AI & the crazy results

Hi Jean. I have been totally underwhelmed with AI trip advice. Like you, I have toyed with it on locations and routes I know personally. I will stick with doing my own investigating,

It only has a chance when there is a huge amount of previous material to train on. Any less popular or niche subject won't have that. If they could even scrape the material nowadays. That's getting harder to do.

I toyed with it a bit when planning a US road trip for this past July. It suggested I stay at a hotel that didn't exist

Posting examples of wrong AI results seems to be a popular subject on this forum. I think people forget sometimes, that we could could also very easily fill an entire thread with examples of incorrect or outdated information in the RS or other guidebooks. Or with examples of incorrect information posted here on the forum….

I regularly use AI in my trip planning and I find it very helpful. I for example asked AI how to get from the airport in Lisbon to my hotel using public transport. It gave me directions, told me where to change and ways to pay to the ride. I always stick to my rule; “one source isn’t a source “ and double checked the information, it was 100% correct.

“one source isn’t a source “

Thank you Dutch Traveler, I'll add that excellent phrase to my vocabulary.
My multiple sources do not include AI for lots of reasons.
But sometimes one source is ok. If I'm on the AAL website planning a trip, I don't need to know if G**gle knows about the flight to Dublin, I'm looking at the real AAL schedule from AAL.

My most recent AI experience was an eye opener. I asked ChatGPT to Create a map locating the ten most popular tourist sights in Budapest. It placed 6 out of the 10 in the wrong locations. And not just a little wrong ... a lot wrong. Why would it trust it to plan a day on holiday? No telling where I would end up.

A few weeks ago when this last cropped up, I specifically asked AI about Cle Elum (where I was meant to have been this week on the way to CdA) and the results were far from impressive.
Two examples- I asked it where the nearest grocery stores are and got a wrong answer omitting the newest one (I have just checked, it is still out of date, but then Google Maps is also out of date and doesn't show the store either).
Secondly I have just asked Gemini about nearby Episcopal Church Services this week (knowing the answer)- it gave me the correct website, but omitted tonight's service. Argh.

So I asked it about events in CE this week- guess what it omitted the biggest event of the week, one which was key to my week, and I even have an invitation to.

But neither this forum or TA shines either. The last time I saw CE mentioned on here 40 year old information was given, and on TA an OP had clearly done their research (in that case about the Teanaway) and was flatly told by all the 'west siders' that their information was wrong when it wasn't.

At least AI (when it is accurate) is unbiased, in the forum and TA cases above the responses showed personal bias, and where they think enquirers should go to.

At least AI (when it is accurate) is unbiased,

I think that is a good point in favor of AI. On a forum, we all have our biases (I'm happy to use cash vs. I'm proud to never use cash, etc...).

But in the future, will AI have biases? I think they will. AI exists for the purpose of making more money for the billionaires, and spending less money on the workforce they replace. So is it not possible that one day, an AI owned by a U.S. conglomerate, has instructions that if there are two similar routings to London, the AI should show the United Airlines flight rather than the British Air flight? More likely, the conglomerate owns shares in United Airlines (or United Airlines paid the conglomerate for the preferential treatment).

Crazy idea, huh? Most of us (not me) use G00gle to search for anything. Ya think G**gle might give you search results heavily biased on who paid to have a higher lisiting?

Perhaps biasing the results is an intended revenue stream for the AI provider, exactly as it is for most search engines. It comes down to trust, and there's not much of that these days.

"AI" as currently available is an extremely impressive search engine, but like any search engine it gathers enormous amounts of data very quickly without grading the quality of what it is gathering and then spits out a probability-based output, depending on its training sets. Anyone who has been confused by the multiple rail stations in a single European City, or the fact that the same names (or a version of them) exist in different parts of a city/area will not be surprised that a Chatbot/AI aggregator will also mix and match things that are not the same. For travel planning AI/Chatbots will essentially zip through every guidebook in existence on the web and come out with results, ie. recommendations. I don't think it will turn out to be much of a shortcut for folks because you will never be able to rely on it for accurate information for an individual trip based on a person's actual interests rather than what every guidebook says you should do.

Maybe the Chinese ones give better results?

What If I said:

President xxxx is a brilliant guy.

or I said:

I'm SURE that President xxxx is a brilliant guy. Haha.

Would AI consider both of those statements to be identical?
Does AI do nuances, or sarcasm? Just curious.
From what I wrote, humans would infer something about me, but would AI be able to do that?

It is absolutely not accurate to say that AI (or, more specifically for this discussion, LLMs) are unbiased. They are trained on large datasets of information generated by humans, who have their own biases. They are then given weights and programmed by humans, who, again, have their own biases. They absorb all of this along the way.

I'm going to paste something Bostonphil7 just wrote on another thread:

"I have read that booking may have bit off more than they can chew by offering other services besides lodging."

Would AI recognize that "booking" refers to "booking.com"?
Would AI know the meaning of "bit off more than they can chew"?

I’m planning a trip to Rocky Mountain National Park and Google Gemini confidently told me to take Old Fall River Road to get back down the mountain…problem is, that road is one-way, up the mountain only.

I agree with Dutch_traveler. AI bashing is a popular pastime on this forum. I have my hypothesis as to why that is, but I will keep it to myself. Personally, I have had good experience with AI in travel planning. I posted a specific example in another thread. Use it, don't use it, your choice. 'Trust but verify', as a former President used to say.

Dutch_traveler, I agree. All I can say is that the responses depend on the prompts you give it. Just like any other tool, you have to know how to use it.And quite frankly, I don't think people are always willing to take the time to learn.

That's fine of course, but keep in mind that if you don't take the time to learn how to use it, then you're going to be stuck with some bad results. For those who do take the time to phrase a well-thought out prompt and feed AI the info it needs, it can be an invaluable resource. I can't tell you how much information I am getting on my current trip to Scotland and how much time I am saving by having AI generate this information. Of course, I check the information, just like I would check it in any guidebook or check responses here on this forum. And I do find occasional errors. But certainly not enough to make me stop using it.

There was a good article in the New York Times the other day that gave some tips on using AI for travel planning. It's definitely worth a read if you're serious about using it. AI Travel Planning Tip Prompts

Mardee, i wouldnt call pointing out flaws and recommending checking or not relying on results "bashing". Well no worse than we bash each other on the forum.

I will say that I discovered those 6 buildings that AI put on the map in the wrong locations was not the fault of AI. Apparently a 150 years ago the Hungarians built them in the wrong locations. Its Orban's fault, not AI.

What does truly scare me is when AI cites a RS Forum post as a source.

I agree that this post and others like it are not AI bashing, They are AI warning.

I've seen many users on a variety of social media, here and elsewhere, swear that AI, Gemini, ChatGPT or whatever is the final answer regardless of subject.

I copied the thread and gave it to ChatGPT with the question. ChatGPT said it wasnt bashing: end of discussion.

...if you don't take the time to learn how to use it, then you're
going to be stuck with some bad results. For those who do take the
time to phrase a well-thought out prompt and feed AI the info it
needs, it can be an invaluable resource.

The same could be said for well-thought out questions on these forums? As much of the "W" info provided as possible - WHEN planning to travel; WHERE planning to travel; for WHAT length of time /WHAT interests in seeing/doing; WHO is taking the trip - the better responders can help customize a trip. Not so dissimilar?

Mardee, I wonder if AI provides small but important details such as, oh, a day trip that's not the best for a city whose museums are closed on that particular day?

For those of you bashing AI, if you hear a knock on your door tonight, make sure it isn't Arnold Schwarzenegger before you open it.

Mardee, i wouldnt call pointing out flaws and recommending checking or not relying on results "bashing".

I am fairly certain I did not us the word "bashing," Mr. E. Did you find it someplace in my post? I'm looking at it, and I don't see it. I merely pointed out that having a better understanding of AI and researching how to ask questions helps get better answers. 😊

What does truly scare me is when AI cites a RS Forum post as a source.

Ha ha, I think it's scary that you think it's scary. Does that mean that you don't trust the Rick Steves Forum posts? It also pulls from Reddit, TripAdvisor, blogs, and other resources.

Mardee, I wonder if AI provides small but important details such as, oh, a day trip that's not the best for a city whose museums are closed on that particular day?

Kate, It really depends on how much information it has, and how the question is worded. Like I said, the more information and detail you give it, the better the response. For example, if I were asking it that question, I would probably phrase it something like this.

I would like to go on a day trip to Edinburgh, while I'm staying in Glasgow. I am especially interested in visiting museums while I am there. Please give me some suggestions for days that would be best for traveling there, especially considering opening days and times of the major museums.

Ha ha, I think it's scary that you think it's scary. Does that mean
that you don't trust the Rick Steves Forum posts? It also pulls from
Reddit, TripAdvisor, blogs, and other resources.

No, I dont trust the answers as fact. But you do?

"Bashing" was in your a post i thought was yours along with a statement that not you suspected you knew why, but were going to keep it to yourself. If i was hallucinating then I apologize. Again, my apologies.

Mr E; you might want to check who actually wrote the post about bashing as it wasn’t Mardee.

All sources of information are error prone. Guidebooks, traditional web search, or AI. Google Maps has errors, as do all the other online maps. Plus, if you follow enough travel groups on Facebook or elsewhere, you'll read of countless examples of people being given wrong information by airline employees, hotel employees, travel agents, and other humans. The humans who post on this forum make mistakes too, believe it or not!

It's easy to single out AI because I think we expect it to be better. But first off, AI technology is still in its infancy. And second, where does AI get its information? From all those other error-prone sources.

Enjoy doing your research by any means you like. But wherever you get your information, be sure to verify it for accuracy.

Then I stand corrected. Mardee my apologies which is why i always leave room for potential hallucinations.

No worries, Mr. E. I think many of us are at that age where we can't tell the difference between reality and hallucinations, lol!

I hallucinate about where I left my reading glasses all the time!

There you go...

I use duck.ai, which gives gives you a choice of several free AI models. I find the default model to be very useful when using carefully considered questions. Of course, double-check, as you should with any information garnered on this forum.

I received good and accurate answers for a couple of Jean's questions above.

It also shares my view of the term "must-sees" - "vague, overused, and faintly breathless."

Thank you Mardee, you are the best.

I've also become a fan of AI as a preliminary search tool to compile ideas. From there I do my own fact checking, but it's certainly a quick way to gather info.

And I do find occasional errors.

I can't recall the situation but I called out AI for some false info it gave me on something travel related. I called it out as being wrong and to try again. It actually apologized to me and said it was wrong and I was right and it corrected itself. I suspect this is how AI learns.

It actually apologized to me and said it was wrong and I was right and
it corrected itself.

Gotta love AI with manners!