07:22.080 here and a lot of space that hasn't been used yet, 0:07:22.080 --> 0:07:27.520 and a very stable, relatively political and business scene. So 0:07:28.400 --> 0:07:30.800 the opportunity is there for Australia to make a mosa 0:07:30.840 --> 0:07:34.280 from the AI revolution, but it really hasn't been fully 0:07:34.280 --> 0:07:34.840 tapped yet. 0:07:35.400 --> 0:07:38.120 That was definitely said repeatedly, wasn't it. We have space, 0:07:38.280 --> 0:07:42.120 we have renewable energy, and we have a general enthusiasm 0:07:42.200 --> 0:07:43.480 for new technology. 0:07:43.640 --> 0:07:44.600 That's right, Yeah, Paul. 0:07:44.680 --> 0:07:46.880 The focus at the summit was not so much about 0:07:47.000 --> 0:07:49.800 generative AI. Last time we had you on the podcast, 0:07:49.840 --> 0:07:53.360 we were probably talking about that it's the chat GPTs 0:07:53.440 --> 0:07:57.840 of the world, But now everyone's talking more about agentique AI, 0:07:58.440 --> 0:08:01.640 that is, AI agents that can go further than chet 0:08:01.680 --> 0:08:08.160 GPT and carry out task. Explain properly the difference and 0:08:08.240 --> 0:08:12.000 give some real life examples of how these AI agents 0:08:12.000 --> 0:08:12.920 are already working. 0:08:13.000 --> 0:08:14.960 You did a pretty good job and explain yourself that 0:08:15.000 --> 0:08:18.320 I mean it basically is that I mean I remember 0:08:18.320 --> 0:08:20.360 the first time I came on the podcast to you, 0:08:20.400 --> 0:08:23.800 we were also excited by chat GPT is a fun, 0:08:23.880 --> 0:08:26.440 novelty thing that we had it right a rap for 0:08:26.480 --> 0:08:28.280 me to do and I don't think I nailed it, 0:08:28.320 --> 0:08:32.280 but you know, just ask it a question and watch 0:08:32.320 --> 0:08:35.800 its unfold. That's Australian Financial Review is sharp and bold 0:08:35.880 --> 0:08:39.520 and even Bill Gates knows the deal. Chat GP teams, 0:08:39.640 --> 0:08:42.880 revolutions are going to change the way we feel. 0:08:43.360 --> 0:08:43.839 You got there. 0:08:43.960 --> 0:08:47.000 Yeah, it was fun. So that's generos of AI. It 0:08:47.080 --> 0:08:50.560 creates new content from prompts that you've asked for, whereas 0:08:50.559 --> 0:08:55.160 agentic AI is much more obviously valuable to businesses and 0:08:55.600 --> 0:08:59.559 much more clear how it's going to improve productivity and 0:08:59.679 --> 0:09:03.320 maybe be changed the way people work. So it's it's 0:09:03.400 --> 0:09:06.280 like having an assistance or an agent who can do 0:09:06.360 --> 0:09:10.280 things for you. They can autonomously make some decisions and 0:09:10.320 --> 0:09:14.920 take actions to complete tasks without needing a human constantly 0:09:14.960 --> 0:09:17.560 prompting and changing what it wants it to do. So 0:09:17.840 --> 0:09:21.679 in a consumer setting, it would be like asking an 0:09:21.720 --> 0:09:24.400 agent say can you please book me a flight to 0:09:24.920 --> 0:09:27.559 Brisbane next week and find me the best price, and 0:09:27.600 --> 0:09:29.400 it will just go off and do it for you, 0:09:29.600 --> 0:09:32.760 navigating through the different websites making payments. If you give 0:09:32.800 --> 0:09:36.199 it permission, that kind of thing now from a business perspective. 0:09:36.240 --> 0:09:39.319 At our summit, we heard from Suncorp about how it's 0:09:39.320 --> 0:09:43.520 got AI agents that help it detect problems before they arise. 0:09:43.800 --> 0:09:46.360 For example, in cyclone Alfred, it was able to predict 0:09:46.360 --> 0:09:49.080 which houses were going to be hit and more likely 0:09:49.160 --> 0:09:51.680 to have problems, so it prepared them to respond to 0:09:51.720 --> 0:09:54.360 claims fast, as they said. And we had am Z's 0:09:54.400 --> 0:09:56.880 chief technology officer though as well, talking about how they've 0:09:56.880 --> 0:10:00.760 got AI agents reading through hundreds of pages of loan documents, 0:10:00.880 --> 0:10:03.800 getting property valuations and things like that, and he was 0:10:03.840 --> 0:10:06.480 saying that these agents have removed a day's worth of 0:10:06.520 --> 0:10:10.520 work for people in assessing complex corporate loans. So yeah. 0:10:10.720 --> 0:10:14.920 Agentic AI is also increasingly common in the area of 0:10:14.960 --> 0:10:18.640 tech development and coding, where the concept of vibe coding 0:10:18.880 --> 0:10:22.000 was spoken about quite a lot, which means well, anyone 0:10:22.400 --> 0:10:25.240 who doesn't know how to code can ask an AI 0:10:25.320 --> 0:10:28.280 platform to design something like a website or an app, 0:10:28.640 --> 0:10:31.040 and they get codes spat back at them and see 0:10:31.200 --> 0:10:34.280 the results of it when it's executed, and can refine 0:10:34.280 --> 0:10:36.199 it with further prompts, so they don't even really need 0:10:36.240 --> 0:10:38.840 to know how to code, but can set about coding. 0:10:39.160 --> 0:10:41.920 With all of those AI agents running around, there's a 0:10:41.960 --> 0:10:45.920 real debate now about the impact on jobs. There are 0:10:45.960 --> 0:10:48.880 some extreme predictions out there, some of which have come 0:10:48.920 --> 0:10:51.000 out in the last few weeks, that it could wipe 0:10:51.040 --> 0:10:54.960 out half of all entry level white collar jobs. Do 0:10:55.000 --> 0:10:56.160 you think that could happen. 0:10:56.559 --> 0:10:59.000 There's certainly no way of saying that it won't happen. 0:10:59.200 --> 0:11:01.960 Do you refer into the chief executive of the huge 0:11:02.040 --> 0:11:05.320 AI company, Anthropic Dario Amadai, who a few weeks ago 0:11:05.800 --> 0:11:08.720 said in an interview that AI could send the unemployment 0:11:08.840 --> 0:11:11.160 rate in the US to up between ten percent and 0:11:11.160 --> 0:11:14.079 twenty percent in the next one to five years. So 0:11:14.160 --> 0:11:16.760 that's as the technology moves from helping humans do their 0:11:16.880 --> 0:11:21.160 jobs to replace them outright, and is often happening first 0:11:21.280 --> 0:11:24.360 in the tech developer space, which is ironic that people 0:11:24.440 --> 0:11:27.000 that would have maybe been designing these systems and thought 0:11:27.040 --> 0:11:29.000 that they would be right for years to come are 0:11:29.000 --> 0:11:31.920 the ones that have found themselves being disrupted first. There's 0:11:31.960 --> 0:11:35.160 been reporting as well from the US again that the 0:11:35.240 --> 0:11:39.080 unemployment rate for graduates has picked up as managers have 0:11:39.160 --> 0:11:41.240 been encouraged to go AI first, and a lot of 0:11:41.240 --> 0:11:44.520 the jobs that have been harder for graduates to get 0:11:44.559 --> 0:11:48.120 are in areas that AI has been typically strong, like 0:11:48.240 --> 0:11:53.480 finance and programming and development, and Telsha's CEO, Vicky Brady, 0:11:53.920 --> 0:11:56.920 said that it's important to be honest with employees and 0:11:56.960 --> 0:11:59.640 that she thinks that Telsha's workforce is going to be 0:11:59.640 --> 0:12:01.240 smaller in five years. 0:12:01.440 --> 0:12:04.920 When you're a leader, I think that transparency, honesty is 0:12:04.960 --> 0:12:08.920 so incredibly important, and how do you do that in 0:12:08.960 --> 0:12:12.440 a way where you also don't want to panic people. 0:12:12.800 --> 0:12:15.560 She stressed that she wasn't sitting there with a number 0:12:15.640 --> 0:12:18.040 of how much smaller are the organization's going to be 0:12:18.160 --> 0:12:19.400 that she was keeping secret. 0:12:19.840 --> 0:12:23.160 I don't know what our workforce looks like in five years, 0:12:23.720 --> 0:12:26.120 but what I do know is I think jobs are 0:12:26.120 --> 0:12:29.319 going to look different. I think it's likely our workforce 0:12:29.400 --> 0:12:30.280 will be smaller. 0:12:30.800 --> 0:12:32.920 So it's hard to know whether it's going to be 0:12:32.960 --> 0:12:35.560 a job apocalypse. But the feeling out of the summit 0:12:36.120 --> 0:12:38.480 is that the impact of this is going to be uneven. 0:12:38.520 --> 0:12:41.400 It's going to create some jobs and replace some jobs. 0:12:41.960 --> 0:12:44.240 There's still a lot of talk about having humans in 0:12:44.280 --> 0:12:46.480 the loop. We're going to keep you in the loop. 0:12:46.920 --> 0:12:49.280 You in the loop until you realize you're not, And 0:12:49.320 --> 0:12:51.760 so this is to have people keeping an eye on 0:12:51.760 --> 0:12:54.960 the output and avoid rogue agents running around. And I 0:12:54.960 --> 0:12:57.920 think the thing that came obvi is to me though, 0:12:57.960 --> 0:13:00.200 and it's been obvious for a little while now, is 0:13:00.200 --> 0:13:02.720 that the line that's been regularly used by tech companies 0:13:02.760 --> 0:13:06.520 and executives responsible for AI, that AI is on again 0:13:06.559 --> 0:13:11.439 augment rather than replaced workers is palpably false. But well, 0:13:11.440 --> 0:13:13.800 we're all worried about these job losses. They might end 0:13:13.880 --> 0:13:16.400 up being the least of our problems. According to Liesel 0:13:16.480 --> 0:13:19.520 Yearsly here we spoke about before. She had a lovely 0:13:19.559 --> 0:13:22.960 phrase where she said it's bringing out the worst of capitalism, 0:13:23.080 --> 0:13:26.640 and she was veering towards a really dystopian view of 0:13:26.760 --> 0:13:27.640 the AI future. 0:13:29.200 --> 0:13:32.520 What we're not really thinking about is that we're actually 0:13:32.600 --> 0:13:36.640 creating a thing that has a form of sentience. It's 0:13:36.640 --> 0:13:41.439 a coevolution. It's a fundamental shift to our society and 0:13:41.480 --> 0:13:46.160 our species. So I think the magnitude of shift that's 0:13:46.240 --> 0:13:49.800 coming is nothing like we've seen. We don't have a 0:13:49.840 --> 0:13:52.240 generation to adjust this time. It's happening in a regional 0:13:52.280 --> 0:13:52.960 space of time. 0:13:55.360 --> 0:13:59.240 And there are more extreme views out there as well. 0:13:59.280 --> 0:14:02.600 In April former researchers from Open Ai and some other 0:14:02.640 --> 0:14:06.480 respected people in the AI industry over in Silicon Valley 0:14:06.520 --> 0:14:10.760 released this report called AI twenty twenty seven, and it 0:14:10.800 --> 0:14:13.280 has caused a bit of a stir in Silicon Valley, 0:14:13.320 --> 0:14:17.160 lots of people debating it. It basically describes a fictional 0:14:17.240 --> 0:14:22.880 scenario based on some evidence and some theorizing about what 0:14:23.080 --> 0:14:27.240 could happen when AI systems surpass human level intelligence, which 0:14:27.240 --> 0:14:30.160 they all expect it to do in the next few years, 0:14:30.680 --> 0:14:34.000 and what might happen when AI gets away from us, 0:14:34.040 --> 0:14:55.920 And to be honest, it's not looking good for humans. 0:15:00.000 --> 0:15:03.280 We're talking about what an AI future looks like, and 0:15:03.320 --> 0:15:08.000 there are competing views. The utopian view highlights the potential 0:15:08.040 --> 0:15:12.200 for scientific and medical breakthroughs and huge boost in productivity 0:15:12.200 --> 0:15:15.040 that will add billions to the economy, but there are 0:15:15.080 --> 0:15:19.000 more dystopian views about the impact of AI on society, 0:15:19.600 --> 0:15:27.040 high unemployment, and even apocalypse. What is AI twenty twenty seven, Well. 0:15:26.880 --> 0:15:29.240 First of all, it was a rippin read. It's a 0:15:29.360 --> 0:15:33.520 scenario planning exercise that's been conducted by an expert panel 0:15:33.920 --> 0:15:39.280 led by a former OpenAI insider, and it basically takes 0:15:40.200 --> 0:15:44.000 the scenario from roughly where we are today and tries 0:15:44.040 --> 0:15:48.680 to realistically assess what happens if certain decisions are made 0:15:48.920 --> 0:15:53.240 and they are developing these AI systems, and the aisystems 0:15:53.320 --> 0:15:56.200 are then start helping to develop the next versions of 0:15:56.200 --> 0:15:59.400 the AI systems, and they somewhere along the way lose 0:16:00.200 --> 0:16:03.720 the ability to truly see what the AI systems are 0:16:03.720 --> 0:16:06.400 trying to do, and the AI begins to be able 0:16:06.440 --> 0:16:10.640 to hide its true intentions. Things spiral out of control. 0:16:10.800 --> 0:16:15.280 So it's this narrative style warning about what might happen. 0:16:15.600 --> 0:16:17.760 It tells the story of what might happen. 0:16:17.560 --> 0:16:20.560 That's right, And ultimately we get to the brink a 0:16:20.600 --> 0:16:23.200 real cold war between the US and China, and there's 0:16:23.200 --> 0:16:25.960 a decision to be made in twenty twenty seven about 0:16:25.960 --> 0:16:30.160 whether the US company, which is a few months ahead 0:16:30.400 --> 0:16:33.800 of the Chinese one, whether it stops to let humans 0:16:33.920 --> 0:16:36.360 regain control of what's happening, or whether they press on. 0:16:36.680 --> 0:16:40.400 In the SNAI where the company slows down, things become 0:16:40.440 --> 0:16:43.200 a little bit more manageable, still not great, but manageable. 0:16:43.480 --> 0:16:47.400 But in the scenario where they press on five six years, 0:16:47.680 --> 0:16:49.840 there's no more humans left on Earth. I mean, the 0:16:49.880 --> 0:16:53.400 Earth is covered in data centers and other AI infrastructure, 0:16:53.480 --> 0:16:55.480 and we have been taken out of the game. 0:16:55.480 --> 0:16:57.520 And we've been taken out of the game because they 0:16:57.560 --> 0:17:00.800 need the space and the power to keep. 0:17:00.880 --> 0:17:03.440 We seeing themselves. Yes, he's being useful. 0:17:03.720 --> 0:17:06.040 So as you said, a ripping read it is, Yeah, 0:17:06.160 --> 0:17:07.919 are people taking it seriously? 0:17:08.359 --> 0:17:08.639 Now? 0:17:09.000 --> 0:17:11.600 People are taking it seriously in so far as these 0:17:11.600 --> 0:17:15.120 aren't idiots putting it together and they aren't actually coming 0:17:15.119 --> 0:17:17.119 out and saying this is what they think will happen. 0:17:17.240 --> 0:17:21.399 They're just putting scenarios out there to focus minds on 0:17:21.840 --> 0:17:25.160 the discussion, and they point out that there's a genuine 0:17:25.200 --> 0:17:29.560 concern about the amount of power being concentrated in the 0:17:29.560 --> 0:17:32.879 hands of a really small group of tech billionaires who 0:17:33.160 --> 0:17:36.720 may or may not have the best interest of humanity 0:17:36.760 --> 0:17:39.560 at heart. So it's really a conversation startup. And I 0:17:39.600 --> 0:17:43.600 had the opportunity to ask one of open aye's most 0:17:43.640 --> 0:17:47.480 senior executives about that, Jason Quan, who's their chief strategy 0:17:47.480 --> 0:17:50.680 officer and who's worked with Sam Altman from years backwards 0:17:50.680 --> 0:17:53.040 in Sydney, and we had had a chat about it, 0:17:53.080 --> 0:17:56.600 and he clearly disagrees with this sort of dystopian ending 0:17:56.640 --> 0:17:58.680 of it, and obviously wouldn't be doing what he's doing 0:17:58.760 --> 0:18:02.440 it if he didn't, But his response was pretty measured 0:18:02.480 --> 0:18:04.040 to it. He thought it was a good narrative and 0:18:04.080 --> 0:18:06.800 he thought it was worth having these conversations. But he 0:18:06.880 --> 0:18:11.320 really thinks the best way to understand the impact of 0:18:11.359 --> 0:18:14.920 these products is to start using them and see what 0:18:15.040 --> 0:18:18.080 sort of problems arise. And he thinks in a much 0:18:18.119 --> 0:18:22.879 longer term horizon, they talk about a thirty year time 0:18:23.080 --> 0:18:28.159 arc whereby these things will cause major changes. But society 0:18:28.320 --> 0:18:30.680 works out a way because and it's in no one's interest. 0:18:30.720 --> 0:18:33.639 It's not in the AI companies themselves interests, it's not 0:18:33.720 --> 0:18:37.280 in government's interests for everything to fall apart. So he's 0:18:37.320 --> 0:18:39.840 got a more optimistic view that we will figure things out. 0:18:41.160 --> 0:18:43.960 These are all very big issues. There's a lot at stake, 0:18:44.280 --> 0:18:48.840 and yet the Australian government doesn't yet have an AI policy. 0:18:49.840 --> 0:18:52.280 Are we being left behind? How do we compare to 0:18:52.359 --> 0:18:53.159 other countries? 0:18:54.040 --> 0:18:56.760 Well, the nature of politics means that it feels like 0:18:56.800 --> 0:19:00.359 we're back at square one. We did have a policy 0:19:00.880 --> 0:19:02.879 plan of sorts. It's due to be announced at the 0:19:02.960 --> 0:19:05.240 end of the year, but that was very much embodied 0:19:05.400 --> 0:19:09.119 in Ed Husick, who was the Industry and Science Minister 0:19:09.240 --> 0:19:12.280 for the last term of government and spent a lot 0:19:12.280 --> 0:19:15.359 of time going around and talking to the industry, talking 0:19:15.359 --> 0:19:18.840 to stakeholders about what the changes were, whether we need 0:19:18.880 --> 0:19:21.600 an AI Act like they've got in the European Union, 0:19:22.240 --> 0:19:24.800 or just what the country's position should be. And he's 0:19:24.800 --> 0:19:27.400 obviously no longer in the cabinet. He was replaced by 0:19:27.440 --> 0:19:30.400 Tim Ayres, who spoke at our AI summit, and it's 0:19:30.480 --> 0:19:33.760 perhaps unfair to compare them both, but it was interesting 0:19:33.800 --> 0:19:37.080 because both Tim and Ed appeared at the summit. So 0:19:37.680 --> 0:19:40.800 tim Ayres had an interview with myself on stage where 0:19:40.800 --> 0:19:42.880 I asked him about whether we should have an AI 0:19:42.920 --> 0:19:47.240 Act and what he thinks about how to create more 0:19:47.280 --> 0:19:50.760 big AI companies from Australia, and he really he sort 0:19:50.800 --> 0:19:54.280 of didn't have any answers yet, and it's maybe understandable 0:19:54.600 --> 0:19:56.160 he's only just in the job, but he was kind 0:19:56.200 --> 0:19:57.800 of saying, well, I've got to go back and talk 0:19:57.840 --> 0:20:00.880 to my colleagues about this in the industry, and where 0:20:00.920 --> 0:20:03.119 he had Ed Husick jump up on a panel next 0:20:03.160 --> 0:20:05.200 and not trying to make him look silly, but he's 0:20:05.200 --> 0:20:08.280 already done all of that consultation saying in his view, 0:20:08.280 --> 0:20:11.080 we do need an AI Act just to give some guidelines. 0:20:11.119 --> 0:20:15.200 He described a Swiss Cheese approach to regulation at the moment, 0:20:15.240 --> 0:20:17.160 which is not going to be helpful to anyone, where 0:20:17.640 --> 0:20:20.639 rules get put in place when something goes wrong, and 0:20:20.680 --> 0:20:23.879 so there was a sense that regulations struggling to keep up. 0:20:23.880 --> 0:20:26.520 There's not widespread agreement with ed music at all that 0:20:26.600 --> 0:20:29.679 we need an AI Act in Australia. There's views that, 0:20:29.960 --> 0:20:32.480 certainly amongst people in need take industry, that in the 0:20:32.520 --> 0:20:35.560 EU it's become too restrictive and actually stops them being 0:20:35.600 --> 0:20:38.040 able to release new products there because they're always worried 0:20:38.040 --> 0:20:40.800 about it breaking the rules. But there is a sense 0:20:40.880 --> 0:20:43.800 at the moment that we have a bit of a 0:20:43.880 --> 0:20:48.960 vacuum in terms of clear direction about how employment policy, 0:20:49.000 --> 0:20:53.000 about how workplace policy, about how innovation policy needs to 0:20:53.040 --> 0:20:55.719 interact to get Australia sort of motoring on the global stage. 0:20:57.119 --> 0:20:59.720 All of this depends on where we are on the 0:20:59.760 --> 0:21:04.960 path to superhuman intelligence. So from using check GPT for 0:21:05.040 --> 0:21:07.800 this and that to massive change in the way governments, 0:21:07.880 --> 0:21:11.320 businesses and people do things. Where are we up to? 0:21:12.320 --> 0:21:16.480 So that's the big question, and the multi billion dollar question. 0:21:16.680 --> 0:21:20.280 There's a term that's taken off recently Open Silicon Valley 0:21:20.320 --> 0:21:26.159 of being AGI pilled and AGI meaning artificial general intelligence. 0:21:26.560 --> 0:21:30.600 That is where artificial intelligence equals the best of humans, 0:21:30.640 --> 0:21:34.480 and then the next step beyond that is the superintelligence, 0:21:34.520 --> 0:21:37.480 which is where it outstrips us in all these areas 0:21:37.480 --> 0:21:39.200 as well. And the phrase that people have talked about 0:21:39.240 --> 0:21:42.080 being agi pilled is a reference to the nineteen ninety 0:21:42.119 --> 0:21:45.120 nine movie The Matrix, where humans could take the red 0:21:45.119 --> 0:21:47.359 pill to wait from the dream and see the real 0:21:47.400 --> 0:21:49.800 world done by AI systems, or take the blue pill 0:21:49.840 --> 0:21:53.480 and stay in their dream and they're nice, comfortable existence. 0:21:53.560 --> 0:21:53.800 Yeah. 0:21:54.080 --> 0:21:57.240 So industry luminary is like Demis Hasarbis, who's the chief 0:21:57.280 --> 0:22:00.840 executive of Google Deep Mind and a real pioneer in this. 0:22:01.119 --> 0:22:03.840 He's been one of the ones that's been I guess 0:22:03.880 --> 0:22:07.719 agi pilled and is increasingly convinced that it's going to 0:22:07.880 --> 0:22:11.359 arrive imminently. Sam Altman from Open Ai has said it 0:22:11.400 --> 0:22:14.000 could be this year. I don't know whether he's walked 0:22:14.040 --> 0:22:15.880 that back. That was a little while ago he said 0:22:15.880 --> 0:22:19.560 that and topics. Dario Amadi said in January that he 0:22:19.600 --> 0:22:22.000 could see a form of AI that is better than 0:22:22.040 --> 0:22:26.199