136: Robo Hacks & Dodgy Degrees

Ready to buy a car, a home, or just want to take control of your money?

Your FICO score matters, and 90% of top lenders use it to make decisions.

Check your FICO score for free today without hurting your credit score.

Visit myfICO.com/slash free or download the MyFICO app today.

My FICO gives you the score lenders use most, plus credit reports and real-time alerts to help keep you on top of your credit.

Visit myfico.com/slash free and take the mystery out of your FICO score.

You got the VPN, the encrypted messenger, the private browser.

But if your mobile carrier is selling your data, those apps are just window dressing.

Meet CAPE, the mobile carrier that never collects your name, address, or data.

Because what they don't store, they can't sell.

Premium service, nationwide coverage, CAPE is just $99 a month.

Your first month is just $30.

And with code CAPE33Off, you'll save 33% on your first six months.

Go to CAPE.co, lock down your network, not just your apps.

Page 94, The Private Eye Podcast.

Hello, and welcome to another episode of Page 94.

My name is Andrew Hunter Murray, and I'm here at the Private Eye office with Adam McQueen, Jane McKenzie, and Ian Hislop.

We're here to discuss everything, not everything, three things that have happened in the news since the last issue of the magazine was published.

And we have a little bit of parish business just before we start.

Finally, we're going electric.

We're doing a live show at the Cambridge Literature Festival, and it's going to be on the 26th of April.

Sorry to let you all know about this now because it's sold out in the room but you can buy a streaming ticket if you like listening to this podcast but you think it needs to be more visual.

I need their faces.

Well now's a perfect opportunity to do it.

It's a way to lose listeners when they see what we actually look like.

The disappointment.

You're absolutely right.

So

if you want to buy a streaming ticket to see the thing happening then you can go to private-i.co.uk forward slash pod live

and you will be able to get one there.

I'll redirect you to the Cambridge Literature Festival website.

And the other thing we need to know before we start the episode, for that episode, for the live, what should we call it?

This live.

It feels like it needs a big title.

Well, it's an incredible development in technology.

We move from doing this thing that's recorded on the wireless somehow and put out to doing it in a room, a bit like theatre.

So advances again.

But it's recorded on a wax cylinder.

This is sent out to all of our subscribers.

Well, if you'd like to ask us a question that we will read out and then answer in that room as part of that show then we would love that and you can email them to podcast at private-night.co.uk.

Right, parish business ends.

Let's go straight to topic number one this week.

As we've all ascertained, we're very, very technically proficient, AI.

Adam, various newspapers have been making various announcements about how the future's bright, the future's AI.

Yes, they have indeed.

Yeah, and some of them not making announcements when they should have been, possibly, as well.

So we've written in the last couple of issues of Street of Shame about The Guardian, who have done a big deal with OpenAI, which is probably the most famous of the AI firms, it's the one that's behind ChatGPT, and they have not only handed over in return for some cash access to their archive of stuff, which actually is what a lot of newspapers are doing.

They're not alone in doing this, making their online archives available for a price to train AI models on.

So, other people who are doing similar things include the Associated Press, Reuters, DMG Media, who are behind the Daily Mail titles, and the I.

The FT have done it, News Corp have done a deal with an AI provider as well.

Now, this isn't necessarily a bad thing, because this is about training artificial intelligence models.

So, actually, training them on reasonably reliable stuff.

You won't necessarily say that, but all those newspaper titles I've just quoted, but it's better than just sending them out willy-nilly to kind of steal stuff and crawl all over the internet and take stuff from whatever sites the AI models might fancy without any sort of judgment on it.

So, this is quite a good deal.

If I want to write an article in the style of Owen Jones, and I always do,

this will make it easier for me to do that.

Well, thankfully, you have access to an AI model of some of the nation's best parodists who are quite good at doing this sort of thing.

So, you probably don't need a bot to do that.

But, yeah, that is one part of the deal.

Now, the other part of the deal, which Guardian staff did not find out about until a press release went out announcing it, was also with OpenAI.

The Guardian will be developing new products, features, and tools.

Now, that's caused quite a lot more consternation.

That's not quite the holier-than-thou version of we train them on a proper, reliable model, is it?

No, that sounds a lot more like we get them to do things that could replace journalists, which I think has caused quite a lot of understandable worry and paranoia.

We talked before on this podcast about how the NUJ, the National Union of Journalists, are extremely powerful at the Guardian and they are already involved in negotiations over the use of or possible use of AI and trying to sort out a policy with Guardian management, which essentially, from the journalist point of view, would be don't bring in AI, we don't want it.

There was consternation also, you'll remember the Guardian staff went on strike back in December a couple of times over the sale of the Observer to Tortoise.

And at that point, there was a lot of concern that the kind of skeleton management staff who were left to put out the paper for a couple of days were using AI models to come up with headlines.

So there's a lot of disgruntlement at the Guardian generally, but particularly over this kind of thing.

Now, since the last issue of the I came out, the Independent, remember them?

Online-only newspaper, yep, yep, owned by Yevgeny Lebedev, friend of the podcast.

podcast it's announced that it will be launching something called bulletin which will be like the independent but shorter um ai summaries of stories from the independent for time-poor people people who are just too poor in time to read an entire story and need to have a robot do it first for them instead now they have assured their journalists and readers that everything that's put out in this bulletin will be reviewed and checked by human beings.

But they've also said that it's going to go out with the original bylines on it, which if I was an independent journalist, I think I'd be campaigning to have my my byline taken off something that had been rewritten by a robot and put out.

And if it has to be rewritten by a robot and then checked by a human being, how about a human being writing it?

Yeah, you could call them something like sub-editors, like, you know, those archaic job titles people we used to have on Fleet Street back in the old days.

What justification...

does any organisation put forward?

And I need to know this for when I sack the three of you.

What justification is there apart from just getting rid of journalists?

A lot of it is to do with the fact that it's just the big new thing and it's very, very exciting.

I mean, you hear Kirstalmer talking about AI.

He obviously thinks it's the future.

Rachel Reeves has been talking about increasing efficiency in the public sector and in the civil service by using AI for all sorts of things.

I mean, you run into an immediate problem, which the Independent have already identified.

You also do have to have some humans looking over this stuff to make sure it's not introducing mistakes, which is the experience of the use of AI in journalism already.

So one of the companies has been using it for a while is Reach PLC, who are that publishing BMOF, who published the Daily Express, Daily Star, the Daily Mirror, and over 100 local papers as well.

And they have this tool called, and this is a nice historic reference, it's called Guten after Gutenberg, who you remember, the inventor of the printing press who made the mass media possible back in, where was it?

15th century, 16th century?

A lot of scribes redundant.

All of those monks suddenly not having to illuminate manuscripts.

Yeah, they were probably up in arms as well.

But they've been using it a lot, and that's something that essentially rips pieces of copy that have been written for one reach outlet and can republish them across a lot of other ones.

So, giving them, in theory, a sort of a new geographical nosing on it, that you know, it's a story about something that's happening in Liverpool, but you resell it to people in Birmingham or things like that.

But the problem that I've heard from people within REACH with that is that it has a tendency to work kind of like a thesaurus.

And you can't just pluck words from a thesaurus without some human involvement.

You know,

you need to make sure they are the appropriate word, not the inappropriate word.

So it will tend to introduce one of the things I've had cited, and this is anecdotal, is that it can't quite tell the difference between category B drugs and category A drugs, which is quite important in court reports of people who are being charged with selling drugs.

And even that on occasion, and I've no idea whether these actually made it to publication or not, whether they were spotted, it was introduced the names of completely erroneous crown courts and things.

So, you know, quite important details, but it just muddles up in the way that a not entirely trained and still quite experimental technology might be expected to.

There was something about a boulderizing software that had taken the name of the Enola Gay, the US, the plane that was used to drop one of the atomic bombs and renamed it the Enola Homosexual.

That was

well, it is what it reminded me of.

Actually, I remember Jane, you writing these stories years ago about when spam filters first came in in their untested and not entirely reliable form.

And I remember you did a story about the Horniman Museum in South London.

Yes, any organisation that had a rude word anywhere in its sort of title was being affected, such as ask and Thorpe Town Council.

Are we assuming that AI will sort all this out and that within a year or five minutes or whatever the time scale is with AI, that it will be perfect and it will essentially no longer need humans.

So the choice will be, do we want humans to do anything?

Is that the dilemma?

I think that's the fear certainly yeah yeah I mean these things do get better that's that that that that is a problem of 15 years ago isn't it the one we gave the example there of the spam filters they're still not entirely reliable but these things do

and I'm just going to drop in the word algorithm here as I understand what I mean but they do they do learn things and they do get better at these things and in that sense making you know sort of reasonably reliable newspaper archives that have some sort of authority to them available for the training models rather than as I say just the sort of nonsense that's churned out on various dodgy websites.

A very good story in the latest edition of the eye, which I thought was really interesting, which is about Russian use of AI for propaganda.

And the idea that actually, because of the way that these models train now, you don't even need to get it out there and read by lots and lots of people.

You can just, through very clever search engine optimization and kind of bumping up your fake news websites,

persuade the bots that they are reliable and then the bots crawl all over them and the stuff doesn't get read on your on your dodgy initial website, but it then turns up in the AI-assisted results on Google as reliable stuff.

So, actually, you know, the move towards newspapers doing deals in order to have reliable information out there that these things are training on probably is quite a good one.

That is better than the Russians training the bots.

Yes, but that is a better bet, isn't it?

That's essentially it.

If they're crawling all over

Guardian stories and FT stories, that probably is better than them crawling all over dodgy Russian websites.

Yeah.

We long had the issue of circular confirmation, where something would

in a newspaper article and therefore become part of the footnotes of Wikipedia and then be referenced by a newspaper article and then be referenced in the footnotes on Wikipedia and the source became its own source repeatedly.

Yes.

It was one of my proudest boasts in journalism.

It still is that I got the rules of Wikipedia changed because we pointed this out.

Oh, this is about 20 years ago now.

And I was contacted by a slightly shame-faced editor from Wikipedia who said, okay, all right, yes, no, you've made us look silly.

Now we're going to change these rules.

We're going to look at changing them.

But is there any likelihood of the rules changing that a bot has to be checked?

A friend in journalism pointed me to an advertisement for a reporter, one-year contract, and the job was to write pieces in order to train AI.

That's literally would you like to dig your own grave and then fill it in in it.

Bring your own spade, isn't it?

And I can see the argument it's helpful to have a proper reporter training it, but you're just wiping yourself out, aren't you?

What worries me is the idea that

it's going to go the way of the supermarket till, where you will have one person overseeing a load of tills,

and you know, they'll come over and help you out if there's a problem.

But fundamentally, shoplifting has gone up a lot since that has been introduced.

You know, it's brought its own problems with it.

And then, I mean, I can see there might be a case for some of the sort of the most rudimentary reporting, which is almost programmatic language.

You know, I'm thinking weather reports, things like this.

Yeah.

Maybe not needing a full human involvement.

But then again, the question is where that lies.

We've got to be balanced.

We've got four reporters here saying that the case against AI is pretty much conclusive and that we don't need it.

Yeah.

Outside of journalism, we've seen issues where AI systems sort of entrench biases, like when it was making probation decisions.

And it's basing its decisions on past probation decisions, but it just sort of every iteration around the system, it got a bit more racist because there'd been some initial racism in the way that those decisions were made but every time you ran an AI it deepened that and worsened the problem so it wasn't making unbiased decisions it was taking the human bias and multiplying it every time it ran and that presumably is the effect if you have state agents involved in AI they put what they want to hear into AI, the Russians, the Chinese, you know, whoever, and then that deepens.

Yeah, every time it comes across it, it confirms it to itself.

So it just comes around the next time and tells itself that that was definitely right the last time.

That is identical to the Wikipedia problem that you identified.

Well, there were various people who would ring up and

contact Wikipedia and change their own birthday if they were a public figure and Wikipedia had got it wrong.

And someone said, Yeah, but I've got a copy of the paper here that

says

it's the 12th of August, so sorry.

No, it is true now.

So that's not been solved.

But I mean, the government does have an AI bill which is planning, as far as I can tell, to completely open the doors to all of this stuff and legitimise the large-scale use of what I would think of copyrighted texts.

Well, it may be too late for this because the other thing that's emerged since the last issue in The Atlantic, the other paper which our esteemed and absent colleague Helen Lewis works for,

was that Mark Zuckerberg's meta have simply gone ahead and stolen a load of copyrighted materials, books, which they've used to train their own AI models on.

It probably, and I should say for the benefit of the private lawyer who will be looking over this before it goes out, it's not piece of robot.

Don't worry about him.

Yeah, we could look at getting some AI from that.

That would save us some money, wouldn't it?

No, for the benefit of the lawyer, I'm not accusing Mark Zuckerberg being a thief.

He's actually receiving stolen goods because this is an entire library called LibGen, which is just made up of pirated books, millions and millions of them.

And the Atlantic found memos within Meta, which simply said that it looked like it would be

unreasonably expensive to actually pay any of the authors of these books

for the use of their work to train AI on.

So they simply lifted this entire already illegal library, LibGen, and they've stolen all of my books.

I checked, they've stolen all of yours as well, Andy.

Get out.

They've stolen the private I Annual.

That's in there as well.

Now, this is serious.

They've also stolen an awful lot of your other books, Ian, which I discovered were all about the Catholic liturgy.

And then I looked at the publication dates on them and they were all from the 1950s.

So I didn't know you said you as all theologian relative.

Well, that was my big discovery.

That sounds absolutely bang on.

Look, I was younger then.

I took it.

Seriously, this stuff is just being stolen.

And that was also what the newspapers were very big on.

You remember last month, the front pages right across all of the national papers, except I think the FT and all of local papers as well, saying

copyright law is there for a reason that writers and creators and artists or anyone just should not have their work ripped off by very, very, very rich tech companies in order to train things that eventually are going to put us all out of jobs anyway.

That's very disappointing to hear that

our books were all on there.

I mean, it's never the most sympathetic spectacle.

I know.

People complaining about something because it's put them out of work.

I mean, you know, the poor old Luddites.

You know, people didn't like them much, and they don't like it much when we point this out.

But it is amazing, wide-scale theft.

And when the Chinese deep seek appeared and then all the tech bros said, you've stolen our material, it did make us all laugh because it's all pre-stolen, as we say now.

Do you want some good news?

Yeah.

Because so little of Private Eyes material is online, the bots cannot train themselves on it.

They have not got a clue.

So I asked, preemptively, thinking of my own career, Chat GPT last week, write a story about The Guardian in the style of Private Eyes Street of Shame section.

It's got a headline on it.

The Guardian's Guardian Angel, a Street of Shame exclusive, which, as you know, is how we start all lasting.

Hang on.

Yeah.

I'm embarrassed already.

Here's your intro.

In the ever-bustling streets of Leftyville, home of the wokes, the Guardian stands tall as a bastion of progressive journalism, or so it likes to think.

But recently, it seems that the paper's reputation is slipping faster than an unpaid intern's ability to secure a byline.

What do you reckon?

Ian straight in with this one?

Leftyville.

The most worrying thing about this, I have to say, is that it was based on a complete bit of fiction because the story it pitched to me was about the Guardian introducing a premium-rate paywall paywall on their website which is something that simply has not happened so not only did it write it in the wrong style it's just making facts up as well which is sort of limey

dreaming of buying your first car or a new home knowing your FICO score is the first step in making it real with my FICO you can check your score for free and it won't hurt your credit you'll get your FICO score full credit reports and real-time alerts all in one simple app your credit score is more than just numbers it's the key to building the future you've been working toward.

Visit myfico.com slash free or download the MyFICO app and take the mystery out of your FICO score.

Okay, we come now to our second story this week.

Jane, the last edition of the Sunday Times had as its front page, very exclusive investigation, a big story about fraud in student loans.

It did.

Yeah, and this was, I'm sure, news to you.

Well,

it was a little bit familiar, really.

The story is that a lot of universities franchise out some of their courses to other providers.

These are private colleges and business schools and things like that.

And it turns out a lot of them are just running scams to get student loans, not necessarily providing anything like the courses that they're advertising.

Talk me through where the student loan goes and who's benefiting from it.

Well, a student loan, hard as it is for

actual, real students to pay back,

you know, could be a significant financial drain for young people, but it is a good loan rate if you are just trying to finance things.

Especially if you're never planning to pay it back and write it off, and the student loan company is not going to be able to pursue you because perhaps you've taken it to another country, then getting a student loan is a great way to get some money fast.

Is it the fault of the institutions which are taking on these students who are perhaps less qualified or not willing to pay back the loan?

Or is it the fault of the people, the students, quote-unquote students who are applying?

I think there's some blame to spread around.

Certainly actual fraudsters are the first kind of line.

However, there's

very long been an issue with universities not keeping a close enough eye on their franchised providers.

And this goes back years in terms of private eye writing about the issue with franchise providers.

Long before the fraud issue specifically came up, we've talked about some of the terrible pseudoscience quackery courses that they were actually sort of allowing to be accredited under their names.

We had things like animal chiropractic and homeopathy degrees being signed off by Bay Universities, but the the courses were being run by these separate providers.

and the universities just ticked the box.

Yes, you can have a degree with our crest on it.

A lot of work was done on reducing the amount of that, but there's still some of it out there.

And that's a way of the university getting some money in because they don't have to run the course in animal homeopathy or whatever it might be.

But they still get some tuition money for that.

Right, okay.

I mean, I would say this is yet another woke lefty

attack on the very fine business of outsourcing, which Private Eye seems to write about endlessly.

It it just doesn't work very well in the education field, does it?

Well, no, I mean

so again, it's not new that courses have have turned out to be fraudulent.

A few years back, Ofsted was looking at one in in further education at college and discovered that many of the students that were registered for this college not only weren't doing the course, but hadn't even heard of the college.

They were just on a list of names stuck on a register so that they could claim more money from the funding bodies.

Does the money go to the outsourced provider?

Does it go to the student themselves?

How does it work?

And this is student loan fraud, not tuition fee.

There's whole other issues going on with where the tuition fees go and universities have different deals with different providers in terms of whether they're having a partnership or a franchise deal.

But in terms of the student loan fraud, that's a student's relationship with the student loans company.

But if you're doing organised crime, then you can set up a course and get an organised group of people to lend their names to claiming lots and lots and lots of loans.

And this is where we come back to the Sunday Times story, is not news to private eye.

As just over a year ago, we wrote about this happening with the University of Northampton and one of its franchised providers.

They found that 600 students had effectively submitted the same coursework.

At which point they asked some questions.

Are these students

real?

Are they attending the course?

What's going on there?

At this point, reported themselves.

I mean, at least they found out.

It's, you know.

Northampton reported themselves to the Office of Students.

So, this is how we found about it, because they had to put a line in their annual report saying that they were going to have to pay £6.1 million

back in terms of tuition fees.

Wow.

So, there was clearly.

You spotted this with your eagle eye, £6 million

as being slightly unlikely.

Well, it was

what was helpfully flagged up was that the National Audit Office did a report on issues with the student loans company.

And although they anonymised it, they did make a note that one university had had an issue where they'd spotted a 6.1 million fraud, which made it a lot easier to look for, because they gave the exact amount to look for in the university accounts.

I'm just imagining whoever had the job of marking all that coursework, because they would have had an unbelievably easy time of it.

It's really good.

And so that was a franchised thing at the University of Northampton.

That wasn't wasn't in-house, that was exterior.

What is to stop the four of us doing this now?

Let's say the four of us approach the University of Buckingham or wherever, just using them as an example, and we say, look, we've got this great course in, shall we say, investigative journalism?

Should we say satirical journalism?

Satirical journalism.

Satirical journalism.

That sounds good.

Invite a lot of students to take part in this.

How does that relationship work?

Do we know?

This sounds like a trip to Romania.

Well, that's the thing, isn't it?

A lot of it is Romanian-based, all the recent forests.

Well, in terms of money having left the country, that does appear to be where significant amounts set off to, yes.

So the students from an unknown, possibly EU country,

come over, they take out the loan.

At the end of it, someone says, well, you don't appear to have attended any courses.

And they say, oh, I was working from home and you can pick it up.

And they sent me the texts.

and presumably quite difficult to prove.

And in the meantime, they've got a loan which they possibly have no intention of paying back.

Or even if they did, if they've had a loan at extremely beneficial rates compared to a bank loan, they haven't actually been a student for, so that would still be fraud.

But in the case of these particular

scams, these are just find a list of people who are willing to have their names put on a list.

It seems an incredibly obvious scam.

in higher education.

I'm not saying that they're worse at spotting it, but I mean, shouldn't it have been detected a bit earlier?

So, universities can be very, very proud of their global reach and their kind of their London campus and, you know, by being sort of more than just University of Mugglesville, then they now have courses all over the place, and their name is getting out there.

And a lot of universities, we should say as well, are in pretty dire financial straits at the moment, aren't they?

So, they do need to be considering ways of getting money in to keep the doors open at all, don't they?

They are getting some income through these deals.

That's just what I was going to ask you, Jane.

In terms of the scam that the four of us are going to start running, how are we going to benefit from that financially, please?

The tuition fees.

We get the tuition fees.

They get the cheap loans.

We get the tuition fees.

Okay, great.

We don't provide any teaching, so there's no costs.

Well, we don't have the seats for it.

Yeah, that's fine.

But the Sunday Times, who did this piece quite well, I mean, they had some funny pictures of these sort of outsourced teaching institutions that appeared appeared to be above a chip shop or a couple of chairs or a brass plaque somewhere.

Yeah, absolutely.

Studied Wizarding Worlds.

University of American hard candy.

Yeah.

I think the Eastern University of Coppitz isn't it?

He's working at the number one vape shop in the whole of the country.

We're very impressed.

If people say, oh, I'm not going to pay the student loan back, is that because the figures suggest that lots of people don't?

Yeah.

Genuine students often are not paying them back.

Obviously it depends on their

income and

a lot don't manage to earn enough to be paying a student loan back

for a long time, if at all.

Or

they manage to lose people in the system that aren't going to pay them back.

So in particular if you're just a fraudulent name on a list then you're going to be quite hard to find because

we don't know exactly how accurate the information they're getting about people is.

And this sort of thing sort of looks less unusual because all sorts of universities are opening weird campuses overseas and things anyway.

You've written a lot about that, haven't you, Jane?

Absolutely.

So universities are constantly trying to extend their reach and build their empires and open their campuses in dubious regimes.

And perhaps if you were just the university in one city with sort of one structure, you might be able to keep a better eye on your overall finances and what you're putting your name to.

You're suggesting growth isn't always good, Jane.

I think I am, yeah.

I think growing in Kazakhstan or, you know, it might be sort of worth having a second think about.

So, this has been cracked down on?

Is going to be cracked down on?

Azion asks.

Yeah, it's going to be investigated by the public sector fraud investigation body, although they very much sort of nip round and ask the people at the National Audit Office who did the investigation last year

and a previous investigation in about 2017 and a previous investigation in about 2012.

The National Audit Office have been kind of banging on about this for a while, that it's a massive risk with the way student loans work is that

you can wander off with your full student loan and never pay it back and not do a course.

So this predates all of the COVID loans stuff that's still being looked into, doesn't it?

This is a very long-running thing that suddenly we had.

I mean, the Sunday Times stories, unlike Private Ion's Cat, came with a column from Bridget Phillipson, the education secretary, saying it's time to crack down on this sort of thing.

Yeah, I mean, the government has recognised it needs to crack down on it.

They've been running a consultation on how to crack down on it since January.

So

consultations

around.

They're trying to work it out.

Okay, that's it for this episode of page 94.

Thank you so much to all of you for listening.

We'll be back again in a fortnight with another one.

Just a reminder, if you want to send in a question for page 94 live at the Royal Abbott, sorry, Cambridge Literary Festival,

you just email podcast at private-i.co.uk.

Thanks very much for listening.

Thanks to everyone for participating.

And thanks to Rethink Audio for producing.

Bye for now.

Hey, if you listen to iHeart K-pop with Jojo, let me say thank you and turn you on to something next level.

Hello, Soju's sparkling Soju.

It's light, sparkling, and packed with five delish flavors.

My two faves, peach and Asian pear.

Oh my god.

Smoother than hard seltzer and much more fun than beer.

This drink is all about good times and sharing vibes.

And trust me, once you try it, you'll get why everybody's talking about it.

Order now and take 15% off your first order.

Just enter code JoJo15 at checkout at hellosoju.com.

Hello, Soju.

Every sip is a hit.

Please enjoy responsibly.

A happy place comes in many colors.

Whatever your color, bring happiness home with CertaPro Painters.

Get started today at Certapro.com.

Each Certapro Painters business is independently owned and operated.

Contractor license and registration information is available at Certapro.com.

Time.