The Braindump Blog

From Ars Technica:

After months of resisting, Air Canada was forced to give a partial refund to a grieving passenger who was misled by an airline chatbot inaccurately explaining the airline’s bereavement travel policy.

This feels like a good, important, decision. Would Air Canada have fought so hard if it was a human customer service agent that misled the company? There is no practical difference to the customer. Although to be fair, the ruling suggests that maybe they would have, with the remarkable line that:

Air Canada argues it cannot be held liable for information provided by one of its agents, servants, or representatives

In which case why even have a customer services department? Are you supposed to be able to guess whether or not an agent is telling the truth every time you call them up?

In this case it’s not like we’re talking about something that threatens the existence of the company. Air Canada, a company that measures its quarterly revenue in billions of dollars, isn’t going to go bankrupt if they give the customer the $880 refund he was expecting.

In any case, if a company is going to replace its human customer service agents with chatbot customer service agents then they should surely be equally as liable for any misinformation.

As the judge notes:

In effect, Air Canada suggests the chatbot is a separate legal entity that is responsible for its own actions. This is a remarkable submission.

If we want to regard chatbots as being “responsible for their own actions”, well, some kind of radical legal shift would presumably be needed. It’s also kind of insane. If chatbots have responsibilities then do they also have rights? Do we put them in chatbot jail if they say something wrong?

If what they actually mean is that whoever sold them the chatbot should be held responsible, well, there are surely already legal routes that can be used by Air Canada to claim damages against another company if they believe they’ve been mis-sold something.

In no circumstance should it matter than the actual bereavement policy was somewhere else on their website. It’s not the customer’s responsibility to research the same question several times and try and guess which version the company actually meant to provide.

More from the ruling:

While a chatbot has an interactive component, it is still just a part of Air Canada’s website. It should be obvious to Air Canada that it is responsible for all the information on its website. It makes no difference whether the information comes from a static page or a chatbot.

While Air Canada argues Mr. Moffatt could find the correct information on another part of its website, it does not explain why the webpage titled “Bereavement travel” was inherently more trustworthy than its chatbot. It also does not explain why customers should have to double-check information found in one part of its website on another part of its website.

This multi-year miscarriage of justice is embarrassingly new to me, but then again so was an event that it has recently been compared to - the Post Office Horizon scandal.

A decade ago, one condition for foreign students who applied for a visa to come study in the UK was that they first had to pass a specific English test called the Test of English for international communication or Toeic. Although I haven’t seen the questions, it sounds like it’s quite basic, something that anyone reasonably fluent in English would do OK at. It’s intent is merely to “measure the everyday English skills of people working in an international environment” rather than check whether you’re PhD-level familiar with the literature of the English-speaking world.

The Home Office authorised 4 approved test providers that students could use to do this. One of these was US-based company, “Education Testing Service” aka ETS.

Back in 2014, TV show Panorama found that there was cheating going on in a couple of their centres. As far as I can tell everyone agrees that this part is true: cheating definitely went on in 2 of around 90 centres ETS ran.

But rather more controversial was the follow-up investigation that claimed that of the ~58,000 students who sat that test in their centres over a 5 year period, over half “had used deception” and another 39% were questionable, entirely on the basis of some unspecified voice recognition technology analysis. Yes, the claim was that 97% of entrants cheated.

Consequently, these students had their visas revoked. This made it illegal for them to remain in the UK. At best they could hope to be simply thrown out of their (expensive) university courses. At worst some were locked up in detention centres and eventually deported, only to live a life in disgrace as their nearest and dearest were led to believe they were nothing but a foolish cheat.

From politics.co.uk:

Then people started being rounded up. Families were woken in dawn raids, the husband separated from the wife and sent to detention centres ahead of deportation. It would have been a profoundly harrowing experience. These were students, not criminals. Their only crime had been to do a test recommended by the Home Office. But the full sadistic force of the state was brought down upon them.

Not even a entitlement right of appeal was given at first, at least without having first left the country.

All in all, 2500 were forcibly deported and another 7.2k have left the country after being threatened with arrest. Others have ended up destitute and homeless, depressed or suicidal.

Since then at least 3600 of those students have won immigration appeals, although the public records don’t specify whether that’s because they’ve been found to be innocent of cheating or for some other reason. But it was certainly the supposed cheating that put them in the position of needing an appeal in the first place.

Unsurprisingly, many of the students have claimed they are entirely innocent and that any supposed evidence against them is extremely weak. The Home Office hadn’t previously acknowledged let alone taken action based on that. But they were certainly aware of the issue.

In 2019, the National Audit Office found that the Home Office didn’t have the expertise to validate the evidence provided by ETS, and that it had simply taken it at face value. They’d made no real effort at all to investigate whether or not the accused had been incorrectly classified as cheats

Even earlier, in a 2016 tribunal case that challenged the actions of the Home Office on these matters, the presiding judge found that none of the people representing the Home Office:

had “any qualifications or expertise, vocational or otherwise, in the scientific subject matter of these appeals”. At no time, in fact, had the Home Office had advice or input from a suitable expert. The criticisms of their witness statements “were not addressed, much less answered, in their evidence”.

and that:

The Home Office’s case that these students had committed fraud was “entirely dependent” on ETS. The tribunal found the Home Office accepted uncritically everything reported by the firm.

One of the complainants in that case, a Mr Qadir who had been accused of cheating the test, of course spoke “excellent English” when testifying.

Thousands of the other accused people were waiting on the results of this case to use as evidence in their own appeals. Some of their cases had specifically been put on hold until this tribunal’s verdict was out. However, at the time the committee refused to report on their findings, as important and universal as they were, making it very difficult to cite elsewhere.

The decision means that the case cannot be cited, except under very strict and laborious conditions, in other appeals. It means many thousands of people who have been unjustly deported will not even know of its existence. The decision makes the ruling against Theresa May legally useless. It’s as if it never happened.

Lawyers went on to accuse the committee of “an abuse of power”, with their decision to not publish what went on as being “plainly hostile to common sense”. Accusations flew that the decision whether to publish or hide the proceedings of the cases seen by this committee were very much dependent on whether the case showed the Home Office in a positive light.

The years this was all happening in were well into the still-ongoing inhumane period where the at-the-time British Home Secretary Theresa May was explicitly trying to create a “hostile environment” in theory for illegal immigrants, but in practice for anyone that in any sense didn’t “look British”.

Being able to tout ever bigger numbers of punishments against and deportations of immigrants who were in Britain illegally - the situation these poor students found themselves in when stripped of their visas due to having been accused of cheating on a test - was exactly what the government, particularly the Home Office, would have wanted. Clearly there was little incentive - outside of, you know, the basic tenets of justice and humanity - for May’s department to look into it.

May now regrets the phase “hostile environment”, but so far that doesn’t seem to have helped many of the folk who unjustly suffered, nor lessened the hostility of the environment itself.

The “97% of people are cheats” figure just seems prima facie implausible to me. Surely it’s well beyond the threshold that basic common sense should have caused an immediate halt to the punitive proceedings and further investigations to have urgently been done a decade ago.

In the unlikely case that cheating was really that endemic, there have also been scenarios suggested whereby it may have been the centres, rather than the students, that were cheating. And if it had turned out that 97% of individual students had cheated, well that surely points to a systemic problem with the structure of the testing regime than something individually defective about almost every single person who took the test.

But, in any case, that’s all a meaningless thought experiment when it seems like there’s no real evidence that cheating on anything like this scale ever happened in the first place.

Also to add to the mix: the inconvenient fact that several of the students entangled in this mess were provably perfectly fluent in English well before they took the test. To the extent it’s basically inconceivable that they could have failed what is a fairly basic test. Hence there would be absolutely no incentive for them to have paid someone else to take it for them, which is generally the allegation being made, or engage in any other such nefarious scheme.

One such example of this is poor Kishor. He successfully completed a degree in English literature, worked as an English teacher, and worked with NATO troops in Afghanistan, which required passing an English test, before coming to the UK to pursue a business degree which is the point at which he took the Toiec test. He probably speaks better English than I do. Nonetheless he was amongst the crowd that were accused of cheating.

Learned a new phase last week: stinge-watching.

It’s the opposite of binge-watching. Particularly for the cases where you could in theory binge all episodes right now just like the streaming services want you to do but instead you go out of your way to space them out.

📺 Watching The Apprentice - season 18.

It’s sometimes reassuring when things stay the same - and here once more we have a bunch of buzzword-laden people, with apparent levels of self-confidence most of us could only dream of, doing whatever it takes to impress the master of awkwardly delivering badly-scripted jokes and pointing at people, Lord Sugar, all in the name of getting an investment.

I was just looking up the iconic 2010 contestant Stuart “The Brand” Baggs to be sure of getting his justifiably famous quote right - ‘I’m not a one-trick pony. I’m a whole field of ponies’ - only to discover that sadly he died in 2015, aged just 27. Lord Sugar left a floral tribute.

Another thing I recently learned is that in order to keep who wins a total secret until broadcast, the show apparently films two different endings, one where each of the two finalists win and the other one loses. Even the two candidates concerned don’t learn who won until much later - the day before the final show is broadcast. Same for the production team.

Sometimes even the same ending has to be filmed more than once. Harpreet Kaur, the winner of the 2022 season, had to have her “winning” ending filmed twice because she didn’t have “the right level of excitement” the first time through. I suppose it is probably hard to muster the TV-level appropriate extremes of performative excitement when there’s a good chance what you’re celebrating never happened. I’m not sure how she reconciles that with her quote in the same article that “there’s no acting” in the Apprentice, but there we go.

I’ll watch with interest when it’s time for the 2024 final to see if this cohort’s acting-like-a-winner skills are better than one might imagine.

For anyone that finishes Wordle too quickly, Puzzmo - “the (new) place for thoughtful puzzles” - looks to be a cute ‘n’ fun newspaper-puzzle-page style site.

It’s got the classic crosswords, slightly reimagined for the digital age, along with a mass of other puzzle types, old and new. A good chunk of it is entirely free to play. Not sure how to fit more puzzles into my life, but it’s pretty fun and that’s even before experimenting with the interesting sounding “play with a friend” options.

Their manifesto is also pretty cool. You deserve better, you are smart, play is healthy - all good affirmations.

They’ve also already been acquired apparently. That was quick. Let’s hope (against all hopes?) it doesn’t tarnish the original vision. First up it seems like they’re offering revenue-share collaborations with other “brands” - that’s why for instance you could choose to do the same puzzles alongside Polygon branding if you really want to.

Here’s an embedded “Spelltower” puzzle to keep you busy. Click and drag the letters to form words, submit by clicking on the last letter again, in order to clear as many letters as you can.

TIL: Elon Musk wasn’t the founder of Tesla. Tesla’s real founders were Martin Eberhard and Marc Tarpenning.

Musk was merely an early investor in the company. Who seemingly started out as he meant to go on.

From The Verge:

Musk was brought on as a crucial early investor but soon used his clout, money, and even a few strong-arm tactics to oust Eberhard and Tarpenning and eventually install himself as CEO of Tesla.

Eberhard ended up suing Musk for all manner of things: libel, defamation, breach of contract, infliction of emotional distress, and failure to pay his wages, The only reason Musk is even allowed to call himself the founder of Tesla is because he effectively purchased the right to do so as part of the legal settlement.

Interesting that “social media” is such a toxic brand these days that Snapchat’s recent marketing campaign is all about how Snapchat very definitely isn’t social media

In fact, it was built as an antidote to social media

This is rather different to how they previously touted it, at least to organisations who might want to pay them something. Less than one year ago they were busy business-blogging that you should throw your advertising dollars their way given that:

Snapchat is a prime example of a social media platform.

NearbyWiki.org is a fun site that plots Wikipedia encyclopedia entries over an interactive map. You can then, for instance, see what stuff-famous-enough-to-be-on-Wikipedia exists near where you live.

One thing this led me to learn is that there are a lot of entries for pretty unexciting roads on Wikipedia. I suppose there are about 262,300 miles of them to go around in the UK.

📧 Reading the Bits About Money newsletter.

I’d never spent too much time thinking about how the financial infrastructure behind organisations such as banks and payroll providers was set up. It turns out it’s pretty interesting to read about, and may even help explain a lot of the weirdness about how such organisations treat customers.

All previous issues live here if you’re more of a web-reading fan.

TIL: Buckingham Palace pays a council tax bill of £1,828 per year.

That’s lower than 46% of households in the UK according to The Economist, who rightly argue that the whole system needs a shakeup for reasons beyond palace inequities. It’s an extremely regressive tax as it stands.

Elon Musk has implanted a microchip in someone else’s brain.

No, this isn’t some weird conspiracy theory, it actually happened.

Well, it probably wasn’t him, but rather someone a bit more qualified to do so from a company he founded, Neuralink whose stated mission goal is to:

Create a generalized brain interface to restore autonomy to those with unmet medical needs today and unlock human potential tomorrow.

It’s an FDA approved operation. The end goal is to let folk use their computers just by thinking, developing a “Brain Computer Interface”. This has obvious applications for people who can’t use the limbs for instance. In the future the same technology might be used to restore vision to the blind.

So it’s all exciting and potentially “good-for-the-world” stuff. If only the show wasn’t being run by Mr Musk I’m sure I’d be slightly less worried about it. It seems like The Guardian couldn’t resist a bit of a commentary on the guy and what he chooses to spend his time on in-between whatever running his businesses involves:

In follow-up tweets sent in between arguing about video games and bantering with far-right influencers, the businessman said the first Neuralink product was called Telepathy.

Telepathy of course, despite its potential for revolutionary good, also sounds like something straight out of Black Mirror. Musk might have a vision for it beyond the medical it seems:

Speaking in 2017, Mr Musk said the ‘merger of biological intelligence and machine intelligence’ would be necessary if humans were to stay economically valuable.

I feel like a bare minimum criteria for the criminal abrogation of the UK’s international responsibilities and sense of humanity that is the plan to send our asylum seekers to Rwanda should be that people from Rwanda are not currently finding the need to seek asylum from Rwanda in the UK.

Nonetheless, from the Observer:

Four Rwandans were granted refugee status in the UK over “well-founded” fears of persecution at the same time as the government was arguing in court and parliament that the east African country was a safe place to send asylum seekers.

I mean, are we just going to put those who were compelled to flee Rwanda on a flight to…Rwanda?

Even if Rwanda was the paradise-for-all that the Government attempted to legislatively wishcast into being the case recently, the idea is entirely unworkable, unethical and illegal anyway. But come on, surely even some people who would otherwise support this dreadful culture-war nonsense might concur that the idea that is is OK to send asylum seekers to a country that legitimate asylum seekers come from is a non-starter.

One person fleeing from Rwanda saw a successful asylum claim here literally the day after the Government finished arguing that it was safe in the Supreme Court.

📺 Watched season 6 of The Crown.

This is the final season of this “is it fact or fiction?” drama documentary, which nonetheless will also probably become the primary reference material for royal life in the minds of some of its viewers.

It mainly focuses on the years from somewhere around 1997 to maybe around 2006, although with shoutouts here and there to times longer ago - we are all a product of our past to some extent, even, or perhaps especially, the royals - and perhaps as recently as 2022 in the slightly strange finale.

1997 is of course the year that Princess Diana died, an event that the series focuses on plenty, before and after. These are also the Blair years, her rivalry in terms of public opinion depicted as giving the Queen some very strange dreams.

Focus then moves to teen drama as Prince William tries to win the heart of Kate Middleton, which her mother is pretty enthusiastic about.

Prince Harry is in the midst of his initial Bad Boy phase, dressing up as a Nazi, drugs, partying, all that kind of classic Prince Gone Wild stuff. Although it has to be said that the very idea that he, his brother, and a load of other posh people routinely go to fancy dress “colonials and natives” parties in the first place isn’t wildly reassuring.

The Queen Mother dies, and the final episode then dwells on the preparation of of Operation London Bridge - the planning for the death of the monarch, an operation that had actually existed since the 1960s in one way or another.

Not everyone loved this season. The reviews were bad enough that it made Dominic West, the actor who played Prince Charles in it, stay in bed for two days. TV critic Nick Hilton sees it as a symbol of the Netflix’s ‘decline in popularity and quality’. Even its creator, Peter Morgan, is relieved to see it end. But when you’ve watched the first five series and actually remember some of the events of the sixth actually taking place, it’s hard not to give it a go.

📺 Watched season 2 of The Traitors (UK version).

Exactly the same setup as in the first season, but with a new bunch of people for me to worry about their future sanity.

Doing tasks by day and trying to figure out who’s leading a double life by night, it remains extremely compulsive even if it’s something I’m not sure is overly healthy. The last episode is quite something. I got far too invested in that emotional rollercoaster.

Cory Doctorow reminds us that AI doesn’t actually have to be better than us at our jobs in order to threaten our current livelihoods. It’s only necessary that our managers come to believe that an AI can do a just-about-adequate version of something akin to our work whilst costing less.

That’s the true cost of all the automation-driven unemployment criti-hype: while we’re nowhere near a place where bots can steal your job, we’re certainly at the point where your boss can be suckered into firing you and replacing you with a bot that fails at doing your job.

This very explicit cookie consent / surveillance message jarred me a little today. All I wanted to do was read a single article on Teen Vogue of all places. Doing so by default would entail my personal data being sent to 143 different companies, including active scans of my devices and my precise geographic location at the time. Whilst “caring about my privacy”.

I’m sure Teen Vogue is no worse than its peers. I actually applaud the explicitness of the message. We should know exactly what we’re agreeing to when we mindlessly hit the “yes ok sure if you must” button whilst surfing.

If you hit “Show Purposes” you get a list of the types of information being shared. Some, but not all, of them you can disable. But just for fun, here’s the list. The numbers in brackets after each one shows the number of partners that are allowed to use whatever they can scrape about me from my web visit for the given purpose.

• Functional Cookies
• Performance Cookies
• Targeting Cookies
• Strictly Necessary Cookies
• Audience Measurement
• Store and/or access information on a device (“131 partners can use this purpose”)
• Personalised advertising and content, advertising and content measurement, audience research and services development. (140)
• Use precise geolocation data (50)
• Actively scan devices characteristics for identification (15)
• Ensure security, prevent and detect fraud, and fix errors (106)
• Delivery and present advertising and content (92)
• Match and combine data from other sources (89)
• Link different devices (87)
• Identify devices based on on information transmitted automatically (94)

A friend shares the UK National Careers Service’s job description of an astronaut with me. I’m slightly underwhelmed with the salary to be honest.

“You could work away from home” seems something of an understatement. No WFH for a spaceperson :(

🎶 Listening to Sweet Justice by Tkay Maidza.

Tkay is a Zimbabwean-Australian rapper who recently(ish) released her second full album, 7 years after the first.

This was the first album for a long while that I learned about by virtue of hearing it on the radio, just like it’s the 1990s. Sweet Justice itself is a multi-genre extravaganza, apparently themed on the idea of breaking up with one’s own past, leaving the negative influences behind.

This (taken from some random Discord if I remember correctly) might be a little simplistic in some cases. And perhaps it even plays into the more ludicrous conspiracy theories around the WEF. Nonetheless it remains a constant source of astonishment to me how many big and important problems humanity at least has a good idea of how to solve but seemingly chooses not to.

For the most part I suspect that comes more from us having inadvertently created structures such that we are disincentivised from doing so - especially at the individual level. Rather than, for instance, pure malice.

1 mobile phone gets stolen every 6 minutes in London, back in 2022 at least. That year 90,864 got stolen, almost 250 a day.

To be fair London has a lot of people in it - around 9.5 million at the time - and hence lots of phones available for the taking.

Just when I imagined I couldn’t learn anything new about the often dreadful conditions that many of the people doing some of the most important and oftentimes challenging work imaginable - caring for other people and their needs - are subject to in the UK, I learn that the sector may be increasingly infused with actual slavery.

From The Guardian at the end of last year:

Post-Brexit restrictions on the free movement of workers from the EU have contributed to modern slavery becoming “a feature” of the care sector in England the Care Quality Commission has told MPs.

The Care Quality Commission made 4 referrals about modern slavery in the period of 2021-22, vs a predicted 50 by the end of the most recent year.

Unseen UK, an organisation that runs a modern slavery and exploitation hotline, reported that over 700 care workers called in 2022.

Unseen’s chief executive, Andrew Wallis, says the current approach has led to a rise in “labour abuse and exploitation” and is “a disaster” for many workers. “Very vulnerable people are being cared for by very vulnerable people,” he adds.

The University of Nottingham Rights Lab produced a report in 2022 that detailed what makes the conditions especially of migrant “live-in” care workers - those who stay in their client’s home around-the-clock - particularly vulnerable to these conditions.

1. Employment status, business models, and the role of intermediaries
2. Information asymmetry between care workers and intermediaries
3. The emotionally and physically intensive nature of live-in care work, blurring of boundaries between work and private life
4. Barriers to exercising rights at work: sick leave, time off, redundancy/notice, health and safety at work
5. Individual risk and resilience factors

This isn’t the only driver by a long margin, but I can’t help but think that any scheme that makes your right to live in a country dependent on you doing whatever your employer demands of you is ripe for abuse. This of course isn’t limited to the care sector.

TIL: Ireland’s Dublin Airport holds an annual “Blessing of the Planes” ceremony.

Once a year, ever since 1947, a priest works their way through the airport giving the planes a “festive blessing from God”.

From CNN:

…[Father] Doyle was shown carrying a chalice of holy water onto the airfield, accompanied by another priest, as well as airport police.

As the job got busier though - more planes = more blessings - some shortcuts have had to be made. It got moved to Christmas day, when no flights take place (meaning some staff have to go to work that day), and it’s changed from 1 blessing per plane to a single general blessing.

It’s not restricted to Aer Lingus' Saint-named planes as it once was; even Ryanair gets a lookin.

The annual blessing these days is ecumenical and covers the general fleet.

📚 Finished reading: Time Shelter by Georgi Gospodinov.

This was a fascinating and, at times, darkly hilarious read. The narrator tells the story of his life after meeting Gaustine, a mysterious intermittently-present therapist who develops a new treatment for Alzheimer' disease.

The treatment involves recreating the environment of their past for the period during which the patient felt most at home and secure, the period their body remembers most clearly irrespective of their state of mind.

It’s very successful, leading to the opening of many such clinics with rooms that replicate the 1950s, 1960s, 1970s and so on in perfect detail.

As time goes on though it’s not only patients with Alzheimer’s disease that want to attend these clinics, these ‘time shelters’. People of all kinds are drawn to re-experience their favoured period of the past, a place-in-time that they remember - accurately or not - as being a safe and happy time to live. Anything to take a temporal break from the stress and angst of modern-day life.

The movement expands way beyond the medical. Nostalgia increasingly permeates everything. Radio stations from decades past are re-set up. Newspapers from 30 years ago are reprinted, as though the events 3 decades past are today’s news. Politicians of course see the potential for using this nostalgia to the end of their own electoral gains and get in on the act.

Whole towns from past eras are re-built in something close to their original form. Eventually entire countries have referendums on which decade was their golden one, which perceived era they should legally and culturally return to.

Different countries naturally see different periods as their most glorious days, leading to a set of time rather than location based international alliances. As well as, I suppose, time-travel-adjacent experiences when crossing borders between them.

State sponsored re-enactments of historical events become the norm, although down that route some danger lies.

Just the now-triggering word ‘referendum’ makes it hard not to see some satirical intent regarding a referendum my own country held now 8 years ago where some people consider a good amount of voting preference may have originated from a longing for (semi-mythological) times of yore. A time when Britain ruled the waves, when men were men and women stayed at home, that kind of thing. But the book’s treatment of this theme is done in a way that never grates.

The novel won a Booker Prize last year. I’m not surprised. Embedded within the story is a wonderful mix of philosophy, politics, ethics, history, psychology, sociology and more, although it is never hard to read. It has moments of hilarity, moments of darkness, moments of poignancy. Very recommended.

Computerphile takes us through some of the technical issues found with the Post Office Horizon computer system that led to several sub postmasters' lives being ruined via false accusations of fraud.

It seems that many of them came from the failure of the system to maintain ACID principles.

From Wikipedia:

In computer science, ACID (atomicity, consistency, isolation, durability) is a set of properties of database transactions intended to guarantee data validity despite errors, power failures, and other mishaps.

The speaker is Professor Steven Murdoch, who also wrote a blog post about this that contains examples of the faults he’s talking about taken from the judgement report of the 2019 legal case of Alan Bates and others vs Post Office Limited.

It’s not only books, music and product listings that are being created, or contaminated, by generative AI content. It’s entire websites in some cases.

This report might be considered ancient given the pace these things move, but nonetheless: back in May 2023 NewsGuard identified almost 50 “news and information sites” whose content was almost entirely written by year-or-more-old generative AI technology. Why? Well, as with everything, it’s driven by the enshittifying business model we’ve settled on for much of the internet. It’s always the business model.

Artificial intelligence tools are now being used to populate so-called content farms, referring to low-quality websites around the world that churn out vast amounts of clickbait articles to optimize advertising revenue, NewsGuard found.

This motivation is nothing new. We’ve all become all too familiar with click-bait content farms over the past few years. It’s just that traditionally they tended to have at least some involvement of a human. Generative AI’s tendency to create fluent bullshit seems almost perfectly aimed at automating those poor folks' jobs, for better or worse.

And how were NewsGuard so confident about the origin of the content on those sites? Well, at least in part, it comes from the now increasingly ubiquitous technique of of looking for ChatGPT-style error messages that made all the way through the “publication process”, if that’s not too grand a word for the resulting low-effort word spew

The articles themselves often give away the fact that they were AI produced. For example, dozens of articles on BestBudgetUSA.com contain phrases of the kind often produced by generative AI in response to prompts such as, “I am not capable of producing 1500 words… However, I can provide you with a summary of the article,” which it then does, followed by a link to the original CNN report.

They provide a screenshot of an absolutely perfect example. “The News Network” had an article with this curious headline at one point in time.

There were also some rather more subtle clues as to the AI source in some cases, their article goes into more detail.

That was then. More recetnly they’ve assembled a newer list of 634 AI-generated sites that each meet all of these criteria:

• There is clear evidence that a substantial portion of the site’s content is produced by AI.
• …there is strong evidence that the content is being published without significant human oversight…
• The site is presented in a way that an average reader could assume that its content is produced by human writers or journalists…
• The site does not clearly disclose that its content is produced by AI.

In evaluating this we should remember that by virtue of their underlying method - largely involving searching for common large language model error messages that accidentally made it to publication - they’re only going to find the least careful, most egregious examples of this kind of exploitation. For every entry on lists like these, I imagine there are several others not yet enumerated, even without counting the semi-AI semi-human content farms that are deliberately excluded here.