EarthCache tasks solved by ChatGPT

[+SpeedCore]

I keep receiving EarthCache answers that are generated by ChatGPT. The finders simply copy all the tasks into the prompt and forward the output through the message center. They didn't even bother to delete the part where the AI explains that it can't take a picture of itself at the location because it doesn't have a physical body. How should we deal with such log permission requests lacking human brain activity? Are they allowed to log the Earthcache as found since the answers are correct or aren't they allowd to log their find since it says "descibe in your own words"? What do you think and is there an official guideline for that?

 

Quote

I am a large language model and therefore cannot perform real-world actions such as taking photos or collecting rocks.
* My answers are based on my current level of knowledge and therefore may be incomplete or incorrect.
* I am still developing and constantly learning.
Thank you for your understanding!
Please let me know if you have any further questions.

 

[+lee737]

Easy. Delete the log.

[+niraD]

A seventh grade English teacher recently read me the first paragraph of an essay that one of her students recently submitted. We found it amusing that anyone would think that a middle school student could have written it.

 

But it turns out that the EdTech (education technology) systems have built-in AI that flags student submissions that look like they were AI generated.

[+arisoft]

10 hours ago, SpeedCore said:

What do you think and is there an official guideline for that?

 

If the answer was correct, you should accept it, because the player has the correct answer. You can not undo the result.

Compare this to a situation where a player asks someone, who knows the subject better, to answer questions.

 

[+lee737]

10 hours ago, SpeedCore said:

Are they allowed to log the Earthcache as found since the answers are correct or aren't they allowd to log their find since it says "descibe in your own words"? What do you think and is there an official guideline for that?

Do you think they were at GZ? ie - photo proof etc.

Without that, I'd delete that log, based on the admission they weren't there.... "can't take a picture of itself at the location because it doesn't have a physical body."

[+barefootjeff]

1 hour ago, lee737 said:

Do you think they were at GZ? ie - photo proof etc.

Without that, I'd delete that log, based on the admission they weren't there.... "can't take a picture of itself at the location because it doesn't have a physical body."

 

Except if they're going to the trouble of using AI to answer the questions, they're just as likely to Photoshop themselves into a picture. This isn't much help for existing ECs, but going forward, it's probably a good idea to have at least one question that requires observing something at GZ so it can't be answered by AI bots.

[+hzoi]

On 6/16/2024 at 4:18 AM, SpeedCore said:

How should we deal with such log permission requests lacking human brain activity? Are they allowed to log the Earthcache as found

 

You're kidding, right?

 

Challenge the answer. Tell them to prove they visited by producing photos and logging answers using their own brains. If they cannot, delete.

[+thebruce0]

Agreed, the fact the answer is right means it couldn't be deleted solely because an AI generated it. If someone researched the answer on the 'net and sent it, same reasoning. You can't force the method of how someone attains an answer.  Once they have the correct answer, you can't require them to somehow "write it themselves" (it could have been AI, or their teacher, or their grandmother).  They 'learned' the answer (whether correct or not quite correct) somehow, attained it and provided it.

 

But, if they weren't at the location, then you can snag them on that. Fake photo, no evidence of being there, etc.

[+thekingofbattle]

This is actually super interesting...and I understand the reasoning behind those that would say "accept it." 

 

I think part of this will come down to ensuring the logging tasks are written in such a way to ensure the cacher actually visited the location. Tying that observation/analysis at GZ to the lesson...so that going to GZ actually occurs. 

 

A few weeks back, I wanted to see if ChatGPT could write an EC...oh, it generated something...but it wasnt really useful. Later I went back to see what it would come up with for a cache I'm writing and I asked it a very specific question. What I got back was completely wrong. So, it definitely has some limitations.  

[+hzoi]

4 hours ago, thekingofbattle said:

A few weeks back, I wanted to see if ChatGPT could write an EC...oh, it generated something...but it wasnt really useful. Later I went back to see what it would come up with for a cache I'm writing and I asked it a very specific question. What I got back was completely wrong. So, it definitely has some limitations.

 

Yeah, on the reviewing side, I've been getting some of these from other cachers. So far, they do not meet the guidelines.

 

It's all well and good if folks are using AI to polish up a listing they research themselves, but outsourcing the research work to ChatGPT is not the preferred technique.

[+Awesnap]

So, in lieu of these AI answers... should I be making photos a find requirement on my cache pages? As of now, I only ask is that the earthcache logging answers are sent on a majority of my EC's. It just seems exhausting to have to enforce a photo requirement in everyone's logs too.

[Neos2]

16 hours ago, Awesnap said:

So, in lieu of these AI answers... should I be making photos a find requirement on my cache pages? As of now, I only ask is that the earthcache logging answers are sent on a majority of my EC's. It just seems exhausting to have to enforce a photo requirement in everyone's logs too.

My personal opinion as a cache owner here.

I haven't received any AI logs yet, but it sounds like they are easy enough to detect. I know I did a search engine query the other day about a play I had seen, and the AI tried to write a synopsis of possible plays scripts it created. They were nonsensical. It was easy to see that no human had written those entries. 

I suggest people add photos on my EarthCaches, not because I want to police the photos so much as I want to be able to give people a break if their answers are off quite a bit. If they do include some photo from the area, it assures me they were there and I can send them back the answer I was looking for, so they can learn something--and I can allow the log to stand in good conscience. 

I usually have one or two questions and say that photos are welcome. I've seen COs flat out suggest adding a photo in case your answers are wrong.