AI being used by TripAdvisor

[The Colorado Kid]

Might have been going on for awhile. I just noticed this Traveler Insights section in TripAdvisor is using Artificial Intelligence to summarize reviews:

The TripAdvisor rating in general for Outer Banks Beach Club is 4.5 stars

 

[jorcus]

Interesting but how useful is it? Is it all too generic? I was listening to a tech CEO on tv and he made the point that AI is only useful if people trust it. Getting over the wall of doubt is a big hurdle. People write a lot of stuff on tripadvisor that you always have to sort though to make a real evaluation. When I look for information I want it from people who have posted more than once. I think most of us can tell who is giving a fair description of a resort as opposed to the flip post of I liked the bartender. I find many of people who have long negative reviews had special needs or special problems. That may be relevant if I have that need but resorts may not be all things to all people. I am no where near the point where I would trust a generic AI review vs a well thought out post by an experienced traveler.

 

[The Colorado Kid]

Interesting but how useful is it? Is it all too generic? I was listening to a tech CEO on tv and he made the point that AI is only useful if people trust it. Getting over the wall of doubt is a big hurdle. People write a lot of stuff on tripadvisor that you always have to sort though to make a real evaluation. When I look for information I want it from people who have posted more than once. I think most of us can tell who is giving a fair description of a resort as opposed to the flip post of I liked the bartender. I find many of people who have long negative reviews had special needs or special problems. That may be relevant if I have that need but resorts may not be all things to all people. I am no where near the point where I would trust a generic AI review vs a well thought out post by an experienced traveler.

I agree...I also heard one AI analyst say that after a few years...all AI will sound the same because it will all be sourced from....other AI messaging

 

[Ken555]

Yes, I saw the same yesterday when I was deciding on a hotel for an upcoming trip. It’s a practical use for AI.


Sent from my iPad using Tapatalk

 

[dioxide45]

I just listed an item on Ebay and there was an option to have AI write my description. I used it and it was very generic. AI so far isn't all that intelligent and really isn't really AI in the sense that we think about AI. It certainly isn't living up to the hype. The question is, when will the bubble pop as it seems these tech stocks are the only things keeping the stock market as high as it is.

 

[easyrider]

I have been using AI for ideas and it seems to always have some. I'm currently using googles Bard which has given me a few leads. While it isn't perfect it does produce ideas and clues. I think AI will only get better with time. The problem might be that many people will stop thinking and just rely on AI.

Bill

 

[Tia]

The AI on my current phone when I text is always changing what I wrote

 

[PigsDad]

The AI on my current phone when I text is always changing what I wrote

Umm... that's autocorrect and not AI. Two different beasts.

Kurt

 

[jp10558]

I've been using some AI tools for a bit, and I think people both under and over estimate it a lot. LLM tools like Claude2 are really good at looking through docs and getting answers... most of the time as long as they're not hallucinating. Like with simple scripts (computer programs sort of) or needing to know the one command for this vendor vs that vendor tool - it works pretty well IMO. What can trip it up is when there's a paid and a free version of the same tool - it doesn't seem to know to ask you which you're using, and can conflate the things like file locations.

I find for summarizing they do pretty well. But to some extent there's been computer summarizing for almost 30 years of varying quality, tending towards the better over time. So I think processing reviews for a consensus is something it can do well. What would be better is also just checking the review meta or competitors for whether the reviews seem to be fake or not (or somehow be able to do that in line). Taking into consideration how many reviews are 1 star vs 5 star, and looking at the review quality. A lot of that can also be done with traditional systems because there's currently pretty simple heuristics in use anyway, but with some of the processing LLMs can do should be able to make it better.

So in a lot of ways, as an assistant I think a lot of people underestimate the tools. In a lot of ways they're also really helpful for things like D&D games (not a business case obviously) - coming up with better flavor text, instant NPC generation, Quickish image creation...

Where AI falls down is still the higher level cognitive decision making. I haven't had a lot of luck feeding it a question like xxx error message on yyy computer - what should I do? It tends to come up with the same basic sorts of things that low effort forums would - probably cause that's what it's pulling from. It often creates like the most generic Microsoft or whatever KB sort of page of text that rarely if ever actually helps, but does get you to reboot, reinstall, repair some stuff a few times with no fix. Given that, I probably wouldn't currently rely on it to fully solve a problem the way I could an experienced colleague.

But now that I have some contractual private AI access, I'll be testing some stuff - I haven't really fed it any info to start with yet.

 

[Tia]

Umm... that's autocorrect and not AI. Two different beasts.

Kurt

Okay, well it's annoying for sure

 

[easyrider]

Okay, well it's annoying for sure

I agree.

Bill

 

[Talent312]

I've been told that one fast-food chain is using AI
as an aid to help with interpret verbal orders.
It's responses will Include:
"110 double cheeseburgers. Is that correct?" and
"Do you want fries with that?" when appropriate.