AI generated text is forbidden with the exception of automated translation

Stayjuice

admin what about people who use reword on their keyboard for anti linguistics and ti aid their grammar? That’s ai but it just improves your grammar basically. We will get banned for that?

de0u

Stayjuice I think if the ideas come out of your head based on facts then it's OK if a program helps you fix spelling and grammar.

A big problem with LLM output is that LLMs routinely make things up, and nobody here has time to correct plausible hallucinations made by a linear algebra box (xkcd 1838). Eventually the quality of AI systems may improve to the point where this isn't an issue, at which point the moderators might well adjust the policy.

Stayjuice

de0u
Yeah. I mainly use it to improve my typing and for anti linguistics. The other so I use is for research but I never post said ai I use it for research and spend time fact checking because I know how inaccurate and pants it can be with facts since it just basically vacuumed all the data. I was watching a video and llama they basically pirated a ton of ebooks and fed it to the llm and used user data to teach it which sucks tbh

Bimmy

Thank you very much to admins/mods that keep a watchful eye.

I just had an experience, where an odd thread was about to catch my attention. For all interested, some details about specific case of today:

Collection of indicators why I found it suspicious:

It was from a user, which joined just few hours ago,
contained random quote blocks but without cited origin,
contained an URL to a non-familiar subdomain without a description where that might link to -- which I definitely don't click without confidence in its harmlessness
while on the same hand, the very meaning and intention was unclear to me.

I suspected, but wasn't entirely sure, so I began to invest some time to formulate a response to get clarification. I don't want to come across as hostile to a potential new user that might not speak English fluently and maybe just posted an unfortunate, tool-based translation.

I am nonetheless happy, that that thread is now gone. Apparently, you identified it already as against the rules, while I was still typing. That saves me and other users any further time to deal with the matter. So I "rerouted" my writing effort to post this thanks.

matchboxbananasynergy

Bimmy I know what thread you're referring to, and yes, it did seem to be some LLM thing, which is why it was removed. It wasn't clear what it was an answer to. It was quoting stuff that wasn't ever said.

userofgos

matchboxbananasynergy ah, I was wondering where that thread went. I was sceptical of it as well, and am thankful for the quick moderation. There's a lot of new users joining lately that are just starting to use GrapheneOS and are easy targets for these threads/comments. Greatly appreciated!

DeletedUser393

Kira902 neither do I. GOS is amazing all around

Ejd186

Excellent stance you've taken. Clearly you've taken the time and effort required to put a lot of thought into this post. Great job!
/s

admin

Ejd186 What's the issue you have with this thread?

Ejd186

admin none at all. I was just making a stupid joke trying to sound like an llm.

jh72de

DeletedUser227 There's no actual data to support this claim.

DeletedUser495

jh72de the data is all around you if you look well.

DavidXanatos

Is it permissible to put once own posting into ChatGPT with the prompt "Please Improve and fix spelling" or would that be a rule violation?
My spellign is lets say very crative and I am nto a native english speaker so my englisch may sound very german at times LOL.

de0u

DavidXanatos Microsoft Word, Google Doc, LibreOffice Writer, etc., all have spell checkers. Another option might be writing in German and using a translator app. An ongoing problem with general-purpose LLMs is that, unless one is extremely careful about checking the output, plausible-seeming junk will creep in. In May, Anthropic's lawyers were in hot water because one of them asked Claude (Anthropic's LLM!) to properly format a legal citation to a real scientific paper, and instead it provided a citation for a paper that never existed (source).

If one has the knowledge and time and patience to carefully check every word that comes out of an LLM, they can be useful, but at present the key thing to know is that they are fundamentally unreliable.

matchboxbananasynergy

It's a gray area which I would say is probably okay, as long as ChatGPT doesn't take any liberties and starts adding to what you're saying.

The primary reason for the rule is that we don't want to deal with an avalanche of incorrect claims made by an LLM. A lot of people have a habit of asking AI questions and presenting their incorrect answers as facts that we have to debunk, gets tiresome real quick.

wizoatk

de0u using a translator app.

DeepL is arguably one of the more prominent examples of a special purpose language translation LLM. The free verion(s) are reasonable for internet posts.

https://www.deepl.com/en/translator

GoS_anonuser1

de0u Slightly off-topic, but I wish everybody would understand this. I've been trying to hammer this to the heads of my friends for years now, but they genuinely believe that LLMs "think", are capable of "reasoning" and that they are "less biased" than humans. Good grief. That's the power of full on marketing right there.

As for the AI generated text ban here at GOS forum, I think it's brilliant.

Panda-na

wizoatk

I would also recommend Kagi Translator. It is based on AI (like DeepL), but I have had better results with Kagi. It also has more languages, also Sci-Fi languages (for example Klingon or Dothraki).

Xtreix

GoS_anonuser1 Exactly, two relevant readings on this subject.

wizoatk

Another relatively recent academic paper from a different angle:

Beware of Botshit: How to Manage the Epistemic Risks of Generative Chatbots
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4678265

Abstract
Advances in large language model (LLM) technology enable chatbots to generate and analyze content for our work. Generative chatbots do this work by ‘predicting’ responses rather than ‘knowing’ the meaning of their responses. This means chatbots can produce coherent sounding but inaccurate or fabricated content, referred to as ‘hallucinations’. When humans use this untruthful content for tasks, it becomes what we call ‘botshit’.
[...]

« Previous Page