This is an audio transcript of the FT News Briefing podcast episode: ‘What if AI knows your death date?’
Marc Filippino
Good morning from the Financial Times. Today is Friday, January 19th and this is your FT News Briefing.
[MUSIC PLAYING]
More tax cuts could be coming for the UK. And the conflict in the Red Sea is raising questions about price increases. Plus, you’ve heard about ChatGPT. But what about DeathGPT?
Anjana Ahuja
Just as you can use language and you can mine that language for words to predict words and sentences in ChatGPT, you can use data about life events to predict mortality.
Marc Filippino
I’m Marc Filippino and here’s the news you need to start your day.
[MUSIC PLAYING]
UK Chancellor Jeremy Hunt wants to cut taxes again before this year’s general election. He said yesterday that he’d need more data to make his decision, but that’s the direction he wants to go in. Hunt’s Conservative party is 20 points behind the Labour in the polls. So Hunt is under a ton of pressure to announce new tax cuts before the election. He already announced a £20bn tax cut back in November to try and boost growth. Hunt added that he would not want to fund the tax cuts by borrowing money. That’s how former prime minister Liz Truss’s tax cuts were supposed to be paid for. And we all know how that ended: a total meltdown in the gilt markets and a new record set for the shortest-serving prime minister.
[MUSIC PLAYING]
The big economic story over the past year has been inflation, and that it’s finally getting under control. But the conflict in the Red Sea between Yemen’s Houthi rebels, the US and a bunch of other countries has people concerned over one of the world’s biggest trading routes. Sam Fleming has been looking into the economic impact of the attacks on ships in the Red Sea, and he joins me now. Hi, Sam.
Sam Fleming
Hi.
Marc Filippino
So what percentage of world trade usually passes through the Red Sea?
Sam Fleming
Roughly 15 per cent of global sea trade goes through the Red Sea. You’re looking at 12 per cent of seaborne oil, 8 per cent of grain. So it’s an enormously important shipping lane. There’s been significant interruptions, as you’ve alluded to. Ever since the first Houthi attack, which happened on the 19th of October, we’re seeing mass diversions down around the south of Africa and all the way back up. Container ship traffic around the Cape of Good Hope has increased by more than 200 per cent this year, which gets you a sense of the enormity of the change. So these lead to much longer shipping times, potential disruptions in orders. It’s an important development.
Marc Filippino
And how costly is this for shipping companies?
Sam Fleming
It’s led to an increase in shipping standard freight rates, as you’d expect. If you imagine that they’re thinking for a standard container being transported from China to northern Europe, previously, we’re talking about $1,500. It’s, talking about more than tripling, if more, since then.
Marc Filippino
Do we know what kind of knock-on effect this will have for companies and maybe even, eventually, consumers?
Sam Fleming
The way that economists are looking at this at the moment is that in itself, it’s not something that they’re seeing as a major uplift in terms of . . . It’s certainly an uplift in costs, but not necessarily an uplift in overall consumer price inflation. That’s because there is capacity for companies to absorb some of these increased costs, not least because we’re in a period where we’ve actually seen slowing inflation, so in goods inflation overall. We’re also in a relatively soft period in terms of demand because of the impact of all those interest rate rises that we’ve seen, and thus it might not be that easy for companies to directly pass on these higher costs to consumers. And then this is something that companies should be able to adapt to. Ultimately, we’re talking about longer shipping times, but companies can factor that into their order plans, their production plans, and absorb some of these longer shipping times and therefore higher costs.
Marc Filippino
So, Sam, what are you looking out for as this conflict in the Red Sea continues?
Sam Fleming
I think the main question that people are asking is: does this herald a broader escalation of the hostilities in the Middle East? Are we in a period where we’re seeing more shoes dropping and the conflict worsening and widening? And if you start to see that, then begin to impact, for example, oil shipments and that knock-on into oil prices or gas prices, then I think economists would start to get more concerned about what that could mean. So, for example, if you see a 10 per cent rise in the oil price in the eurozone, that could then lift inflation by 0.4 per cent within a year’s time. So a big rise in the oil price, coupled with these disruptions to commercial goods shipping, could start to become more of an important factor in terms of inflation.
Marc Filippino
Sam Fleming is the FT’s economics editor. Thanks, Sam.
Sam Fleming
Thank you.
[MUSIC PLAYING]
Marc Filippino
I’m sure at one point or another we have all wondered how and when we’re gonna die. Now, I don’t mean like, when is the entire human race gonna go extinct. I mean, really, when is each one of us, individually, going to kick the bucket? It turns out researchers in Denmark have developed an algorithm that can predict a person’s lifespan. Here to talk to me about it is Anjana Ahuja. She’s a science commentator who writes for the FT. Hey, Anjana.
Anjana Ahuja
Hi. Hi, Marc. Thanks for having me.
Marc Filippino
Good to have you on. So let’s start at the beginning. What’s the premise for using new tools like algorithms to track the way someone’s life might unfold?
Anjana Ahuja
Just as language is a sequence of words and sentences, lives are a sequence of events. And just as you can use language, and you combine that language for words to predict words and sentences in ChatGPT, you can use data about life events to predict mortality. That’s essentially it — it’s about sequences. You can think of it as a kind of LifeGPT, or if you’re really morbid, then maybe a DeathGPT.
Marc Filippino
Interesting. So tell me a little bit more about the algorithm that this DeathGPT uses. What were some of the parameters that are included?
Anjana Ahuja
So it’s based on registry data in Denmark on its 6mn inhabitants. And this data set captures things like address, salary, occupation, working hours, injuries, medical visits, schooling. And the algorithm is trained on all that sort of rich, rich data. So it’s scale and detail. And then what happened was, is that the researchers let it loose to see how well it could predict whether people survived.
Marc Filippino
So did it work? I mean, can it actually predict a person’s lifespan?
Anjana Ahuja
It can do better than the next best predictive models. And I think what was interesting is the researchers chose to predict of all things mortality, because that’s what the insurance company does. So when you look at how these models work, it can certainly beat out the actuarial tables and the next best predictive models. And yes, it seems to do better than those.
Marc Filippino
So Anjana, you mentioned that insurance companies are interested in technology like this. I can understand why, but I also assume that there are some drawbacks that might come with this.
Anjana Ahuja
Yes, and I think what’s really interesting is that the researchers themselves are very clear that they don’t want this technology being used by insurance companies yet, because they do point out that even though we have things like actuarial tables, the better we get at predicting mortality on an individual basis, then the harder it becomes to pool insurance risk. And that’s really, that’s the premise of insurance, isn’t it? Everybody pays into a pool. And then the ones who get unlucky just get the payout. And I think because this exceeds anything that’s on the market now, I think certainly the researchers themselves feel that there needs to be more of a public conversation about how the tech is used.
Marc Filippino
Obviously, this is a little scary. We don’t really know how this might unfold and the consequences of it, but are there good things that could come from this?
Anjana Ahuja
That’s such a good question, Marc. I think one of the most exciting findings from this research is that you don’t need to pre-specify to the algorithm what you’re looking for. And when you think about the richness of that registry data, could researchers end up finding correlations that go way beyond what we know already? So, for example, we do know that wealth and life expectancy is correlated. But could you go beyond that? Could you find other correlations which you weren’t even thinking of in the data, that you can increase people’s chances of having a longer, healthier life? So I think from a policy point of view, I think this is really, really interesting work.
Marc Filippino
It is, but I don’t know that I would use it. Would you?
Anjana Ahuja
No. I mean, it’s a tough one to ask because I think people felt like that at the beginning of genetic testing. Even if your life story is predictable to some extent, ChatGPT doesn’t get everything right. I think we still have a sense that we control our own destiny and we can make our own decisions. We can defy the odds to some extent.
Marc Filippino
Anjana Ahuja is a contributing writer on science at the FT. Thanks, Anjana.
Anjana Ahuja
Thank you.
[MUSIC PLAYING]
Marc Filippino
You can read more on all of these stories at FT.com for free when you click the links in our show notes. This has been your daily FT News Briefing. Make sure you check back next week for the latest business news.
[MUSIC PLAYING]
The FT News Briefing is produced by Kasia Broussalian, Sonja Hutson, Fiona Symon and me, Marc Filippino. Our engineer is Monica Lopez. We had help this week from Manuela Saragosa, Josh Gabert-Doyon, Sam Giovinco, David da Silva, Michael Lello, Peter Barber and Gavin Kallmann. Our executive producer is Topher Forhecz. Cheryl Brumley is the FT’s global head of audio, and our theme song is by Metaphor Music.
[MUSIC PLAYING]
