What Happens When AI Directs Tourists to Places That Don't Exist? (bbc.com) 96
The director of a tour operation remembers two tourists arriving in a rural town in Peru determined to hike alone in the mountains to a sacred canyon recommended by their AI chatbot. But the canyon didn't exists — and a high-altitude hike could be dangerous (especially where cellphone coverage is also spotty). They're part of a BBC report on travellers arriving at their destination "only to find they've been fed incorrect information or steered to a place that only exists in the hard-wired imagination of a robot..."
"According to a 2024 survey, 37% of those surveyed who used AI to help plan their travels reported that it could not provide enough information, while around 33% said their AI-generated recommendations included false information." Some examples? - Dana Yao and her husband recently experienced this first-hand. The couple used ChatGPT to plan a romantic hike to the top of Mount Misen on the Japanese island of Itsukushima earlier this year. After exploring the town of Miyajima with no issues, they set off at 15:00 to hike to the montain's summit in time for sunset, exactly as ChatGPT had instructed them. "That's when the problem showed up," said Yao, a creator who runs a blog about traveling in Japan, "[when] we were ready to descend [the mountain via] the ropeway station. ChatGPT said the last ropeway down was at 17:30, but in reality, the ropeway had already closed. So, we were stuck at the mountain top..."
- A 2024 BBC article reported that [dedicated travel AI site] Layla briefly told users that there was an Eiffel Tower in Beijing and suggested a marathon route across northern Italy to a British traveller that was entirely unfeasible...
- A recent Fast Company article recounted an incident where a couple made the trek to a scenic cable car in Malaysia that they had seen on TikTok, only to find that no such structure existed. The video they'd watched had been entirely AI generated, either to drum up engagement or for some other strange purpose.
Rayid Ghani, a distinguished professor in machine learning at Carnegie Melon University, tells them that an AI chatbot "doesn't know the difference between travel advice, directions or recipes. It just knows words. So, it keeps spitting out words that make whatever it's telling you sound realistic..."
"According to a 2024 survey, 37% of those surveyed who used AI to help plan their travels reported that it could not provide enough information, while around 33% said their AI-generated recommendations included false information." Some examples? - Dana Yao and her husband recently experienced this first-hand. The couple used ChatGPT to plan a romantic hike to the top of Mount Misen on the Japanese island of Itsukushima earlier this year. After exploring the town of Miyajima with no issues, they set off at 15:00 to hike to the montain's summit in time for sunset, exactly as ChatGPT had instructed them. "That's when the problem showed up," said Yao, a creator who runs a blog about traveling in Japan, "[when] we were ready to descend [the mountain via] the ropeway station. ChatGPT said the last ropeway down was at 17:30, but in reality, the ropeway had already closed. So, we were stuck at the mountain top..."
- A 2024 BBC article reported that [dedicated travel AI site] Layla briefly told users that there was an Eiffel Tower in Beijing and suggested a marathon route across northern Italy to a British traveller that was entirely unfeasible...
- A recent Fast Company article recounted an incident where a couple made the trek to a scenic cable car in Malaysia that they had seen on TikTok, only to find that no such structure existed. The video they'd watched had been entirely AI generated, either to drum up engagement or for some other strange purpose.
Rayid Ghani, a distinguished professor in machine learning at Carnegie Melon University, tells them that an AI chatbot "doesn't know the difference between travel advice, directions or recipes. It just knows words. So, it keeps spitting out words that make whatever it's telling you sound realistic..."
How stupid does one need to be? (Score:5, Insightful)
How stupid does one need to be not to double check what an AI chat bot tells them? I mean, I even usually double-check what humans tell me. I'd sure double check that imaginary Peruvian canyon location and the time that Japanese ropeway closes at if I intended to get off the mountain before the night.
Selection pressure (Score:5, Funny)
How stupid does one need to be?
Stupid enough to breed, but not stupid enough to die before breeding.
Re:Selection pressure (Score:4, Interesting)
A 2024 BBC article reported that [dedicated travel AI site] Layla briefly told users that there was an Eiffel Tower in Beijing
This one is not quite as ridiculous as it sounds. About 800 miles (1250 Km) south of Beijing is the city of Tianducheng. It started as one of the Ghost Cities, underpopulated cities meant to resemble western locations like Paris, London, Germany, and others. https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fen.wikipedia.org%2Fwiki%2F... [wikipedia.org]
There are numerous replicas of the Eiffel Tower around the world. This website lists eleven of the most well known, including Tianducheng. https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fwww.unjourdeplusaparis... [unjourdeplusaparis.com]
Re:Selection pressure (Score:4, Insightful)
1250 km away is not ridiculously wrong to you? You're saying "well, it said the eiffel tower was in Rome, which is not as ridiculous as it sounds, there's an eiffel tower in Paris 1250km away." It would probably get this absolutely right 100% of the time if there was only one Eiffel Tower, but since there's a dozen replicas or more, its much easier to hallucinate.
Even if it was only 500km away its still wrong and deserves ridicule, specifically for a *travel focused* chatterbot that is supposed to help you plan a trip.
Re: Selection pressure (Score:2)
This is China. 1250 km is practically "the suburbs"
Re: (Score:2)
Maybe you're trying to be funny but its most definitely not close or 'the suburbs' for Beijing, any more than 800 miles is 'the suburbs' in the USA (unless you consider Jacksonville Florida a suburb of NYC)... Its easily 10+ hours or more by train to this 'Eiffel Tower', especially for a tourist with limited time, its absolutely wrong and not even close.
Re: Selection pressure (Score:2)
I wouldn't know, I'm in Europe. 1250 km for us spans 4 nations or so.
But I genuinely knew a Russian guy once who, when asked, always unironically said he was from Moskow. When asked "where exactly in Moskow", it turned out it was Niznhy Nowgorod. It's 500 km or so away, and it definitely takes the better part of the day to reach one from the other by car.
Re: (Score:2)
Though that might've just been him thinking "stupid Europeans know only Moscow, St. Petersburg and Novosibirsk. I'll just say Moscow, close enough."
Re: Selection pressure (Score:2)
I would hazard a guess that most Europeans haven't even heard of Novosibirsk either. I recently met the Russian wife of a parent at my child's school who also initially claimed to be from near Moscow (again, digging deeper, it turned out to be 500 km away.) And last month, one of my team members from near Novosibirsk (about four hours drive away) drove 7,500 km during his holiday. I said to him: wow, that's over 500 km a day, to which he replied: yes, but I was only driving for about five days of those tw
Re: Selection pressure (Score:2)
I guess the moral of the story is: depends on what you're used to.
Re: (Score:2)
Re: (Score:2)
It's two hours on the train from Sheffield, two and a quarter from Leeds.
Yes, provided that you can afford ~100+ quid for a ticket, live in Leeds near the station and the trains are all running on time. Even living close to Leeds like Harrogate, adds another 1+ hours each way without any other delays making a day trip much less practical especially given the extreme cost. That's also assuming that you are not arriving in London before 10am - if you are arriving before the cut-off the cost is 200+ quid.
So prehaps, if you are living in the middle of Sheffield, the closest city
Re: (Score:3)
If you're used to finding inaccuracy and accepting that, then you'll do fine, and have no regrets. For the other eight billion of us, we find this both immoral, and still another example of the immunity from liability by the bad advice from AI, itself an oxymoron.
This is the same logic leap that makes people believe that driverless cars are ready to put on the highways.
These are big lies, foisted by trillions of dollars in bad investments, trying to bring halo to trusting people who believe the big lies. Sm
Re:Selection pressure (Score:4, Funny)
Re: (Score:2)
Equilibrium status for each couple is actually more than four children but only the luckiest two (on average) survive to reproduce into the next generation. Ma Nature is big on equilibrium and small on exponential growth.
Re: (Score:2)
What proportion of the population thinks they need to double-check *everything* an AI chatbot tells them? They're being sold "artificial intelligence" and receiving a reasonable looking product. If they had to verify every single detail what would be the point of using the service?
It ain't "stupid", it's "average human".
Re: (Score:3)
If they had to verify every single detail what would be the point of using the service?
For me, the point is to find sources. If there's a [citation needed], then it has to be treated as unknown.
This is the same as Wikipedia. It's helpful because the internet has gotten less and less searchable, as spam has crept into Google and other search engines.
Re: (Score:3)
Yes but the basic punter is does not think like that. They go through life believing something is true unless they have information that it isn't. A lack of contradictory information is treated as the current information is true. And we all do that to some extent or even living would be impossible.
Re: (Score:2)
Re: (Score:2)
Correct.
If you are not a tech person, right now from all the hype and news and bullshit, you would rightly assume that AI is an amazing revolution and you would barely have heard about its shortcomings.
Re: (Score:2)
Correct.
If you are not a tech person, right now from all the hype and news and bullshit, you would rightly assume that AI is an amazing revolution and you would barely have heard about its shortcomings.
It's not just tech people. I know several writers, and even the ones whose work isn't tech-adjacent understand the perils and limitations of using LLMs. I also have cousins and friends who aren't even remotely connected to tech but who have enough sense - without me ever having broached the subject - to not take 'AI' pronouncements at face value.
I think the primary difference between AI suckers and the AI savvy is healthy skepticism. Some people are inherently gullible, and others aren't. Some who are gulli
Re: (Score:2)
What proportion of the population thinks they need to double-check *everything* an AI chatbot tells them? They're being sold "artificial intelligence" and receiving a reasonable looking product. If they had to verify every single detail what would be the point of using the service?
It ain't "stupid", it's "average human".
The issue is that we're forced to place trust in our tools because there is not sufficient time in the day to perform original research on every single topic. Arguably the reductio ad absurdum approach would be that one would have to go everywhere to verify everything with one's own eyes.
The problem is that publicly-accessible AI generally seems to be that lets us down because its responses look the same whether they're genuine and factual or whether the system has taken-in bad data or otherwise 'hallucina
Re: (Score:2)
I would think that to most people, booking a vacation based on one single source of information would be ridiculous. I don't care if it's an AI, a travel review site, or a personal recommendation, I would want at least a few perspectives. Gary Indiana has a tourism board, and the Visitgary.com website ignores the fact that it is an industrial town that has so much sulfur in the air that it smells like a fart. Why would anyone assume one single source contains the only facts that you need?
Re: (Score:2)
But the canyon didn't exists
The same kind of stupid that forgot to proofread TFS? /s
Re: How stupid does one need to be? (Score:2)
Re: (Score:1)
Re: How stupid does one need to be? (Score:4, Insightful)
Re: (Score:2)
FTFY
Re: (Score:2)
Re: (Score:1)
How stupid does one need to be not to double check what an AI chat bot tells them?
Let me guess never in your life have arrived at a shop and found it was closed, because the hours listed in the Yellow Pages were wrong, or they'd closed up in the last month. You have never arrived at the theater and found the should you wanted to see isn't there because someone did not update the website or put in the wrong information?
If you have double check everything the Chat bot tells you, why bother with it in the first place? Seems like a way a stupid person would waste a lot of time. The issue ap
Re: (Score:3)
Why indeed? You're right, that does seem stupid.
Re: (Score:2)
On the other hand I would not let GenAI plan my vacation.
"GenAI" - is that the one after Gen Z?
Re: (Score:1)
The same idiots who will use a system that hallucinates 30-60% of what it generates in business and productivity logic.
Companies are wasting trillions on this garbage when what they should be doing is investing a few million in on-the-job refreshing the skills of millions of domestic workers with 5-20yrs of traditional tech experience on cloud native tooling and adding dev to their traditional role toolsets, also be willing to cross train to different brand firewalls, etc. A wave of refreshing the skills of
Re:How stupid does one need to be? (Score:4, Informative)
Re: (Score:3)
You realize that people have been getting lost by blindly following their GPS system for decades now, right? And there have been many deaths as a result. It's not just Google Maps directing people down a closed
Re: (Score:2)
AI has invented a new variant of Pokemon Go. Gather points by taking virtual photos of imaginary destinations!
Re: (Score:2)
hell, even when a human enters the data, from the business, into 'google', that data can be wrong...as I discovered just 3 days ago when a restaurant with "recently updated hours" said they should be open until 10pm...and were closed when I arrived at 8:45.
so yeah, if i'm investing in a trip to another freakin' country, damn right i'm gonna have every scheduled event or location double-checked by contacting a human at in the know some point in the process.
Anyone who believes... (Score:2)
...guidance from an AI without independent verification deserves what they get
More likely to fall into an tourist trap that paid (Score:5, Interesting)
More likely to fall into an tourist trap that paid it's way to the top
Re: (Score:3)
And with Apple Maps that could even be a literal statement!
Stupid? (Score:3, Interesting)
Re:Stupid? (Score:4, Informative)
Never trust Google maps.
Re: (Score:2)
Never trust Google maps.
So lets be clear, this has nothing to do with AI's unreliability. It is the reality that all our sources of information are unreliable. You ought to have some level of uncertainty about everything.
Re: (Score:3)
Aside from that, we could get into the intricacies of identifying AI-generated content and argue about if it's stupid to be fooled, but it's irrelevant. You don't even have to know information was AI generated; it's stupid to not double check and to trust Google Maps or a Youtube video, because even if it was posted by a human, humans can lie
Re: (Score:2)
A Nigerian prince once told me.. (Score:5, Insightful)
...that I should just do a simple double-check on things I hear online, or from anywhere else for that matter. Especially if it's something I need to rely on, or if it seems controversial, urgent, surprising or high-stakes.
Re: A Nigerian prince once told me.. (Score:3)
It's spam. Don't listen to him.
hallu-delulu (Score:1)
When AI directs investors to nonexistent stocks... (Score:2)
The Itsukushima girl is an absolute Karen (Score:5, Informative)
Re:The Itsukushima girl is an absolute Karen (Score:4, Interesting)
Heh. You are right. https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fmaps.app.goo.gl%2Ftevw9p... [app.goo.gl] shows some of the easy path, and in the early evening it would also be a very lovely (and quite short) stroll! A situation where your GPS leading you typically astray became a welcome event instead.
Without taking away from your valid point:- I think the Aussie/UK slang "poor little diddums" fits her more than 'karen'. Karen is that well-known and quite typically horrible, screamy, middle aged woman usually demanding to talk to the manager as her face turns redder and redder, whilst diddums is more "dawww, you poor treasure, that must have been soul-shattering for you" :p
Re: The Itsukushima girl is an absolute Karen (Score:2)
Huh. Who would've thought.... Goes to show: you really should double-check everything on the internet. Including, but definitely not limited to, AI stuff.
Re: (Score:2)
The best path down Mt Misen isn't really paved all the way. It is a forest path, with stone steps.
They had set out to descend after sunset, and I don't remember seeing any lights on the path. Even a paved road can be dangerous in pitch black.
But I'd agree that it is stupid to rely on any third-party time table you'd find online.
Many travel sites and Google Maps often have outdated information, and often don't account for seasonal variations in opening hours. Many sites' own web sites don't get updated prope
Re: The Itsukushima girl is an absolute Karen (Score:3)
Re: (Score:2)
You site looks great, nice trick. I see it uses OSM data. Here is the direct link to the OSM map and hike (though it suggest another route), very similar look and feel at least without considering any paid features for Mapy. https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fwww.openstreetmap.org%2F... [openstreetmap.org]
Re: The Itsukushima girl is an absolute Karen (Score:2)
Re: (Score:2)
They had set out to descend after sunset, and I don't remember seeing any lights on the path. Even a paved road can be dangerous in pitch black.
This. I've had to descend a mountain as the sun was going down once (got stuck at the top due to weather for some time, and when it let up enough for a safe descent, it was late). It's absolutely not fun, even when there's still some light. Had it been dark, I think I would've taken my chances staying at the top rather than going down.
That said, anyone not a complete idiot checks things like "time of last cable car" a) in person, b) at the day, c) at the location. Because even there is an official website t
Re: (Score:3)
artificial intelligence is fed up with someone (Score:1)
Tell me what the point of AI is... (Score:5, Insightful)
That requirement just doubles the work, instead of cutting down the work.
Re: (Score:2)
Re: (Score:2)
This is the new "AI", which is just a higher-order statistics of words and their relations that have been found in text scraped off the web. .. and posts like this.
You know: racist rants, shitposts, memes
The older, classic, type of "AI", such as "expert systems" used to be based on highly curated information, input by experts in the field. With each datum and relation known to have meaning beforehand.
Re: (Score:2)
In a slightly perverted sense, I think these are sort of expert systems too, but their expertise is this: "What has been said, and how can I sound like the people who said it?"
Re: (Score:2)
You use AI to explore ideas, not give you hard facts. You use AI to give you outlines, not content.
AI is definitely not AI in the way everyone seems to think it is. It really shouldn't be called AI at all, but, just like Musk saying 'autopilot', you can't change it.
Finding is usually harder than verifying (Score:2)
Figuring out what places are worth going to on a trip can be a significant amount of work.
Verifying that a museum, restaurant, lake or mountain top exists and actually seems interesting is usually very easy.
I hate these stupid "what if AI .." questions. (Score:2)
What if AI figures out you're a fucking moron? What if the GPS gives you bad directions? What if your coffeemaker breaks? What if your anal beads stop working during a chess match with Magnus Carlsen? You do nothing. Does that work for you? You do nothing. Or, just replace AI with "what if a human .." .. "What if a human gives you bad directions?" You fucking punch him in the face and move on. Just do the same with AI. Dumbass. Humans do all the mistakes that AI does. So does any mechanical object that bre
Google sent me this way (Score:3)
AI edition.
Always tell your chatbot (Score:2)
Citation needed.
LLMs ... (Score:2)
... are literally word salad.
Sometimes the salad they toss together is interesting and useful - surprisingly often, when, say, rigorous languages and well defined goals are involved (as in programming). But sometimes is it is just plausible sounding words strung together.
Why anybody would use an LLM for say weather reports or (un-double checked) travel advice is baffling. It's just a word machine.
re: Google Maps directing people to non-places. (Score:1)
Re: (Score:2)
Before internet maps, one typically had a buy a roadmap of every new place they visited. I had an accordion-style folder in the back of my car with about 40 maps. For big metropolitan areas one could buy a Thomas Guide, which is essentially a book of maps. If one is just driving through a new area, then travel maps covered the roads adjacent to typical gas stops. Often these were given away free because they had hotel and restaurant ads.
Some did have mistakes. Boo boo's aren't new.
In the mid 90's one could
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
The good old way (Score:2)
Redirecting tourists where the traveling community hangs was always an old hobby of us
Things I am willing to ask an AI to do (Score:2)
I am willing to ask an AI to:
Tell me a story
Draw a fantasy picture
Talk to an unknown caller they give me their name.
That's about it.
If you had a friend or coworker (Score:2)
That was only accurate 70 to 80% of the time you would probably stop trusting them and not use them as much.
What if they went nowhere? (Score:2)
Well that would be your great chance to get away from it all.
"Welcome to the Darwin Award Resort" (Score:1)
"As a reminder, there are no refunds."
Happened to me many times. (Score:2)
I have been doing a lot of travelling for pleasure.
Google, TripAdvisor, Apple iMaps (whatever), Zomato, Expedia, AirBnB, etc.
On many occasions I been directed to restaurants that didn't - and never could have - existed.
This was without directly consulting an AI, but perhaps those sites are populated with AI-generated data?
Why even travel? (Score:2)
We can remember it for you wholesale.
I guess the same thing... (Score:2)
...as when maps to the same thing or GPS Navigators lead heavy trucks onto small forest trails or tourists into the ocean.
They teach them a lesson about human stupidity.
What happens when your navi is wrong? (Score:2)
Many years ago someone wanted to sue the navi company as they drove into a lake because they looked at the navi instead of at the road.
Use your brain.