Using AI for trip planning

[Sikkun]

I’ll just say I do research on this type of stuff for a living…and the LLM’s deff make up math all the time.

Just a caution.

Personally I’d just use the tools actually delete to solve these problems.

Sponsored

 

[Sikkun]

I’ll just say I do research on this type of stuff for a living…and the LLM’s deff make up math all the time.

Just a caution.

Personally I’d just use the tools actually designed to solve these problems.

 

[ChrisO]

One thing I do see is different AIs are better for different tasks.

And it is like the discussion on the crash prevention systems. Blindly using either without understanding their limitations can get you in trouble.

I would no more trust ABRP 100% than I would trust what an AI said 100%.

I’m a programmer and I can tell you that AI is quite good for research and sometimes simple examples, but the people that think they are going to be turning out any kind of mildly complicated software are fooling themselves at this point.

I take this use case the same way just another tool to do research with. Find out what it does do good and use that but also watch for mistakes and plan accordingly.

Back in the “ancient times” when expert systems were the ”ai” which were basically rule based programmed in, I always said that you wouldn’t have true AI until we allowed it to make mistakes. I take this to be a requirement, just as I take it as a requirement that I can’t believe 100% of what any given person says.

 

[Tampamike]

I’ll just say I do research on this type of stuff for a living…and the LLM’s deff make up math all the time.

Just a caution.

Personally I’d just use the tools actually designed to solve these problems.

Speaking of making up math, …

Did anybody else notice that the arrival % at Santa Ana was wrong? It has you leaving with an 80% charge for that last short leg of 50 miles and 17% battery depletion. My free-range intelligence comes up with 63% remaining on arrival not 13%.

 

[Mach1E]

One thing I do see is different AIs are better for different tasks.

And it is like the discussion on the crash prevention systems. Blindly using either without understanding their limitations can get you in trouble.

I would no more trust ABRP 100% than I would trust what an AI said 100%.

I’m a programmer and I can tell you that AI is quite good for research and sometimes simple examples, but the people that think they are going to be turning out any kind of mildly complicated software are fooling themselves at this point.

I take this use case the same way just another tool to do research with. Find out what it does do good and use that but also watch for mistakes and plan accordingly.

Back in the “ancient times” when expert systems were the ”ai” which were basically rule based programmed in, I always said that you wouldn’t have true AI until we allowed it to make mistakes. I take this to be a requirement, just as I take it as a requirement that I can’t believe 100% of what any given person says.

It’s not about being right 100% of the time.

The problem with AI in its current state is that it speaks with 100% confidence but is frequently wrong.

It’s basically just researching the internet and repeating whatever it finds, right or wrong. The source data is the problem. The internet is filled with bad and contradictory information.

I trust it to search basic stuff. But anything technical or complicated? It’s more likely to be inaccurate.

To the question at hand, I wouldn’t expect it to know which charging stations are compatible nor what’s available. But there are other existing systems that can…… and they don’t give you a 5 page essay explanation either.

 

[ChrisO]

It’s not about being right 100% of the time.

The problem with AI in its current state is that it speaks with 100% confidence but is frequently wrong.

It’s basically just researching the internet and repeating whatever it finds, right or wrong. The source data is the problem. The internet is filled with bad and contradictory information.

I trust it to search basic stuff. But anything technical or complicated? It’s more likely to be inaccurate.

To the question at hand, I wouldn’t expect it to know which charging stations are compatible nor what’s available. But there are other existing systems that can…… and they don’t give you a 5 page essay explanation either.

A lot of people speak with “100% confidence“ but I certainly have learned to not believe everything they say.

At any rate I like everyone to know that I mostly just found this use for AI interesting and maybe a bit useful, I’m certainly not trying to push anyone to use it.

 

[Mach1E]

A lot of people speak with “100% confidence“ but I certainly have learned to not believe everything they say.

Exactly. And that’s where AI seems to get its data from.

A couple months ago there was a thread on this forum where someone asked a random question.

A few posts later someone used AI to get an answer……. And the source was that exact thread.

 

[MellowJohnny]

No. “AI” is just a glorified search engine that parses and summarizes results in a pretty format in plain English.

Depends what you ask it to do.

Lets say you have a flat organisation of Products but your eComm site needs a hierarchy of parent Products and child SKUs (variants). When we browse a web site we want to see a Product page with all the possible colour & size combinations (the SKUs). I can't do that if I don't have SKUs. In the "old days" this would take a lot of code & testing to figure out how these products are related, define a Parent Product, and associate all the SKUs to it.

Now a decent LLM can do it in seconds.

You give it this as an input - five Products, no variants:

Product Name,Material,Color,Size,SKU
Glove,Leather,Black,Small,G001
Glove,Leather,Black,Medium,G002
Glove,Leather,Black,Large,G003
Glove,Leather,Black,XL,G004
Glove,Leather,Black,XXL,G005

And it gives me this - a Parent / Child relationship of one Product with 5 variants that it figured out on its own:

{"Glove - Leather - Black": {
"variants": [
{ "Size": "Small", "SKU": "G001" },
{ "Size": "Medium", "SKU": "G002" },
{ "Size": "Large", "SKU": "G003" },
{ "Size": "XL", "SKU": "G004" },
{ "Size": "XXL", "SKU": "G005" } ]
}
}

If you use agentic AI agents like a glorified search engine you will get results that look like a glorified search engine.

Ask it to do complex stuff and you have a powerful tool.

 

[Sikkun]

Depends what you ask it to do.

Lets say you have a flat organisation of Products but your eComm site needs a hierarchy of parent Products and child SKUs (variants). When we browse a web site we want to see a Product page with all the possible colour & size combinations (the SKUs). I can't do that if I don't have SKUs. In the "old days" this would take a lot of code & testing to figure out how these products are related, define a Parent Product, and associate all the SKUs to it.

Now a decent LLM can do it in seconds.

You give it this as an input - five Products, no variants:

Product Name,Material,Color,Size,SKU
Glove,Leather,Black,Small,G001
Glove,Leather,Black,Medium,G002
Glove,Leather,Black,Large,G003
Glove,Leather,Black,XL,G004
Glove,Leather,Black,XXL,G005

And it gives me this - a Parent / Child relationship of one Product with 5 variants that it figured out on its own:

{"Glove - Leather - Black": {
"variants": [
{ "Size": "Small", "SKU": "G001" },
{ "Size": "Medium", "SKU": "G002" },
{ "Size": "Large", "SKU": "G003" },
{ "Size": "XL", "SKU": "G004" },
{ "Size": "XXL", "SKU": "G005" } ]
}
}

If you use agentic AI agents like a glorified search engine you will get results that look like a glorified search engine.

Ask it to do complex stuff and you have a powerful tool.

If your starting with that CSV I am confused why any “AI” was needed to get that JSON quickly.

 

[bbulkow]

No. “AI” is just a glorified search engine that parses and summarizes results in a pretty format in plain English.

Web search doesn't do multiple searches on my behalf and summarize the results, which is critical now that most top search results are ai generated drivel that has you reading page after page before getting any answers.

Go ahead, don't use ai. Enjoy your transistor am radio too! Great technology.

 

[ChehRob]

AI is a tool. The poster reported how he used it and what he found helpful. Why all the dislikes? No one is telling anyone what to do.

 

[ChrisO]

Exactly. And that’s where AI seems to get its data from.

A couple months ago there was a thread on this forum where someone asked a random question.

A few posts later someone used AI to get an answer……. And the source was that exact thread.

I know, I’m the one that posted the CoPilot response and pointed out that it was just came from the very thread that I was curious about getting more “viewpoints” on.

 

[ChrisO]

My viewpoint on AI is this. There is a lot of it is hype and it is being setup for a big fall, but on the other hand it is a great tool when used within its limitations. So, it is both a mistake to think it will do everything for us and a mistake to think that it is all fake and useless.

 

[Mach1E]

My viewpoint on AI is this. There is a lot of it is hype and it is being setup for a big fall, but on the other hand it is a great tool when used within its limitations. So, it is both a mistake to think it will do everything for us and a mistake to think that it is all fake and useless.

I think it’s somewhere in the middle. It is a “better search engine.” It also can summarize things quickly and help your write emails etc.

People also seem to either blindly trust it too much or blindly distrust it too much.

And the more mistakes it makes and bad info it sends, the more it becomes distrusted.

I read an interesting article a few months back about how AI could lead to the “death of the internet.”

TLDR; As more online content becomes AI written and less “real human,” the internet will become an AI echo chamber and people will stop using it if they don’t trust it.

Examples from the past- Amazon reviews before they cut down on the fake bot reviews. Bots nearly killed Amazon’s trust.

Craigslist- used to be trusted and used by everyone, then it became mostly scammers and it’s barely used anymore.

The entire internet could end up that way. If you no longer trust the articles you read or the reviews or search results…….. you’ll stop using it.

 

[ChrisO]

BTW all this planning got overridden by a higher power, my wife told me that she wants to stop twice to take breaks.

Sponsored