New rule against AI use on ES

[neptronix]

I've put in a new rule against using AI to impersonate a human, because we are increasingly seeing this lately.
0% of the time are these bots providing anything of value to anyone. They're being used to gain a reputation on the site so they can spam us later.

These bots undermine the human to human communication that people expect on sites like ours, and we won't allow that to happen here.

What to do:
When you see these AI bots, please report their posts; our policy is now to remove this.
If you are copy pasting AI output, please note that it came from AI when posting.

 

[A-DamW]

If you are copy pasting AI output, please note that it came from AI when posting

Thanks for that clarification, was just going to ask about it.

 

[nicobie]

I'm curious on how to determine if it is AI? Is it something I'm capable of? Or is it another judgemental call?

 

[gromike]

On some forums I've been on, if the poster has perfect spelling they are suspect. Not of AI, but suspect nonetheless.
Though, people here spell pretty good.
From the AI post I've read, they can be engagingly chatty, ramble on, and then never really come to a point. Kind of like they are just regurgitating related words and ideas.

 

[neptronix]

I'm curious on how to determine if it is AI? Is it something I'm capable of? Or is it another judgemental call?

Judgement call but the patterns and language style are still fairly obvious.
If unsure, don't remove - but a pattern will emerge with that user.

Here's an example user that i nuked:
https://endless-sphere.com/sphere/members/treasure.100646/

 

[jetpackjbd]

Here’s another example from a small wiki I’m a part of, I’ve seen similar things on other forums but this is the most extreme I’ve experienced:


It’s usually just a bunch of nonsense not related to the forum at all from what I’ve seen; it’s very easy to spot.

 

[neptronix]

Oh man, that's another problem with wikis that are unattended and let everyone in the world edit them. Wikipedia has gradually became corrupt, therefore untrustable, due to this.

Independent websites that think the internet is still a high trust environment like it was in the 90's are today's dodo birds.

 

[Kottbullar]

Thank you for your message.
Your feed back is important to us.
This forum may be monitored for training and quality control purposes.

 

[nicobie]

Do we really need wiki?

 

[neptronix]

Do we really need wiki?

No, but a knowledgebase ( that's different ) is the most requested feature

 

[marty]

Thank you for your message.
Your feed back is important to us.
This forum may be monitored for training and quality control purposes.

So here we have a human pretending to be a robot? This AI stuff has just begun. AI will get much better. Soon humans will not be able to tell if they are communicating by text with a human or robot? Prediction for the future? Robots will learn how to build robots and there will be no need for humans any more. My plan is to have human fun now, because the end is near.

 

[neptronix]

If it comes down to it, we should turn this into a laser/plasma weapons forum; i think we have the collectively knowledge to mount a strong offense during the great ai cyborg uprising of 2029.

 

[jonescg]

My students always use AI to write their reports. And it's either wrong, or so obviously not written by *that* human, I call them out. No point AI checking as it's usually fairly obvious, but at the end of the day, I reserve the right to orally question their knowledge, and that normally sorts the wheat from the chaff.

 

[neptronix]

I mean... you have to, otherwise you are compromising the integrity and utility of your role as a teacher.
And the student is compromising their education and the doors that education can open.

People who use AI to cheat at school are in for it later.. if they have equivalent knowledge/capabilities as an AI and no more.. they will be the easiest to replace with AI in the near future. What they need to survive in a career in the future is to do a better job than a robot ( not hard, but takes actual effort ), because that same robot is coming for their job.

The best thing you can do for them is flunk their ass i f they don't stop leaning on AI.

 

[Papa]

The best thing you can do for them is flunk their ass....

No.... take away their electronics, and boot'em outside...

 

[nicobie]

I just ignore them. They will get what they deserve soon enough.

 

[offGridDownUnder]

They will get what they deserve soon enough.

Perhaps this is meant as a flippant comment in a social interaction, but otherwise it's not real.

Socializing children (done properly) prepares them for adult freedom in a manner that does not lead them to murder, steal, assault. In a manner that contributes to the general good, rather than as parasites that only take to feed their immediate whims.

Yes, you can point to failures - several of the Silicon Valley CEOs come to mind. The failures only emphasize the need for success.

Many people have the wrong idea about school - for example, the idea of creating compliant workers. For the sake of society, school prepares children to be citizens - and then they decide for themself, as citizens, what to do with society.

Those of us that live long enough to become old, will be dependent on the behavior of these children. It is wise to invest in their foundations.

 

[eee291]

 

[neptronix]

We have our first major LLM spam:
Avoid " F I R E" - A little knowledge can save your life

This post contains some incorrect information, some which is dangerously incorrect. This is unedited LLM output. The poster clearly didn't vet it for accuracy, or maybe they didn't understand the topic well enough to do that.

These kinds of posts are cancer. ES is at or near the top of google searches for small EVs. If those search results are polluted by good looking, but dangerously incorrect information, this has an outsized effect on misinforming the internet.. and that erodes this website's usefulness.

What should i do?
Don't post LLM output as if you wrote it in the first person. Always credit the LLM that wrote it.
Always vet it for accuracy, we don't need artificial stupidity on this website on top of natural stupidity

Members: please report this kind of post whenever you see it.
Moderators: move these to the dumpster and wag your finger at the offender

 

[E-HP]

It seems like if AI can’t figure out the answer, it just assembles stuff that uses some of the words in the question to make a nonsensical (to a knowledgable reader) “answer”. Like some people, it doesn’t know enough to say “I don’t know”, which is even more dangerous.

 

[neptronix]

It's a predictive token generator, that, if prompted incorrectly, will output nonsense most of the time. If prompted correctly, might still output nonsense.

It doesn't have a concept of what it doesn't know to begin with, output is based on statistical number crunching of what word comes after the next in a sentence it read before.

 

[E-HP]

They should call the stupid thing AS.

 

[offGridDownUnder]

If you are copy pasting AI output, please note that it came from AI when posting.

Bingo.
If an AI wrote it, I didn't, so this is proper.

 

[neptronix]

Yeah, humans come here for human interaction with each other.

There's a little wiggle room on this rule:
IF
- english is your second language and you need some help
- you're missing a few fingers or have some other physical disability that makes typing hard
- you want to speed boost your way through making helpful articles about stuff you already know
AND
- you vet, edit, and humanize the response so that it doesn't have unnatural language and represents what you want to say

I'm totally cool with you using a LLM here as long as you are honest about using it.

 

[floydr]

Sh*T some of us are old the last thing we need is a YAFA Yet Another Freaking Acronym. AI I understand the acronym LLM even with the most likely AI outputted explanation I still don't understand. "

Large language model​

A large language model is a type of machine learning model designed for natural language processing tasks such as language generation. LLMs are language models with many parameters, and are trained with self-supervised learning on a vast amount of text. The largest and most capable LLMs are generative pretrained transformers. Wikipedia"
So LLM is a long winded way to say AI?

later floyd