The Real Opportunity Behind Underhyped AI

00:00:07 Mark Smith
Welcome to AI Unfiltered, the show that cuts through the hype and brings you the authentic side of artificial intelligence. I'm your host, Mark Smith, and in each episode, I sit down one-on-one with AI innovators and industry leaders from around the world. Together, we explore real-world AI applications, share practical insights, and discuss how businesses are implementing responsible, ethical, and trustworthy AI. Let's dive into the conversation and see how AI can transform your business today. Welcome back to AI Unfiltered. My guest today is from Wellington in New Zealand. Oliver, welcome to the show.

00:00:51 Oliver Hartwich
Great to be with you, Mark.

00:00:52 Mark Smith
I'm really looking forward to the discussion we're going to have today. But before we start, can you tell me a bit about food, family, and fun, and also what brought you to New Zealand?

00:01:02 Oliver Hartwich
Food, family, and fun. Okay, my favorite food, of course, is decent German bread because I am German by birth and by background. Family, I've got a wife and a 12-year-old son. and fun probably also a bit of German engineering. I drive a very nice German car. And what brought me to New Zealand, my current job with the New Zealand initiative. So I got this job 13 1/2 years ago to run a new New Zealand think tank. And I built it up together with my colleagues, of course, over the past 13 1/2 years. And now it's my second baby.

00:01:35 Mark Smith
Nice, What's your first baby?

00:01:39 Oliver Hartwich
My first baby is still our son, of course.

00:01:41 Mark Smith
Yeah. Naturally.

00:01:44 Oliver Hartwich
I don't think he would be wanted to be called a baby anymore, but.

00:01:48 Mark Smith
Yeah, now at 12, right? Yeah, I read an article of yours some months ago and it grabbed my attention. One, that the article was so timely and coming out of New Zealand fascinated me because I feel like as another, you know, somebody living in New Zealand myself, that with the great advancements happening in the AI space, What are we doing down here in our part of the world? So with that in mind, what's top of mind for you right now when you think about AI?

00:02:24 Oliver Hartwich 
Well, I'm a bit of a geek. I love AI and I love everything technology full stop. And so I followed the development of AI for many years. And what worries me a bit is that most New Zealanders actually, I think we can take this beyond New Zealand, most people. really don't quite know, don't quite understand what's happening. They might be loosely aware of tools like Claude and ChatGPT and Gemini. They might have even tried them occasionally, but then they probably wouldn't have got fantastic answers straight away and concluded, it's just a bit of another hype. There's nothing in it. It's probably just overblown, right? And I, of course, completely disagree because I think the potential is enormous. And we're really just at the beginning of this curve. And we are, depending on who you ask, between maybe 5 and 15 years away from artificial general intelligence.

00:03:22 Oliver Hartwich
So I think the future is going to be even more exciting. And so the discrepancy really between what is already possible, what is going to be possible, and what people think is possible. That is what really worries me.

00:03:35 Mark Smith
You've got people like Eric Schmidt, the former CEO of Google. Some would say he's overhyping where AI is going. Same with Sam Altman, Dario Amadei from Anthropic. The folks at Gemini, you know, at Google seem to be a bit more pragmatic, I feel, around timelines of things like AGI. Do you feel AI is overhyped? for its potential impact on humanity.

00:04:07 Oliver Hartwich
I think if anything, it is under-hyped. And you just have to look at what's already possible, what's already happening. So the team at DeepMind have now concluded investigations into all the possible protein forms in the world. All, I think, what is it, 200 million of them. And it took them about a year to compute. Previously, we thought if you put really all the world's PhD, qualified scientists on it, this task would probably take a few billion years. So this is already possible. This is possible with an AI that is not AGI yet. So how can this be overhyped when we can already do things like that? And can you imagine where this might take us over the next few years?

00:04:52 Mark Smith
One of the things that you said in that article was that and probably sheer good luck has protected New Zealanders and even to some degree Australians from potential bad things that have happened in the world based on our geography and remoteness of things. When you think of that in the context of AI, and if you like, even the maturity that we have here in the Southern Hemisphere for AI, is it going to be something that's going to, we're going to wake up and go, Why weren't we at the table? Why weren't we involved? Will we have missed opportunities potentially and that our remoteness won't protect us from?

00:05:33 Oliver Hartwich
Yes, because certainly for AI, geography doesn't matter anymore. You can participate in the AI game no matter where you sit on the planet. What's actually preventing us from participating is not geography, but culture. So when I look at Australia and New Zealand, I see two countries that are actually quite comfortable. They really don't want to be at the cutting edge of AI. They are quite reluctant. They're perhaps a bit skeptical when it comes to their cultural openness to novel technologies. And you can see this actually in opinion polls. So when you survey Australians and New Zealanders, Are you excited about AI? Do you think that the technology will bring more benefits and problems? And Australians and New Zealanders are kind of at the bottom when it comes to that list of excitement. They don't really expect much from the technology. They're not that really forward-looking. They think maybe there are some problems associated with it and they're quite cautious. They are also then conversely very willing to explore regulations of AI. So that's the flip side of it. So on the one hand, they don't really know too much about it. They're not very excited, but actually they're quite willing to give regulation a go. You look at these surveys then and you find that in other countries, it's the exact opposite. And I'm talking about especially low and middle income countries. So your Turkeys and Egypts and Nigerias, they are the complete opposite of Australia and New Zealand. They are super hungry for AI because they see it as a technology that can help them leapfrog the West. And therefore, consequently, they are not that interested in regulation because they actually just want to use it now and they want to get all the gains from AI that this technology has to offer them. And only then as a second thought or maybe a third, they would think about, okay, it doesn't need any regulation whatsoever. So that worries me because I think these other countries will be much faster getting on the AI ladder and exploiting it for their advantages than Australia and New Zealand are.

00:07:34 Mark Smith
What's the opportunity for us in the Southern Hemisphere? As in, what is the opportunity they're overlooking in your mind?

00:07:40 Oliver Hartwich
I think the opportunity for us in the Southern Hemisphere is exactly the same opportunity for everyone on the planet. AI can just get a lot done and a lot faster done. than we could ever hope to achieve with conventional means. So when it comes to condensing information, when it comes to actually being creative, finding new ways of solving problems, and actually just simply becoming more productive, I think AI is a wonderful tool. And it is a general purpose technology, I would say, because it can be applied to all sorts of things. It can work from medicine to energy to journalism, to the arts, to culture. There's practically no sector of the economy that would be immune from the change that AI offers. And these general purpose technologies are quite rare. If you really think about how many of these general purpose technologies have we seen in the past, say, 200, 150 years, I think of perhaps the steam engine. I think of electricity. I think of basic computing. I think perhaps of the internet then as a general purpose technology. And AI really stands in that long line of really seminal technologies that have completely changed the way we live, that changed the way we do business, that changed the way we do science. And I think there is a danger that we're missing the board because this will change everything in the world in decades to come.

00:09:06 Mark Smith
Are you hopeful for us as a people, as a culture, you know, particularly in New Zealand, or are you more, like if you could take a crystal ball and look five years into the future, do you think we would have embraced and benefited from AI? Or do you think, based on your 13 years of being in New Zealand, that, you know, we have Rocket Labs, we've got this one unique, you know, recent times, and we've got zero, you know, from an amazing companies that have made it to world stages. What do you feel about the future here?

00:09:46 Oliver Hartwich
Well, as a native German, I'm always optimistic and hopeful, of course. And of course, I've got a great sense of humor. But no, I would put it in different terms. When I look at the state of the world today, I'm deeply concerned on a number of measures. So, I mean, geopolitically, we have never seen a world that is more dangerous than the one we have today. When I look at the economy around the world, actually, I see extremely worrying signs about increasing debt, about the fracturing of global commerce. So there are so many reasons to be really pessimistic about the world today. But the one thing that still gives me hope is technology. because I believe technology has a wonderful potential to solve some of our problems. Not all of them, of course, but it could at least make us a lot more productive, which would also help, by the way, in an aging world population. So for me, this is the one glimmer of hope I have, that technology might solve some of those most pressing problems that we see all around us. What worries me then is just that we might miss the boat, that we might not even see this as an opportunity and immediately fall back into our old habits and patterns of just being pessimistic, of actually worrying, of asking ourselves, so where could this all go wrong rather than actually seeing it for what it is, an enormous opportunity for us to do better.

00:11:13 Mark Smith
In what practical ways would you advise people in the skills that they need to acquire let's take whatever their role currently is. They might be lawyers, accountants, they might be electricians, more in the blue collar work side of things. What skills should all of us fundamentally be learning to really thrive with this new technology, to lean in and let it really enhance our lives? And I'm not just talking about make more money or even create a friction-free world because I feel friction is an important part of our lives for growth. But how should people, what skills should people be gaining now and that you feel are critical in such a way that allow us to move ahead as AI evolves, that we'll evolve with it. We'll augment our human skills with all the benefits that AI brings at an individual level.

00:12:20 Oliver Hartwich
I think what you really need is a certain degree of curiosity and a willingness to experiment. Because AI is not a technology that you just adopt one day and then you get productive the next. It is a learning curve. And it might be a J-curve for productivity, meaning initially you might even see a bit of a drop in your own productivity once you're going onto AI because everything is different. And it will take a while until you master how to prompt. But then you get out of the first dip on the J, and then you get into a very steep rising curve because your productivity will explode. So what I think people need to do is actually, they need to go into the technology without any prejudice. They should just try to experiment how it works. They should try out different techniques. I actually think everybody will realize that they have probably slightly different ways of prompting. There is no one method that fits all. It depends a bit on your own background. It depends on your background knowledge. And you will find the way that works for you. When I look around my colleagues and we are all using it to various degrees, I think every colleague has a slightly different way. You have to find the way that works for you. The other thing I'd say, will work better with AI generally on issues that you understand yourself. If you are a great programmer already and you know how to code, then I think AI can make you an excellent programmer, but it would not necessarily make you an excellent newspaper writer. Whereas for me, I have probably written, I believe, around 1500 columns or so over the years. I know how to craft a good newspaper column. And I've done this before AI came along. Now I can use AI to actually craft better columns and probably faster columns than I could have done before. But I can only do so because I've done it before. And so one thing that I would always stress is actually that AI does not supplement or replace previous education. You still need deep education. You need deep knowledge. The better educated you are, the more experienced you are, the more you will get out of AI. And incidentally, that's the one thing perhaps that worries me the most, that for juniors, in whatever profession, AI can be very tempting because it will give you a ready-made answer and it will solve your problems. But you as a young researcher or young programmer would never actually be able to do this on your own without AI in the 1st place. And therefore, I think that will also limit your effectiveness in using AI further down the track. So I think AI is something that works really well for mid-career professionals who already know what they're doing, rather than for pure beginners for whom this may not be the best way to start your professional career.

00:15:16 Mark Smith
So let's talk about education for a moment. When you look at your son, 12 years old, and the school curriculum at that age, We know often it takes a long time for the education system to catch up. One of the things in my career of being in computing, I was always surprised at how far behind I felt even things like programming, et cetera, was at university until one day I was told by an educator that when a course is written, it often has to get seven years of return before they'll rewrite a curriculum, that type of thing, or adjust it. And this was like 15 years ago when I had this discussion. And I'm wondering, will the education system adjust to a world where the jobs that are going to be in the future are nothing like what our education system was perhaps designed for now? Will it iterate at a rapid enough pace to help this next generation not just get through primary and secondary school, but then on into university. But just at this point, can you just think about that primary and secondary age? Will the education system evolve quickly enough and how much does it need to change for the new world that we're moving into?

00:16:36 Oliver Hartwich
It may surprise you that I think the answer to preparing for the world of AI may lie in the very distant past. I'm a great fan of the Prussian education reformer Willem von Humboldt, who was education minister in Prussia, Germany, more than 200 years ago. Because I think actually the basic lessons that he taught us still apply today. In Humboldt's world, it was all about equipping every student with deep knowledge across a broad range of subject areas. And Humboldt always said, this is not to prepare you for any specific job, because that's not the role of schools. The role of school is actually to give you enough of a toolkit so that you are equipped yourself and to build your personality so that whatever job comes your way in the future, in future decades, really, you will be able to deal with that. This is an education philosophy from the early 19th century, but I think it has a lot of lessons still for us in the 21st. Because technological change is so fast, we can't even anticipate what our students will be doing in, say, 10, 20, or even 30 years' time. But what we can do is we can equip them with the right kind of mental toolkit so that they can deal with them. And so I think actually, paradoxically, the answer to a rapidly changing technological world is to have a very traditional school system.

00:18:08 Mark Smith
Yeah, that's interesting. It brought to mind a book I read called Range, which the book is about that the world we're going into is going to be for those people that are broad on a whole range of subjects, polymath type, very broad, because they will be able to spot opportunities that we can't define at the moment because we're not that far down that path. So I love that perspective. Just switching gears to universities. I recently ran a training program on behalf of Microsoft across all the universities in New Zealand. And there was seemed to be a conflict between the faculty and the management around the use of AI as part of the education where you had some lecturers feeling that it was absolute cheating to use it and others more like the scientific calculator when it came out that was considered cheating and math and then became everybody uses it as a standard tool nowadays. And one of the conflicts that seem to be happening is that Different educational staff had different opinions, but there was no collective opinion of the university. It hadn't been formed yet. What are your thoughts?

00:19:22 Oliver Hartwich
Increasingly, I think that universities in New Zealand are almost giving up on fighting AI. There was a lot of resistance. They, of course, used plagiarism detection software. They used AI detection software. Increasingly, I think universities are giving up that fight because they know that the software is always a step behind what's possible. And actually, then it comes up with false positives, too. So it tells students, actually, you've plagiarized or you've actually generated this with AI when they haven't. So the software is unreliable, and I think universities are realizing that now. And so that means universities then also kind of have to accept that students are going to use it. The question is actually, can we ensure that they at least use it well, that they know the limitations, that they prompted correctly? And then when it comes to actually assessing whether students understood the stuff they're supposed to learn, well, you might have to go back to oral examinations because there's no other way. Because whatever is written these days and submitted to the university, well, you will never be 100% sure where it comes from. So we might actually have to do oral examinations, we might have handwritten examinations, in big lecture halls under supervision. Yeah, Very traditional methods of actually checking whether students actually understood stuff because everything else is just too unreliable.

00:20:43 Mark Smith
Yeah, interesting. What about careers? You know, I think Xero, and I've thought this for some time, is in a very interesting position because it has become by far the dominant player of businesses, you know, in the southern hemisphere from an accounting perspective. And they progressively over time, Rod Jury and the team there fashioned it that to start with the accountants were on the outside and then they really got the accountants on board and the accountants then moved more and more of their clients to that platform. Have accountants ultimately done themselves out of a job? The potential for AI to be applied to zero, to the ability to understand the tax code of New Zealand intricately, every bit of tax-related legislation, as well as court cases that have run in the past. And then being, fundamentally, and I know banks have been doing this for some time, that if you're an electrician and you have a small business, they're able to model your electrical business based on every other electrician in the country that's also banking with that organization. Something I saw at BNZ some time ago. And I'm wondering, Is Xero set up to really transform that whole accounting space and that an AI accountant could be the way things move forward?

00:22:10 Oliver Hartwich
Well, I think there's absolutely no doubt that Xero is making it far easier for companies to do their accounting. I think it also makes it more intuitive and it makes it more accessible. I'm not sure that the accountants will really go out of business completely. I mean, some of them might retrain and become auditors instead because I think there is no shortage of regulatory complexity when it comes to our auditing needs. And it's actually really difficult in New Zealand these days to find an auditor. We as a company have been going through that because we've been using the same auditing company now for the whole time, which I know is not good business practice because we should actually switch auditors every maybe six or seven years. The problem was actually we could not even find another company willing to take us on because the market is so tight. So yes, you may make accounting a bit more straightforward with AI and with software like Xero. But then again, I think legislators are particularly good at creating new rules, which will then demand more employment elsewhere. So it might just be a case that we're actually shifting people dealing with our accounts from one side of the equation to another.

00:23:19 Mark Smith
Yeah, it's Definitely interesting. For me, it's an area that I thought would be low-hanging fruit opportunity for Xero in the advancements of their business. But businesses get large and they sometimes don't innovate at the state or rate they did in the early days. Time will tell. Do you think there's going to be a dual society going forward where we have, you know, like with the sewing looms and stuff where you have people that are totally anti-technology, And they're going to go, listen, we're going to disconnect. We're going to not get involved. We're going to go offline, so to speak, and not engage in an AI-centric world, if that is possible. I can't think of the word I'm specifically thinking. What's the word for somebody that's totally against technology?

00:24:11 Oliver Hartwich 
Luddite, perhaps?

00:24:13 Mark Smith
A Luddite. That's the term I'm looking for. Do you think we're going to have a dual society of Luddites and then people that are fully embraced in using AI? and all the advantages to augment their own skill sets.

00:24:25 Oliver Hartwich
To a degree, yes, except I think it will become increasingly difficult to be a complete Luddite because it will just be required of you to participate. What I do see, however, is actually that we will polarize society in a different way. Everybody will be using AI in the end, or almost everybody, but not everybody will use it well. And not everybody will use it for productive purposes. And so what we will rather see is actually a society that will fracture between people who really know how to use this for their own great advantage and others who may not even know what they're properly doing. And in that sense, I believe AI is, again, very much like previous technologies. What I mean is when TV became a big thing in the 1950s and 60s, there were some people who only watched soap operas, maybe sports, but light entertainment. And then there were other people who watched the National Geographic documentaries and political stuff and historical stuff. And so in a way, it amplifies what's already there. So if you're coming from a low education background, I'm stereotyping a little bit now, you are more likely to go into the light entertainment category. And so you would basically stay where you are. Whereas if you're coming from a higher education background, you would use TV much more selectively and only watch stuff that will actually enhance your existing knowledge. And so in a way, it amplifies. It makes some punchier, perhaps it will make dumb people dumber and smart people smarter. And in that sense, AI will do exactly the same. If you don't know how to use technology, if you have limited background knowledge, if you have a limited degree of formal education, you are less likely to really get much out of AI. You will not get the most sophisticated applications going and deliver to your advantage. Whereas if you're coming from a highly educated background and you are really capable of using technology, it will boost you. And so that's the kind of polarization I see coming our way. And it will be increased, of course, by the fact that Not every AI is made the same. You can use ChatGPT in the free version, but you can also pay 200 US dollars a month and you get a completely different product. So again, this is another way in which society could fracture.

00:26:51 Mark Smith
Yeah. I liked your metaphor that don't blame the mirror for the reflection.

00:26:57 Oliver Hartwich
Yes, because the mirror is actually our own capabilities. So people who have no idea how to use technology and get dissatisfying answers from ChatGPT, well, they're effectively looking into the mirror. They're seeing their own limited capabilities. But then, of course, they blame the mirror.

00:27:17 Mark Smith
The last question for you is around society in general. There's going to be those of us that can pick up the tools, work with them, and advance ourselves. There's going to be a whole section of society that might not even have access. How do we accommodate and create a pathway for those folks in society?

00:27:38 Oliver Hartwich
That's going to be a massive challenge for decades to come. I mean, to that degree, it's not a challenge already. I think if we wanted to take an optimistic view here, we can expect these technologies to become a lot cheaper over time. So I mentioned the 200 US dollars that you currently pay for the top subscription on ChatGPT. By the way, if you want to be really productive on AI, That would not be your only subscription. You would probably have another one with Claude and another one with Gemini and a few more as I do. Not everybody can afford that unless you're working for a company that actually invests in employees, I think companies should. And so it is quite costly to play at the kind of top edge of the market. But there's a chance, of course, these tools might become a bit cheaper over time. Because as we all know, Moore's Law, computing technology goes up. As we can also see, it becomes cheaper to develop these large language models. And we might not actually need the billions of parameters. We might actually do with a few 100 million and have much cheaper models in the future actually delivering comparable results. On top of that, we know that quantum computing is around the corner. Once we get into quantum computing, again, it's a whole new different ballgame. because then things could be possible, and A, we can't anticipate, things will be possible actually delivering results that we currently get today, but at a fraction of the cost. And so, yes, you are hinting with your question at potential for redistribution to enable people to become part of that new technology. But actually it might also work the other way, that technology over time just becomes more affordable in a way that other technologies in the past have also become more affordable. I mean, in the 1960s and 70s, who would have thought that everybody's got a computer at home or actually in their pockets? And they have become more affordable over time and we have democratized that technology. So why wouldn't that happen to AI as well?

00:29:40 Mark Smith
Yeah. I actually have one last question. It's around energy in New Zealand. And you mentioned in your article, 90 gigawatts of electricity needed in the US, which is equivalent to 90 nuclear power stations. In New Zealand, our energy supply, we've just had AWS say they're spending $10 billion on data centers in New Zealand. Microsoft in the last three years has spent a similar type figure on their data centers, the largest data centers currently in New Zealand. And of course, AI is going to come to those in time naturally as they can get the hardware. Is there a risk with, I mean, I know in Northland, big solar farms and wind farms are going in, but is there a risk that we're going to have an energy issue in New Zealand based on what you see forecasted at the moment and then, you know, demand where a wealthy company will probably pay a premium at the detriment of the individual buyer.

00:30:41 Oliver Hartwich
Well, arguably, we already have an energy problem in New Zealand. We saw this last winter. The energy system is already quite stretched, not least because of some political decisions by the previous government. They're toying with installing battery capacity in New Zealand, which then didn't materialize, but actually held up any other kind of investment. And of course, we had the oil and gas ban, which also had a chilling effect on the entire energy sector. So we've got some political problems actually to overcome. But then again, New Zealand is perfectly placed for data centers because we have an abundance of geothermal. It's the energy under our feet that we could use because it is baseload effectively that you could use for big data centers. So if we wanted to, we could tap that and we don't even have to go down the nuclear path because we have plenty of energy in New Zealand available for us if we want it. The other idea on top of that is, of course, as I said before, quantum computing will be completely different. Currently, what we have is we have an energy system, well, sorry, we've got a data system that requires enormous amounts of energy. But quantum computing is way more efficient. And the question is actually, will we still need all of this energy in the future once we actually move to a different type of compute?

00:32:02 Mark Smith
Yeah. Definitely interesting times ahead. Thanks, Oliver, for coming on the show. Really appreciate your insights.

00:32:11 Oliver Hartwich
Thank you very much, Mark. Great to talk to you.

00:32:14 Mark Smith
You've been listening to AI Unfiltered with me, Mark Smith. If you enjoyed this episode and want to share a little kindness, please leave a review. To learn more or connect with today's guest, check out the show notes. Thank you for tuning in. I'll see you next time, where we'll continue to uncover AI's true potential, one conversation at a time.