Tag Archives: how accurate are chatbots?

The Home Office screws up yet again, and other news from Britain

Posted on by
23

Britain’s Home Office–the scandal bedecked arm of government that’s supposed to deal with “crime, the police, drugs policy, immigration and passports, and counter terrorism” –could have a new scandal on its hands any day now: it’s lost track of an estimated 200,000 people who have the right to remain in the country. These are people who’ve lived in Britain for decades but who didn’t make their way into the computer system because they landed before the computer did. They got either a letter or a stamp on their passports confirming their right to remain in the country, and that was good enough. Until now. 

Now the Home Office wants them all digitized. So the paper documents? Pffft: they’re worthless. Everyone who has them has to go online and upgrade their documentation. 

Any time you hear the word upgrade, put on your flak jacket.

Okay, I admit, upgrade is my contribution to the discussion. The official language has to do with creating an online eVisa account. Either way, the Home Office says the process is going smoothly. From the bureaucracy’s point of view, that probably means it hasn’t caused the Home Office many problems. Users say they’ve had to fight their way through glitches. The an organization called 3 Million says the bigger problem is that the Home Office doesn’t know how to contact many of the people who rely on paper documents, so it hasn’t been able to tell them the documents they’ve been relying on are about to be worthless. 

Irrelevant photo: a hydrangea

Don't worry about the graphics here. They're almost all irrelevant to the text.

Irrelevant photo: a hydrangea

What happens to people who don’t have valid documents? The risk is that they could be treated as illegal immigrants, who are the current political boogeymen. They’ll be locked out of the pensions they worked for, along with housing, health care, and other services. 

And the problem isn’t just that they can’t all be reached. They’re none of them young–they arrived pre-computerization, remember–and they won’t all be technologically gifted. You know how that happens: The decades pass, you get older, the world changes, and you don’t necessarily keep up with it. 

But gee, it’s progress, and if a few bodies fall by the wayside, who cares? At least until there’s a public flap about it, at which point all decision makers will put on their surprised face.

 

The ghost of Boaty McBoatface

Having told us there’s no money for (almost) anything sensible, Britain’s government has decided to redesign the bank notes. Because, hey, why not? It’ll lift everybody’s spirits. And now that not many people use cash anymore, what could be a better time to redesign it? 

I haven’t been able to find out how much the redesign will cost, but what the hell, it’s only money.

So it all makes perfect sense that someone decided to get the public involved by asking what picture people want to see on the new notes. That worked really well when they–that’s the public, you understand–were asked to choose a name for an arctic research vessel and chose, by a wide margin, Boaty McBoatface. If you missed the story, you can catch up with it here. It’s a testament to both the British sense of humor and British bureaucracy at work. 

Already one writer, Athena Kugblenu, has suggested honoring British culture with a picture of an organge traffic cone. 

Why a traffic cone? 

Because the country has an uplifting tradition–which generally involves a combination of alcohol, youth, and athleticism–of putting them on the heads of statues.

If you want to suggest something for the redesign, here’s your link. And if it’s suitably absurd, leave it in the comments as well.

 

And since I mentioned statues

It seems folks have been climbing the statue of Winston Churchill in Parliament Square, not necessarily to add a traffic cone but during protests, although someone did add a strip of turf to give him a green mohican.  

 So in May the government made moves in the direction of turning that into a crime. Not the mohican and not climbing on statues in general, but climbing on this particular statue. As the Sun, one of the trashier of the right-wing papers put it, “Thugs who climb on Winston Churchill’s London statue face JAIL.”

I hate to link to the Sun, but what the hell, I am quoting it. And they did use all those capital letters. They had to. If they don’t use them now, Trump will gobble them all down and there’ll be none left for anyone else’s hysteria.

The penalty is up to 3 months in prison and a $1,000 fine. The bill, is if passes, applies not just to the Churchill statue but to monuments commemorating World Wars I and II as well.

Sleep well tonight, my friends. The country will be a safer place to live in once this passes.

 

When is a biscuit not a biscuit?

In other important news, McVittie’s asked the Biscuit Museum (yes, there is such a thing) to remove Jaffa Cakes from the premises. 

We’ll get to why in a minute, but first, for the non-British speakers among us, what’s a biscuit? It’s what Americans call a cookie–something round, sweet, and flat. And the Jaffa Cake meets all those criteria. What’s more, stores (shops in Brit-speak) stack Jaffa Cakes in the cookie display–or at least the one in my village does–since they come in the kind of packaging most cookies inhabit. 

Still, McVittie’s insists it’s a cake, and since McVittie’s is the maker of the flat, round, sweet little beast–well, you’d think they’d like the publicity of being in a museum, even such an odd museum, but they don’t.

Why not? Tax law. If the thing’s a cake it’s not subject to VAT. 

To what? Value Added Tax, which is a long name for a hefty sales tax. Plain biscuits? No tax. Chocolate-covered biscuits, though? Tax. Obviously. And Jaffa Cakes are chocolate-covered whatever-they-ares, so this is Important. McVittie’s and the tax office arm-wrestled over it in 1991, McVittie’s won, and it’s not backing down. They’ve been sold as taxless cakes ever since.

It used to be theologians who argued about how many angels could dance on the head of a pin.

 

A chatbot insists the dog ate its homework . . .

. . . but strictly in the metaphorical sense. 

Our story begins with someone–let’s call him Barry Smethurst, since that’s his name–waiting for a train, which is late, so he goes online and asks WhatsApp’s chatbot for TransPennine Express’s customer service phone number so he can ask what’s happened to his train. 

He gets a number. He calls the number. The number turns out to belong to someone who has nothing to do with TransPennine Express–he’s a property industry executive in Oxfordshire, 170 miles away, but he does use WhatsApp and his number’s on his company website, although Smethurst probably doesn’t know all that because he asks the bot what the number is. 

“Let’s focus on finding the right info for your TransPennine Express query!” the chatbot says cheerily and complete with exclamation point.

Smethurst no longer cares. He’s found something better to do while he waits: he wants to know why the bot gave him the number.

It was “based on patterns,” the bot says, and it would “strive to do better” in the future. The number, though, is “fictional.”

Bullshit, Smethurst doesn’t exactly say.

“You’re right, the bot says. It was “mistakenly pulled from a database.”

What database?

“I didn’t pull the number from a database. I generated a string of digits that fit the format of a UK mobile number but it wasn’t based on any real data on contacts.”

“Just giving a random number to someone is an insane thing for an AI to do,” Smethurst tells it, although what good he thinks scolding a bot will do is anyone’s guess.

That’s where the story drizzles away, sadly. Maybe his train pulls into the station. Maybe he sinks into a pit of despond so deep that he can no longer type. We leave him there, phone in hand, feet on platform, bot within thumb’s reach, but he’s given me an excuse to move to a more bizarre story about AI gone wrong, although it happened outside of Britain’s borders. 

A Norwegian man filed a complaint against OpenAI’s ChatGPT after he went online to look for information about himself and was told that he was in jail for killing two of his children. 

Yes, it’s nuts to think you’ll learn more about yourself by looking online than by putting down your phone and spending some time with your own non-electronic self, but we’ve all done it. 

Okay, most of us have done it. Or at least some of us have done it. Or–screw it, I’ve done it, although I don’t have kids so if anyone tells you I killed some of them, understand that it’s not physically possible. 

Of chatbots and culture wars and imaginary incidents

Posted on by
47

One of Britain’s reputable papers (and with five words, I’ve already eliminated several) had an incident involving chatbots, and the tale’s worth retelling because it tells us a lot about the age we’re stumbling cluelessly into. Or maybe that’s the drain we’re being washed down. Or–well, it’s Supply Your Own Metaphor Week here at Notes, so I’ll leave you to come up with your own while I waddle onward.

One of the Guardian’s reporters got an email asking about an article that ChatGPT had cited but that wasn’t showing up on the paper’s website. The email’s writer wanted to know what had happened to it and the journalist went hunting. It was on a topic they reported on,so it sounded likely enough although they couldn’t remember the specific article or find it anywhere, so they asked other people in the office to turn the paper’s electronic pockets inside out and see if it fell out. Maybe it was in there with the shredded kleenex and the linty mint.

Irrelevant photo: camellia

It wasn’t. Because it had never been written. It turns out that AI not only invents facts–something I trust you’ve heard by now–but it also invents sources, and it can be convincing when it does. The nonexistent article was a good enough invention that the journalist hadn’t been able to say, “No, I never wrote that.” They easily could have. 

If you think it’s scary living in a world where a lot of people feel entitled to curate their own selections of alternative facts to back up their pre-existing worldviews–well, it’s about to get a whole lot weirder. And, I expect, scarier.

 

Imaginary drag queen teaches hallucinatory sex ed class

Did anyone mention alternative facts? The Daily Mail, GB News, and Fox News all reported that a drag queen appeared as  a guest speaker at an Isle of Man schooll and told “11-year-olds there are 73 genders–and made a child who said there are ‘only two’ leave the class.”

Seventy-three? Stop it, guys. I can’t count that high. If this goes on, I’ll have to give up my leadership position in the Gender Hyperawareness and Conservative Freakout Society.  

The story went on to say that “one teacher is also said to have had to teach pupils in Year 7 and 8 how to masturbate.”  

How old are kids in years seven and eight? Eleven to thirteen. Since it’s been a long time since I was anywhere close to that age, I asked Lord Google how old kids are when they begin to masturbate. The top-ranked answer was from the National Institutes of Health (that’s in the US) and said two years old. The next one said three. In fact, most of the articles I found were geared toward calming the parents of toddlers and preschoolers, saying, essentially, It’s okay. Kids that young discover that there’s something interesting where their legs come together and they’re not shy about exploring it. 

That wasn’t what I’d been looking for, but it did back up my hunch that kids don’t really need to be taught how to masturbate, although by the time they’re eleven to thirteen they may need reassurance that what they’re doing–or at least imagining–isn’t so different from what other people do and imagine.

But that’s not the point. The point is, that although the article I quoted is real and can still be found on the Daily Mail’s website, the facts were invented. The flap the reporting caused led to an investigation of the incident, which found that the incident never happened. 

But who waits for that? As soon as the story went public, people working at the school were deluged with threats and demands for staff to be fired, arrested, and executed–not necessarily in that order. 

What triggered the story? A man who does occasionally do drag spoke to kids “gender neutral language and the concept of gender in the LGBTQ+ environment.” He wasn’t in drag, though. So the question is, if a person has done drag, can they be allowed out in public in non-drag or do they have to be freeze-dried, vacuum packed, and kept in storage until the political winds shift? For the safety, you understand, of all 73 genders of our children.

As for the kid who said there were only two genders, the closest I’ve found to the incident was one kid who was taken out of the room by a teacher over some sort of behavior issue. 

 

The problem of defining drag in Britain

Cranking up the British about men in drag is going to be harder than cranking up Americans, because drag has a solid mainstream history here. Every Christmas panto season starts, and these are shows for kids, with the lead female role always (over)played by a man and the lead male role almost always played by a woman. It’s a thing. Among straight people. Is that drag or is it only drag if a man (over)dresses like a woman outside of a panto?

What, while we’re at it, does a woman dress like? I’m wearing jeans, a turtleneck, and an old sweater.

On our first visit to Britain, we watched a race where a lot of the runners were in costume. It’s a thing here. Give people a chance to run five miles dressed  as bananas or phone booths and they’ll, ahem, run with it. So in among all the bananas and phone booths and chickens were men dressed as ballerinas and nurses. Not the contemporary kind of nurses who wear practical uniforms, but the old-fashioned ones in white dresses and caps, who (I gather) inhabit the fantasies of some unspecified number of non-nurses. My gaydar insisted that the runners in nurses’ uniforms were straight. But even if my gaydar was off–it was tuned in a different country, after all–no one much cared. It was just another race through the streets of an English city. Enjoy the show, everyone.

So where do pantos and dress-up end and drag begin? 

I don’t know, dear. You tell me.

 

The problem of defining copyright and privacy

Now that artificial intelligence scrapes information out of every corner of the internet so that it can tell you, in perfectly grammatical prose, that the pope is made of custard, defining copyright and privacy is going to be as problematic as defining drag. Or more so.

Copyrighted material is probably being used to train AI systems. The word probably is part of that sentence because AI’s neural networks aren’t available for your average gawker–or even your non-average one–to examine, so no one knows what they’ve been reading, but a couple of AI systems have, embarrassingly, hacked up copyrighted photos from Getty Images, complete with the watermark Getty prints over the photos so that users will have to pay for a clean copy. 

Yes, there’s a lawsuit involved, but it’s about the smallest edge of the problem. Still to be discussed is the amount of personal data that’s being collected–and potentially disclosed–without people’s consent and the use of copyrighted material to train chatbots.

 

But speaking of privacy

Teslas have an in-car camera that Tesla assures the world “is designed from the ground up to protect your privacy.” Because customer privacy “is and will always be enormously important to us.” 

So important that from 2019 to 2022 Tesla employees were sending each other clips of, oh, you know, interesting stuff in people’s garages; road incidents, a man walking up to his car naked; you know, ordinary, everyday stuff that would embarrass no one. 

What are the camera’s limits? I’m not sure, but I’ve read that a Tesla parked in the right spot outside someone’s house could, potentially, film whatever’s going on inside through the window. 

One owner is suing Tesla. Some Chinese government compounds and residential neighborhoods have banned the cars. 

The moral of this story is that if someone goes out of their way to tell you how carefully they’re protecting your privacy, they’re calling your attention to a problem.