Next top model: GPT-5 and its AI rivals

0:00With no fees or minimums on checking accounts, it's no wonder the Capital One Bank Guy is so passionate about banking with Capital One. If he were here, he wouldn't just tell you about no fees or minimums. He'd also talk about how most Capital One cafes are open seven days a week to assist with your banking needs. Yep, even on weekends, it's pretty much all he talks about. In a good way. What's in your wallet? Terms apply. See CapitalOne.com slash bank. Capital One N.A. member FDIC.

0:32This intelligence podcast is sponsored by Back Market. the leading global marketplace for refurbished tech. When the kids need phones, laptops, or tablets for school, buying refurbished tech makes sense for families trying to save money without sacrificing quality. At Backmarket, tech costs up to 50 % less compared to new. All devices have been professionally refurbished and come with a one-year warranty and 30-day return policy. Start saving money. Visit Backmarket.com or the Backmarket app to shop now.

1:05The Economist

1:12Hello and welcome to The Intelligence from The Economist. I'm your host, Jason Palmer. Every weekday, we provide a fresh perspective on the events shaping your world.

1:26For a long while in Japan, after the bursting of the asset price bubble in the 1990s and a long era of deflation, investment just didn't look smart. With inflation now on the way up, all that is changing. Across the board, Japanese people have investing fever. And here are some words that until recently didn't really go together in America. Indie Wrestling League. The big purveyors of the country's fakest sport are on the ropes, and lots of smaller, frankly bonkers outfits are drawing dedicated crowds.

2:04But first...

2:12For nearly seven decades of work by artificial intelligence researchers, not a whole lot happened. There was even an AI winter when many thought the field was just dead. Boy, did that change quick. In 2017, Google researchers came up with something called a transformer. No, not that kind. It's a mathematical thing. The next year, a little American company called OpenAI had turned that thing into a generative pre-trained transformer. GPT for short. You see where this is going. In 2022, out came chat GPT, and the world lost its collective mind. Suddenly, not even three years later, AI has gone fully, pervasively global, and OpenAI is still at the frontier.

2:59last week with GPT-5. It can write an entire computer program from scratch to help you with whatever you'd like. And we think this idea of software on demand is going to be one of the defining characteristics of the GPT-5 era. It can help you plan a party, send invitations, order supplies. It can help you understand your healthcare and make decisions on your journey. It can provide you information to learn about any topic you'd like and much more. This is an incredible superpower on demand. Sam Altman, OpenAI's boss, is, as the field demands, a bit breathless. He's got plenty of competition at home and abroad.

3:36So where is it all going on, and where is it all going? GPT-5 is the new AI model from OpenAI. It's the first understandable model number since GPT-4 two years ago, and it is the best AI in the world for now. Alex Hearn is the economist's AI writer. GPT-5 is exactly as good as we expected it would be, which means if you're looking at the graphs of AI capability that veer up and to the right, it takes us another step closer towards this artificial general intelligence that is due 2027 or so. So first off, talk me through GPT-5, what it does. What are the new features? What are the new lights on the dashboard?

4:21It is an incremental upgrade. And I think that's a problem for open AI. GPT-5 is roughly two months better than the last frontier system OpenAI put out roughly two months ago. In the world of AI development, two months of progress is meaningful. And GPT-5 does better on a lot of the benchmarks that OpenAI care about. It's smarter at solving coding challenges. It has a greater general knowledge of scientific questions. it, according to one internal benchmark that we can't externally verify, is very, very nearly as good at expert level knowledge work as a panel of experts that OpenAI compares it to.

5:04It beats them on about 48 % of tasks. With GPT-5, now it's like talking to an expert, a legitimate PhD level expert in anything, any area you need on demand that can help you with whatever your goals are. All of those things are great, but they are incremental. And when Sam Altman, OpenAI's boss, was presenting GPT-5 to the world, he spent a lot of time talking about the vibe, about the way this just feels better, smarter. I suppose it's a fair thing to focus on if wider adoption is the key, the goal for a lot of these companies, though. I mean, what are people saying about GPT-5? It's got mixed reviews on the ground.

5:41Some of that is because they've introduced a new way of using GPT-5, which they're calling a unified model. Until now, you had to, if you cared enough, go to this obscure menu, the model picker, and effectively tell OpenAI how much you wanted the AI system to actually think about your question. You could pick O3 Pro, where it would sit silently and think for 15 minutes before giving an answer, or you could pick 4.0 and it would just blurt out whatever it thought very quickly. GPT-5 is supposed to root those questions to the best way of tackling them possible, but a lot of users have complained that what it actually does is give you a glib instantaneous answer to some questions that deserve to be thought about for longer.

6:23But the loudest people complaining about this have a very different problem. They're people who effectively had a relationship with GPT-4-0. That relationship for some of them was that they used it as a therapist. For others it was more literal. There's whole communities of people who have, over the course of months and months of conversations with their chatbot, started treating it like a friend or in some cases a romantic partner. They view the release of GPT-5 as effectively a personality transplant for a digital human that they got to know and love. Putting aside the sort of sociological and indeed very personal here, you said at the start that this is pretty much where we would expect GPT-5 to be.

7:03What do you mean by that? One of the top level measurements of AI quality that I've been tracking for a while comes from an independent group called META. That's META with an R at the end. It stands for model evaluation and threat research. They've been tracking for some number of months now the most complicated tasks that AI models can solve with 50 % accuracy and how long it would take a human being to solve that same task. And so in 2019, GPT-1, the very first version, could achieve 50 % accuracy on tasks it took software engineers two seconds to do. A year later, it could do tasks that took software engineers around eight seconds.

7:41In 2023, GPT-4 could tackle tasks that took around four minutes. That trend line is a doubling of complexity of duration of tasks around every 200 days. And GPT-5 is on trend. It can handle tasks that would take a human software engineer around a little over two hours to do. We've constrained ourselves so far to discussing OpenAI and its models. Is that in line with what other companies in other countries are doing? OpenAI is regularly at the frontier, right? It frequently has the best model in the world. But this 200-day doubling in capability issue, if you're scared of AI getting too good too quickly, is reflected across the industry.

8:25And that's not just the American industry. Chinese AI labs have also been bringing out AI models that are just better and better at a frankly astonishing pace in China, particularly because they've been doing it backwards and in heels. They have no access to the absolute top flight AI chips made by NVIDIA. Even the second tier chips that were initially made for the Chinese market were put under export controls from April to July until the US government lifted those restrictions. In the last month alone, there's been three major releases from Chinese apps, Zed.ai, Moonshot.ai, and Alibaba, a huge Chinese tech company.

9:05And we're all still waiting to see the next thing that comes from DeepSeek, the Chinese startup that burst onto the scene in January this year with DeepSeek R1. It's clearly developing DeepSeek R2, hasn't released it yet, And if they maintain their quality, the new world number one could come from China, not America. And thinking about that line that describes the doubling of capacity, let's look a little further ahead on it to AGI. This plainly puts us on a path there. Are there any roadblocks in between? I think it's more the other way round, right? It's a very naive analysis. This is a line on a logarithmic chart.

9:44But if the line stays in the same shape and goes in the same direction, well then you end up with AI models being able to do a task that would take an entire working day by 2027, an entire working week before the end of Donald Trump's term in office let's say. That's the sort of threshold at which people start talking about AGI, artificial general intelligence, an AI system that can do all the work a human being can do at least in the knowledge sphere. That, if AI systems get there, would reshape the entire global economy. And so the question isn't, are there any obvious roadblocks lying between here and there?

10:23It's, do we have any reason to expect the progress that we've got so far to slow down? AI is defined by the scaling law. You just throw more data, more compute in a big bucket and stir it with a stick and it learns to do PhD level maths. We don't yet know whether ever bigger buckets with ever more data and ever more computing power will get us to this general intelligence. There may need to be more fundamental breakthroughs. But the incredible thing about the AI industry is it's got us from 2018 to here without really needing to do any fundamental science. To repeat this insane analogy, we've just made the bucket bigger and bigger and bigger.

11:04Alex, thanks as ever for joining us. Thank you for having me. Alex also sat down with my colleagues over on Babbage, our weekly subscriber-only show on science and technology. They really got under the hood, a bonnet, of AI efforts in China. There's now a huge ecosystem of Chinese AI labs, all learning from each other, all cross-hiring with each other, and releasing new frontier models that have changed what we think about Chinese AI. Look for Babbage tomorrow and every Wednesday, wherever fine podcasts are sold and traded.

12:04API checks. It doesn't just help buyers find a car they love. It helps schedule a test drive, get pre-approved for financing, and estimate trade-in value. Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One. Hi, this is Becca Moore from For the Girls, and this podcast is sponsored by Aerie, where you can find the right bra in 59 sizes, soft fabrics, and good fits. If your bra strap is digging into your shoulder right now, it's the wrong bra. We get so used to being in pain as girls that we forget it doesn't have to be this way. Your clothes should not hurt.

12:35I got to the point where I stopped wearing bras altogether and now my life has changed you guys. The only bras I have in my dresser right now are Aerie and I'm not kidding. Aerie's on a mission to make you feel good and that starts with your bra. Your bra should hold you in not back. Visit aerie.com to find the right bra for you.

12:58When Ponchiyo first started investing in Japan back in 2010, he said the practice had a shady reputation.

13:09But attitudes towards investing have changed a lot in recent years. Moeka Iida writes about Japan for The Economist. Gradually, the base of people investing in Japan has grown. It's really not difficult to spot the change. If you walk into a bookstore, you see entire sections dedicated to financial guides. And if you go on the train, you see a lot of advertisements about investing seminars. Now age 31, Ponchiyo has a YouTube channel with almost 500 ,000 subscribers. Hello everyone, this is Ponchi. In this video, we will be explaining how to start a new Rakuten Securities account. Ponchiyo is just one of the many online influencers here in Japan.

13:48He makes tutorial videos on how to build a portfolio or how to open a brokerage account. He says that these days, especially young people are realizing that it's wasteful to leave your money sitting in savings. So Moega, what's behind the change here? Why the change in attitude? Beginning last year, the Japanese government overhauled the NISA scheme, which is a tax-free investing scheme modeled on Britain's ISA. and this has prompted a very big shift in how Japanese people hold their assets. So at the end of 2023, over half of Japanese household assets were held in cash and bank deposits, which is much higher compared to other countries.

14:29And the NISA scheme turned out to be more successful than a lot of people anticipated. Retail investors have opened 5 million new accounts since the NISA scheme was overhauled. And also earlier this year, the total amount of money invested in NISA reached 59 trillion yen, which is around 400 billion dollars, hitting the government's target three years ahead of schedule. Politicians have been very keen to promote this shift as well. If we can get even a portion of this huge amount of cash that is sitting in people's savings into companies and into investment, that could have big implications for the Japanese economy.

15:04But why are people responding so much? As you say, there hasn't been a culture of investing. people thought even that it was shady. Why have they been so enthusiastic about the NISA? So the reason why people haven't been so enthusiastic about it was partly because of the burst of the Japanese bubble economy in the early 1990s, which made people very wary of risk assets. But the younger generations don't have this trauma in their living memory. And they're also living in a very different economic environment compared to older generations. So as I mentioned, Japan had been in a deflationary environment.

15:35So that made it more rational to keep your money sitting in savings, but we start having inflation. So that means if you have cash sitting in your savings, the value of it is going to decline. So there's more incentive for people to put their money into these risk assets. And then on the other hand, you see these efforts among the private sector or companies as well to attract more retail investors. Companies have been working on becoming more shareholder friendly. So in recent years, we started seeing more companies doing share buybacks and raising dividends. And that has made retail investing more attractive.

16:10You mentioned that the sense of risk is different among the young. What about the older people in Japan's economy who play such a strong role? Older generations are more cautious of investing. They think it's dangerous. They have this trauma of the bubble burst, but they are the people who have more savings as well. So the government has been eyeing these older people's bank deposits and have been trying to push older people to invest as well. So now the government is considering a scheme called Platinum NISA. In the current scheme, monthly dividend funds are excluded, but now they're thinking of allowing this for investors over age 65, and that would make it easier for pensioners and older people to invest while drawing down assets.

16:55So this all sounds win-win-win. It's good for older people, it's good for younger people, and it's going to give the economy a boost. Yes, so it seems like the boom is here to stay and that would have a very positive effect on the economy. But there is something that people have been criticizing. People here complain that too many retail investors are investing abroad and not investing enough in Japanese companies, which people say similarly about Britain's ISA. When it comes to individual stocks, Japanese investors are mostly investing in Japanese companies, but the picture is very different when you look at mutual funds.

17:34Some estimates suggest that as much as 80 % or 90 % of NISA money that is invested in mutual funds is actually flowing abroad. Beyond the numbers, I also sense this just talking to retail investors. Investing in NISA has been such a hot topic. When people talk about what to invest in, their default purchases are usually the S &P 500, or there's an index called Oregon, which is short for All Country, which is a tracker for global equities, but mostly American ones and not so much Japanese ones. Some people argue that this is basically capital flight. It's good that people are now investing and there's an investment culture, but Japanese companies are not reaping the benefits.

18:17That might be annoying for policymakers makers in Japan, but maybe it's a sign that Japanese investors are demonstrating wisdom beyond their years. It's always good not to be too exposed to your home market. Thanks very much for your time, Oika. Thank you so much for having me, Jason.

18:53A man wearing an inflatable eagle costume introduces the main contender. He is the great US a-hole, dressed in cowboy boots, a shirt depicting a bald eagle, and a backwards Stars and Stripes baseball cap. Daniel Knowles is our Midwest correspondent.

19:15Ben Camps' opponent, Canadab, A polite father in double denim and a Canada jersey to electronic strains of O Canada.

19:28You know, in the time Canada has come out for your super wrestler super shows, it's true. I've rattled a few cages, eh? As the two warm up, USA Hole shouts out an insult. Bartender, I need some ice And the crowd of about 500 people Many of whom addressed and supported their characters Erupts in boos Super Wrestlers

20:03Super Wrestlers is an indie wrestling league in Chicago It's been going for a few years And this was its 10th edition Super Wrestlers They describe themselves as Earth's strangest pro wrestling show, and I'm pretty sure there's no doubt about that. Besides USA Holt and Canada, the recent show featured Barrowsmith, Stephen Flo, and the pubic noose, wearing an antler hat and a large wig over his groin.

20:38Pro wrestling has long been dominated by the WWE, which had revenues last year of$1.4 billion. Recently, the franchise has been stuttering. SmackDown, the weekly show, has seen lower viewership this year. And WWE Raw, which is a second show, has struggled with a move to Netflix. That revenue growth has tailed off, and All Elite Wrestling, the biggest competitor, has been eating into its market share. Indie wrestling shows are booming, though few are as silly as the super wrestlers.

21:16Game Changer Wrestling, a hardcore wrestling league which started in New Jersey, now puts shows on all over the world. Its appeal seems to be a lot more violence.

21:34Or there's the International Wrestling Cartel, which is an indie league that came out of Pittsburgh and is often seen as a training ground for new wrestlers elsewhere.

21:47Imports are thriving too, such as New Japan Pro Wrestling, which is shown on AXS cable TV channel.