AI-pocalypse: predicting the threat from artificial intelligence

Hey, Ryan Reynolds here wishing you a very happy half off holiday because right now Mint Mobile is offering you the gift of 50% off unlimited. To be clear, that's half the price, not half the service. Mint is still premium unlimited wireless for a great price. So that means a half day.

Yeah.

Give it a try@mintmobile.com Switch upfront payment.

Of $45 for three month plan equivalent to $15 per month. Required new customer offer for first three months only. Speed slow. 135 gigabytes of network busy. Taxes and fees extra.

See mintmobile.com hey, Ryan Reynolds here for Mint Mobile. You know one of the perks about having four kids that you know about is actually getting a direct line to the big man up north. And this year he wants you to know the best gift that you can give someone is the gift of Mint Mobile's unlimited wireless for $15 a month. Now you don't even need to wrap it. Give it a try@mintmobile.com Switch upfront payment.

Of $45 for a three month plan equivalent to $15 per month. Required new customer offer for first three months only. Speed slow after 35 gigabytes. If network's busy, taxes and fees extra. See mintmobile.com.

Hello and welcome to the Intelligence from the Economist. I'm your host, Ore Ogumbi. Every weekday we provide a fresh perspective on the events shaping your world. While many American prisons offer education programs, there's only one that allows inmates to study to graduate level. And this initiative is helping them thrive in more ways than just academically. And we pay tribute to an accidental hub for British pop culture, a home for comedians and notably the undersung incubators of rock music. It's the working men's club. First up, though. Lately it seems that the chances of humanity being wiped out has been high on many people's minds. There's the increasing danger of natural disasters from climate change.

People are dying. We are in the beginning of a mass extinction.

There's the threat to the world's population from new diseases as the global COVID 19 pandemic forced us to reckon with.

Nearly 15 million people around the world.

Have died from the impact of COVID.

During the first two years of the pandemic.

And the war in Ukraine has brought back to the fore a fear that many thought had receded. The possible use of nuclear weapons.

President Putin has ordered Russia's strategic nuclear forces to be placed on high alert.

The famous Doomsday Clock, set up in 1947 by the scientists who created nuclear arms is the closest it has ever been to midnight, a symbol for annihilation.

It is now 90 seconds to midnight.

And as if all those weren't enough for us to be concerned about, it seems you've got something else to worry about now too. Artificial intelligence. And forecasters seem to think it's one of the worst dangers that we currently have to face.

A lot of people are worried about AI at the moment.

Arjun Ramani is the global business and economics correspondent for the Economist.

So in May, a group of bigwigs in the AI industry, including the CEOs of DeepMind and OpenAI, two of the most important labs, signed a one sentence open letter stating that, and I quote, mitigating the risk of extinction from AI should be a global priority alongside other societal scale risks such as pandemics and nuclear war.

Okay, that is quite the statement. Quite an alarming one actually. Are they right to be so worried?

Well, it's hard to say. So artificial intelligence as it is right now doesn't really pose these existential risks. But people worry that if progress in the field continues and AI becomes what people called artificial general intelligence or eventually artificial superintelligence that are perhaps more powerful than us, you could imagine why that might pose risk to us, but that's uncertain because we don't know if progress is going to continue first. And we also don't know if these powerful AIs will actually want to do anything bad to humans or whether they'd be designed in a way in which they could Possibly experts are overestimating these risks. It's not just AI. A consistent pattern has been found where experts in the specific field tend to assess their topic to be more risky than an equivalent estimation from a non domain specialist. And that finding is according to a recent working paper by researchers including Ezra Karger, who's at the Chicago Federal Reserve, Phil Tetlock, a political scientist at the University of Pennsylvania and a team of their co authors.

Okay, so that has calmed me down slightly. Sounds like we can take that big alarmist statement with a pinch of salt. How did those researchers of this working paper study this?

Right, so they ask subject matter experts in nuclear war, bioweapons and AI, you know, three of the big areas where you might think they're existential risks to give forecasts. And they also ask another group of what people call superforecasters to make those same predictions. And these are general purpose forecasting experts who have a proven record of making accurate predictions on all sorts of different topics, like geopolitical events like wars, or the financial markets. And these two groups were given two different kinds of disasters to consider. Consider. So as I mentioned, catastrophes and extinction. And to define them, a catastrophe is an event that kills just a mere 10% of humans in the world. For your context, the Second World War is estimated to have killed about 3% of the world's 2 billion people at the time. So catastrophe would be far worse than that, but extinction would go even further. So this is defined as an event that would wipe out everyone with the possible exception of at most 5000 lucky, or maybe you consider them unlucky people. So they were asked to provide forecasts in the likelihood of everything from terminal events like an AI or nuclear apocalypse, to smaller questions such as how fast AI progress might go in the shorter term, which potentially would give you a sense of which types of people are better at predicting short term events, which could then give a sense of whether their predictions about the long term should be taken more seriously, which I think is really interesting.

Are you saying that the subject matter domain experts predicted consistently worse outcomes than the more general purpose superforecasters?

Exactly. So the experts thought there was about a 20% total chance of a catastrophe by 2100 and 6% chance of extinction, which is very high. The relatively optimistic superforecasters gave those events probabilities of 9% and 1% each. So that's for everything, all the different risks, AI, nuclear and so forth.

Okay, those odds don't sound great, even from the relatively optimistic superforecasters. 9% chance of total catastrophe. I mean, I'd buy a lottery ticket on those odds.

Same here. I think you're absolutely right. When you talk to superforecasters, they'll say pretty clearly that yes, they're a bit more sanguine than the experts, but the probabilities of catastrophe and extinction are still very high. So, you know, they were pretty worried as well, to be honest. But there were differences between the two groups on top of these headline numbers, and they were greatest when you look at AI in particular. So the average superforecaster thought that by 2100, AI presents a 2.1% chance of catastrophe and a 0.4% chance of extinction. AI experts massively increased those estimates. So they assigned the two events a 12% chance. That's 12% for catastrophe and a 3% chance of extinction, respectively. So, you know, for extinction, that's nearly an order of magnitude bigger probability. But the most interesting thing is that despite these general disagreements, both groups actually put nuclear war and AI as much bigger worries. Than anything else, and in particular for extinction. Even the superforecasters thought that AI was the biggest risk.

Why is AI in particular such a concern? I mean, it's still pretty far off.

Yeah, it's a good question. So both groups actually thought that AI progress was going to continue. So that's maybe one reason why they rated it as a big risk. But the other big reason is there's kind of a multiplier effect of AI on other risks. Like nuclear weapons. AI could kill humans directly. You could think of killer robots or a virus that gains autonomy and hacks into the world's infrastructure, or it could act indirectly by hasting another threat. So for example, humans might actually use AI to help design nuclear weapons or perhaps more potent bioweapons. And so you would consider that to be a case where AI contributed fundamentally but indirectly to the disaster.

Okay, but Arjun, in light of this multiplier effects, wouldn't it then be sensible to believe the domain experts who are a bit more pessimistic?

Well, perhaps. And they make the point that superforecasters do have less technical knowledge. And, and it's also true that superforecasters, their track records are based on predicting short term events. So we don't actually know how good they are at predicting these long term tail risks. But despite this, it's worth noting that in a lot of other domains, superforecasters have a very solid proven record of outperforming experts on questions from finance to geopolitics. And even though they're perhaps shorter term predictions, the long term is a series of short term events coming one after the other. But the really interesting thing here is the paper's results basically suggest that superforecasters and experts have different models for just how the world fundamentally works. So technological risk depends not just on how powerful a technology might get, like AI, but also critically about how humans will react to it. So if there is a small scale event that's caused by AI, some kind of damage, societies will respond to that. And in my interviews with the superforecasters and some AI experts, it was pretty clear that superforecaster tended to think that these kind of small scale events would prompt a lot of scrutiny and regulation, which could then reduce the risk of bigger problems in the future. Whereas the domain experts tended to think that commercial and geopolitical incentives to use these technologies more and more would outweigh the worries about safety even after some short term harms.

Arjun, what do you think? Whose predictions should we take on board?

I'd put myself a little bit closer to the superforecasting camp, to be honest. And I think the reason why is I think one, there is a lot of uncertainty over how how much faster we can keep on boosting AI progress because you need a lot more computer power and data, and we're running up against various limits on both of those ends. But also because I actually do tend to have a bit more faith on the part of humanity to kind of respond to risks appropriately. If you actually look at what's happening in the world right now, there's a huge wave of interest from regulators on how to manage these risks. And it seems the AI labs are actually pretty open to common sense regulation that could limit these risks too.

Arjun, thank you so much for joining us.

Thanks Ari.

Look, no one's journey is the same.

That's why Delta Sky Miles lets you do it your way. From earning miles on reloads for coffee runs, shopping, and things you do every day to connecting you to new places and experiences, a SkyMiles membership fits into your lifestyle, letting you do more of what makes you you.

It's more than travel. It's the membership that flies, dines, streams.

Rides, and arrives with you.

Every great journey deserves a great story.

And when you have a membership that's as unique as you are, there's no telling how your story will unfold or where that journey will take you to next. Sky Miles is the membership that will be here for all your big and small moments. The membership that's there for every solo adventure or family trip. The membership that comes with the power of partnership from brands you love. The membership that moves with you. Learn more@delta.com SkyMiles hi, this is Christy.

From Back to the Bar. You've probably heard about GLP1 weight loss medications and the side effects that can come with jumping in too fast. That's why I love Noom. Makes getting started easy. Their microdose GLP1 program begins with a smaller dose and gradually scales up based on how your body reacts. The Noom GLP1 microdose program starts at $99 and is delivered to your door in seven days. Start your microdose GLP1 journey today at noom.com that's n o o m.com Noom micro changes big results average weight loss 8 pounds in first month meds and personalization based on clinical need and not available to all individuals. Medications are not reviewed by FDA for safety, efficacy or quality. Pricing based on first month only.

If you enjoy our podcast, I want to tell you about an app from the Economist called Espresso. It offers quick takes on big global stories with daily shot of the biggest headlines. The World in Brief gives you the very latest news and is updated throughout the day. There are five short articles offering quick reads and analysis on critical developments. There's a daily chart, a fact of the day, a quote of the day, and even a quiz. It's a diverse little digest to get you primed for the day in just a couple of minutes. If you already subscribed to the Economist, Espresso is available to you now. But if you don't go to economist.com get espressoapp to find out more.

June in America is graduation season. Over the past three years, I've been to two, both my sons, both from primary school. This one that you're hearing now was more momentous.

John Fazman is our American business and Society correspondent.

Seven men were receiving graduate degrees from the New York Theological Seminary. There were proud parents and teachers. There were speeches, hugs and tears. But the ceremony wasn't held on the seminary's campus in upper Manhattan. It was in a vast linoleum floored auditorium at Sing Sing, New York's maximum security prison. Prisons are supposed to rehabilitate their inmates to prepare them to become productive members of society, but that happens far too infrequently. When states feel the fiscal pinch, education programs in prisons are often the first things to go. This program at Sing Sing is the only one in America that awards graduate level degrees to inmates.

Maya Angelou says, if a human being dreams a great dream, dares to love somebody.

For many of these men, it's their first experience with academic success, the first time they've had their intellect, discipline and diligence publicly affirmed.

This program is really focused on providing an advanced degree to those who are incarcerated.

Lakeisha Walrand is the seminary's president.

We've been here since 1982. We've graduated over 530 men during that time with an incarceration rate of less than 10% over the lifetime of the program and less than 1% in the past 10 years. Years. So this program is really about restoring humanity. It's about letting these individuals know that they do have transferable skills to prepare them for the best life that they can have while they're on the inside, but also prepare them when they come home. Because what we know is that they're going to need so much resources when they get out in order to be able to live the life that they've been called to live.

What do you attribute that low recidivism rate to.

You know what? I think it is really attributed to our professors and the intent of the program, which is not only about providing a rigorous theological education, but it's also about restoring humanity and restoring faith.

There's a little song we used to sing oh, what a beautiful morning oh, what a beautiful day I've got a wonderful feeling Everything, everything going our way Amen.

Nancy Fields doesn't just have a beautiful singing voice. She's a minister and a professor at the seminary and has taught at Sing Sing for years.

The teacher learns from the students what to teach them. So the teacher is learning and the student is learning, and that is what I bring to this experience. So they are given assignments where they engage with one another. They work out on groups. That's like, whoa, because, you know, I've got to talk to you about my stuff. I'm not used to that. You know, we learn how not to trust because you got to watch your back, all of those things. But in here, in this classroom, they get a chance to really share and bond and, wow. They do just that.

Too often. In criminal justice and in my own discipline, criminology, we view the answer to the crime problem as entirely disciplinary. In other words, punitive.

I first met Michael Hallett almost 10 years ago. He's a criminologist at the University of North Florida and a strong advocate for prison ministry programs. When we met, I was doing a story at Angola Prison in Louisiana, and he was conducting a study on inmates religious lives there that eventually became a book called the Restorative Prison.

One of the things we've learned over decades of studying how it is that offenders succeed is that when offenders stop offending. John. It has nothing to do with the threat of future punishment. Most prisoners have been convicted multiple times. The national recidivism rate is 80% after five years. And so we have really, in criminology, in my view, disproven the notion of that we can punish our way out of the crime problem. Simply punishing people is not getting the results we want. If we can get back to rehabilitation, I think that could change the formula. You know, I think the main reason these programs are valuable is that if you bring people religion, it's a great way to smuggle rehabilitation back into the prison system.

I spoke with the mother of one of the graduates.

I think it gave him a really positive sense of himself and a sense of completion. It was. It's a master's degree program. It was challenging. There was. He had not Used his brain and his skills in this way for a very long time. His brain had to heal, heal for a long time first because he'd been a serious drug user, as well as dealing with mental illness. And so when he got on the correct medications, he was then able to heal and get back to his true self.

And you think this program helped him do that, allowed him to use his brain in a way that he hadn't before?

Absolutely. He's a bright guy, but he really, you know, he's 38. It was a long time ago that he graduated from college and that was not the healthiest period of his life. And so this was really the first time he was engaged in serious study. And I think it gave him some a reason to get up every morning and a community that you would continue.

To bless us to be all that we can be to our ultimate greatness. We ask this in the precious, in the mighty name of the one in whom we serve, Let us all say together Amen, Amen and Amen.

What legends Def Leppard are in the middle of a world tour at the moment.

Michael Hahn writes about music for the Economist.

They recently played at Wembley Stadium, the biggest venue in Britain, for the first time as headliners. They didn't start out that way. They began in much smaller venues in the late 1970s in the working men's clubs of South Yorkshire, they turn up promising to play a set of familiar soft pop hits to the uninterested drinkers. But instead the group would play their own hard rock songs, such as Answer to the Master. In Britain, working men's clubs are being great incubators of rock music, and the venues continue to play an important role at the grassroots level of the music industry today. But Def Leppard's tale isn't unique, because working men's clubs have long played a huge and undersung role in the British music industry. Scores of the bands in the new wave of British heavy metal, which exploded in the late 1970s and early 80s, got their start in these venues. It might seem odd to have a band singing about the occult appearing before the bingo. But even groups like Venom got their start in the working men's clubs, despite the risk of being paid off. That means give them half your money halfway through your show to stop you coming back for more. Pretty much every town in Britain has a working men's club somewhere. Their social clubs set up for a fairly specific kind of man. If we're honest, the white working class men, their places to drink, to socialise, maybe have some entertainment. But the club's role as incubators of rock music, I think would have surprised their founders. The first club was opened in Reddish, which is southeast of Manchester, in 1857. Henry Solly founded the working Men's Club and institute union, the CIU, in 1862, and he had a mission statement for what he was doing. The club rooms in every locality will form the strongest counteraction to the allurements of the public house, he wrote. He hoped his clubs would offer uplifting education for their members. However, as clubs, they were exempt from pub licensing laws and they were able to take advantage of a loophole intended to allow the gentlemen's clubs of Pall Mall to serve alcohol to the ruling classes. That loophole allowed the working men's clubs to become late night drinking clubs. A century later though, and working men's clubs have come to represent something very different. In the public imagination, Working men's clubs were the homes of comedians in crumpled tuxedos telling jokes about their mother in law to ailed up patrons. Everyone graced the stages of the working men's clubs, no matter how famous they were. That's how central they were to British people, popular culture. There was even a TV variety show in the 1970s called the Wheel Tappers and Shunter Social Club, which was set in a fictional working men's club.

Are we ready for the next act? Lovely. I don't know. Books are turns here, but sometimes they're good and sometimes they're bad. We've been unlucky tonight.

It featured the kind of actor who would play in the real life working men's clubs. Comedians like Bernard Manning and the Crankies, the magician Paul Daniels and middle of the road musicians such as the Dooleys and Brotherhood of Man. By the late 1970s, the CIU was the largest private members organization in the world. It had more than 4,000 affiliated clubs. It's not the same now, though. After that peak in the 1970s and 80s, the clubs began a rapid decline and that was linked to the changing economy. The areas where the clubs were based, the old heavy industrial areas of the north, suffered a massive downturn of industry. But in the 21st century, they're back at the heart of grassroots music. They're in a slightly different way. A number of the old clubs have been bought up and they've been turned into really good and widely admired concert venues. The Trades Club has become one of the most beloved music venues in the country. And the band snapped Ankles, recorded a live album at the Trades Club. I'm not sure Solyer recognises vision in the Moth or the Brudenelle or the Trades Club these days, but I think there's a chance he'd approve of their mission. They're still providing uplift at reasonable prices.

That's all for this episode of the Intelligent. Let us know what you think of the show. You can drop us a line@podcastseconomist.com and if you're not a subscriber to the Economist, what are you waiting for? Dive in. Get a free 30 day digital subscription by going to economist.com intelligenceoffer the link is in the Show Notes. We'll see you back here tomorrow.