Openai ChatGPT

Joined
May 28, 2024
Posts
1,045
Likes collected
4,492
Location
Maspalomas, Spain
Funster No
103,439
MH
Giottiline Toscan 74
Exp
2024
Has anyone used ChatGPT or similar for route planning? I thought I would have a look while waiting for my Moho pickup as I'm planning a trip from Gran Canaria to Omagh in Northern Ireland in September.

I introduced instructions that I wanted to drive a maximum of 100-150km a day, wanted to avoid motorways and toll roads, and wanted to overnight in 'Aires' both in Spain and France. I initially put the destination of Folkestone just to see what it came up with.

I got back (in around 10 seconds) a very comprehensive route guide that even included what to see once stopped in an 'aire' I probably will fine tune this, but I'm very impressed with the results. I'll attach the print out in case anyone wants to check it out.

Please remember I am new to all this and am not familiar with routes through Europe and as a newbie thought this is a brilliant resource.

All comments (good or bad) are welcome.

I wasn't sure if this is the correct forum to post this in, so if it isn't, please advise.

All the best.
 

Attachments

Worth noting that ChatGPT and all other LLM do their best to please you and will even lie to make you think it has done well.

Verify everything it tells you to avoid being sent somewhere unsuitable. Not trying to put a downer on it, it is a very useful tool but it does lie.
 
At first glance looks wonderful. However, having recently travelled Seville, Cordoba etc, there are not aires called “ Aire de Cordoba” - or Aire de xxxxxx. It has put that for every town so they are not real places. The towns and routes may be OK but look for yourself on app such as Park4Night to find the actual aires and read reviews.

Subscribers  do not see these advertisements

 
I've started using bard (Gemini) and the bing image generator

Never thought about using for routes. Thanks

Trust it....no way not yet
 
Worth noting that ChatGPT and all other LLM do their best to please you and will even lie to make you think it has done well.

Verify everything it tells you to avoid being sent somewhere unsuitable. Not trying to put a downer on it, it is a very useful tool but it does lie.
Yes, of course. But as a starter it could be helpful for me.

Subscribers  do not see these advertisements

 
At first glance looks wonderful. However, having recently travelled Seville, Cordoba etc, there are not aires called “ Aire de Cordoba” - or Aire de xxxxxx. It has put that for every town so they are not real places. The towns and routes may be OK but look for yourself on app such as Park4Night to find the actual aires and read reviews.
Ok, good tip and thanks for the input.
 
An AI bot, now owned by Microsoft. This is the begging of the end of humanity probably.
Sorry but Microsoft does not own even a little bit of OpenAI. They are entitled to a percentage of the profits due to the partnership agreement, but they do not have any ownership at all.
 
I realise that you want to avoid motorways etc but that route is still nearly 40% longer than one churned out by Google maps requested to avoid motorways etc........2000km as opposed to 2800+km....or did you ask for all the hotspots...?

Subscribers  do not see these advertisements

 
Like any tool it should only be used as an aid, whether chatgpt or a gps always verify what it has told you, i learnt this a long time ago.

I was using google maps to go from Bristol to Hérault, had never been through the tunnel before so thought I would try that route as an option, I was suprised to discover that it only added about 50 miles to my route, so decided to do that route without further verifying something that i was already questioning in my head.

After what felt like a very long drive home, I checked again on google maps, yep only 50 miles further, still puzzled i split the journeys up. Bristol-Portsmouth, Caen-Hérault, then Bristol-Dover, Calais to Herault, the result added up to well over 100 miles different, google had included the distance travelled in the tunnel, and the ferry, not taking into account the fact that you are not actually driving that bit.

As i said it is very easy to take tools for granted and get caught out.

Like they say in carpentry measure twice and cut once.
 
I realise that you want to avoid motorways etc but that route is still nearly 40% longer than one churned out by Google maps requested to avoid motorways etc........2000km as opposed to 2800+km....or did you ask for all the hotspots...?
No ianp, I was merely checking out the possibility of using AI. Your suggestion using Google maps is good, thank you for your input. I will check this out.
 
Like any tool it should only be used as an aid, whether chatgpt or a gps always verify what it has told you, i learnt this a long time ago.

I was using google maps to go from Bristol to Hérault, had never been through the tunnel before so thought I would try that route as an option, I was suprised to discover that it only added about 50 miles to my route, so decided to do that route without further verifying something that i was already questioning in my head.

After what felt like a very long drive home, I checked again on google maps, yep only 50 miles further, still puzzled i split the journeys up. Bristol-Portsmouth, Caen-Hérault, then Bristol-Dover, Calais to Herault, the result added up to well over 100 miles different, google had included the distance travelled in the tunnel, and the ferry, not taking into account the fact that you are not actually driving that bit.

As i said it is very easy to take tools for granted and get caught out.

Like they say in carpentry measure twice and cut once.
Very good points to take into consideration. Thank you.
 
AI regurgitates information it's been trained on and tries to fill the gaps. If it has lots of good info on the subject that's well connected, you'll generally get a true and useful response. If your question doesn't have a lot of training material, or makes it try to link sets of information that aren't well linked... it'll make it up. Referred to as hallucination. It can't tell the difference. It has no concept of the real world, it just chains together words in a common pattern (hence Large Language Model). It'll sound just as plausible when it outputs trash as when it's being useful. Be careful. They are useful tools. But never trust them.
 
AI regurgitates information it's been trained on and tries to fill the gaps. If it has lots of good info on the subject that's well connected, you'll generally get a true and useful response. If your question doesn't have a lot of training material, or makes it try to link sets of information that aren't well linked... it'll make it up. Referred to as hallucination. It can't tell the difference. It has no concept of the real world, it just chains together words in a common pattern (hence Large Language Model). It'll sound just as plausible when it outputs trash as when it's being useful. Be careful. They are useful tools. But never trust them.
Absolutely.

Subscribers  do not see these advertisements

 
Can anyone tell me what's the difference between this new thing and the normal Google that's talked to you for 10 years.?
 
Can anyone tell me what's the difference between this new thing and the normal Google that's talked to you for 10 years.?
Google Assistant is voice recognition, with the result fed into an static and preprogrammed pattern of outcomes. If you didn't ask in the right way, it'd be clueless.

The outcomes for Bard/Gemini aren't pre-programmed. It guesses completely itself based on being fed billions of pieces of information and joining the dots itself. It's far more flexible and useful... and far more unpredictable.
 
Some of the legs are 30-40 km - far short of what you requested.
 
Looks interesting for a bit of fun at least.

Subscribers  do not see these advertisements

 
chaser Did you not read the recent posts on Ian's thread, it was used on there to give an idea on his request for a begging letter?
 
Are you sure that’s correct?
Yes.


1718236948542.png


Microsoft. Minority Economic Interest on this chart.

1718237089813.png
 
Back to Barriesimpson ’s initial question and maybe a little about chaser ’s too.

ChatGPT and other similar systems are “large language models” they ingest large amounts of text and build a statistical model of how text is put together. When you ask them a question they generate a “likely response”. A simplistic example “how do i drive from birmingham to manchester” a very likely response is “go up the M6” because the internet is full of answers like this. ChatGPT is actually working at a lower level than this using sub words and iterating over its partial answer but it’s useful to think of it like this.

That means they’re very good at some things but they have real weaknesses: specifically they don’t “understand” the world or what they’re saying. Worse they do sound very fluent and confident so it’s easy to mistake them for intelligence; the AI branding doesn’t help either :-(

So, they are useful and generally right but they do make things up (also known as hallucinating) and when they do, they sound very confident and believable. These hallucinations tend to be in things like the titles of scientific papers but also real world distances etc. “Aire de Alicante” is a perfect example of this.

They also get confused and regurgitate internet memes as fact because those memes are repeated a lot on the internet (e.g. use glue to keep cheese on a pizza).

So, they’re a good starting point for route planning but they don’t “understand” distance and traffic etc in the same was as Garmin and Google maps do. They do have a good chance of spitting out good campsites and tourist attractions because Google maps data and probably MHF were used to build their statistical model.

Subscribers  do not see these advertisements

 

Join us or log in to post a reply.

To join in you must be a member of MotorhomeFun

Join MotorhomeFun

Join us, it quick and easy!

Log in

Already a member? Log in here.

Latest journal entries

Back
Top