If You love AI, Protect Human Written Texts at any Cost.

[memehunter]

This should be the final nail in the coffin of people arguing (through repetetive crap/ strawmanning etc) for allowing AI generated posts. Experts are claiming that 90% of the content available on Internet would be AI generated by 2026 (synthetic data). Now the problem is with LLM models is that if you train them on synthetic data (data produced by LLM itself) they start collapsing very badly. This could result in AI hallucinating and giving you totally wrong or fabricated responses. Researchers used the word 'poisoning' (rightly).
 



This gives us even more reasons to protect promote the human-written texts.

Source: Research Article published in Nature by Dr. Ilia Shumailov and the team of researchers from the University of Oxford.

[Upgrade00]

AI already gives wrong answers a lot of the time. I have stopped using it as a source of complex questions I have and really important ones, and when I use it I still read through multiple sources to form my own idea of the response rather than having AI do that for me.

On the 2026 prediction, to be fair, lots of the content on social media have been dominated by bots for several years now. Major difference I see now is that video content now has to be scrutinized to ensure it's produced by a human.

As a side note, I don't think this is a meta topic. It would have been if it were discussing AI n the forum, but it's on AI in general.

[memehunter]

AI already gives wrong answers a lot of the time. I have stopped using it as a source of complex questions I have and really important ones, and when I use it I still read through multiple sources to form my own idea of the response rather than having AI do that for me.

Yes, true if you read the research article, in the initial stages (I believe we have already passed this) trails of minor details started to distort and we are indeed experiencing this especially on crucial matters where there is no uniform opinion among experts.

As a side note, I don't think this is a meta topic. It would have been if it were discussing AI n the forum, but it's on AI in general.

You don't have to write specifically that it is about the forum. This is about discussing AI policy in this forum, and the discussion cannot be fruitful in a particular localized context. You have to take a wider context into consideration.

[NotATether]

Why don't the LLM companies simply detect and exclude the AI generated text from training? It would largely curtail the poisoning problem, at least from the second recursion onwards.

I think 90% of the content being AI generated only happens if a chatbot largely supplants Google Search (or if Google does this themselves, with Gemini returning most of the search results).

[memehunter]

Why don't the LLM companies simply detect and exclude the AI generated text from training?

First, it is not so easy to differentiate at scale between AI and human (thanks to our prior negligence). Second, it is inevitable as AI will have access to internet which will be full of crap (if not already). We can't afford that negligence now for sure.

[LoyceV]

How about I hate AI, and want to protect human writing anyway?

I can only encourage chatbots to destroy themselves, so I don't have to read their verbal diarrhea. I can't wait for this hype to disappear again: even every phone and browser is now forcing "AI" because for some reason many people think they need that.

Back to the forum: "AI" is mostly used by people who are too lazy (or dumb) to write something on their own, and fake images. I wouldn't miss that if it's gone.

[Free Market Capitalist]

We continue inside the cave. Meanwhile, outside the cave, Google already gives you as a first result what its AI generates, AI is implemented in certain curricula in universities, just as it permeates everything, new cell phones, you can use it to learn languages, to improve your knowledge on a particular subject, and so I could go on for a long time.

To protect what? If you are a scribe you'd better buy a printing press and learn how to use it.

[hugeblack]

This can be avoided by telling the AI ​​application where to search or adding a source. these applications are useful for summarizing an article or searching for specific information.

As for forums, they are banned because they are more likely to be plagiarism and lead to low-quality posts, especially if they are paid posts.

[ABCbits]

We could discuss about AI all the day, but i doubt admin will ever add rule that forbid AI usage to generate text. It's simply need too much effort to enforce such rule (even with detailed report from other member).

Why don't the LLM companies simply detect and exclude the AI generated text from training?

First, it is not so easy to differentiate at scale between AI and human (thanks to our prior negligence).

What exactly do you mean by "our prior negligence"?

Second, it is inevitable as AI will have access to internet which will be full of crap (if not already). We can't afford that negligence now for sure.

AI Big company behind AI/chatbot already do that, with their massive scraper. It got so bad that CloudFlare release multiple feature related with it,

https://blog.cloudflare.com/ai-labyrinth/
https://blog.cloudflare.com/declaring-your-aindependence-block-ai-bots-scrapers-and-crawlers-with-a-single-click/

Now i wonder whether @theymos use any of those feature on this forum.

[Lucius]

We could discuss about AI all the day, but i doubt admin will ever add rule that forbid AI usage to generate text. It's simply need too much effort to enforce such rule (even with detailed report from other member).
~snip~

My opinion is that some kind of rule is better than none, because when there is no rule, people behave exactly like we see - every day you can easily find at least a dozen profiles that use AI to generate their posts - and that's probably just the tip of the iceberg.

It would be easiest to say that the use of AI in an inappropriate way is subject to a permanent ban as well as plagiarism, but the admin will have a hard time deciding on such a step because it is too much work for the mods - although the forum has enough resources to appoint a few more moderators and start a reckoning with the AI ​​mafia.

It's been almost 5 years since @fillippone opened topic AI writing messages on Bitcointalk.org and I think it's simply too much time for nothing to be done officially - and if we let things take their course, in the next 5 years at least 50% of the content on the forum will be generated with AI and that will turn people away from the forum even more.

[mindrust]

AI is just a tool which is as smart as its user.

A pencil is not just a pencil. It is deadly in the hands of John Wick but some Leonardo might create amazing drawings with it. In the hands of a moron, it is a shitty tool.

If you are afraid of AI getting poisoned by its own texts, well, AI can be trained to ignore other AI written data. It is as easy as that.

*with a fookin pencil m8

[Ziskinberg]

Using AI to generate posts completely is wrong. But if it’s our own idea and we just need AI’s help to polish the language, I see no issue with that. After all, AI exists to help us communicate clearly, so we can’t ignore its role in making ideas more accessible.

Here’s a better question..

Would you prefer reading posts with poor grammar/spelling that are hard to understand, or would you rather people use AI tools to express their thoughts effectively?

[HeRetiK]

Why don't the LLM companies simply detect and exclude the AI generated text from training? It would largely curtail the poisoning problem, at least from the second recursion onwards.

This has been the core premise of generative adversarial networks; though while they've been important for early generative AI I'm not sure how much of a role they still plays these days. Also IIRC this approach has been mostly applied to image generation.

Regardless of that the challenge is that current AI is notoriously bad at detecting AI generated text, leading to both false positives and false negatives.

Would you prefer reading posts with poor grammar/spelling that are hard to understand, or would you rather people use AI tools to express their thoughts effectively?

I would rather have people learn to express their thoughts effectively. What's the use of a forum populated by bots.

[Perfectbaby]

I can't see myself using AI to post because what I think is that those who rely on AI are those who aren't fluent with their writing or that are too weak to write for themselves. As a learned individual I didn't see any reason to keep relying on AI to post or search information, it would be better i make my research and compares different articles to get what I want than solely rely on AI to give me the information I want to have. As a matter of fact, AI is highly prohibited here on the forum and anyone caught using AI to post such post could be penalized either by banning or tags.

[Frankolala]

This is one thing that I have never tried and wouldn't think of using AI for anything. I don't like the concept of using AI, because it doesn't help you to grow in knowledge and writing. It will make you lazy and unable to use your reasoning faculty to think outside the box. However, the world is not seeing it this way because they feel AI is making things more easier for them.

[uchegod-21]

This is one thing that I have never tried and wouldn't think of using AI for anything. I don't like the concept of using AI, because it doesn't help you to grow in knowledge and writing. It will make you lazy and unable to use your reasoning faculty to think outside the box. However, the world is not seeing it this way because they feel AI is making things more easier for them.

Please change your idea of AI.
AI is not a bad tool, unless the users use it in a bad way. Don't let your hatred for AI set you back in this AI generation. AI makes your work easy, especially if you are smart to prompt it well. I will use and have been using AI as much as possible in my daily live. I don't use it here in the forum because it doesn't fit here. This is a discussion forum, so humans should be discussing and not AI.

[Alone055]

if we let things take their course, in the next 5 years at least 50% of the content on the forum will be generated with AI and that will turn people away from the forum even more.

That's precisely the reason why I suggested there should be a rule against AI-generated posts because things will keep getting worse over time. Some people argue that users who are caught generating their posts using AI models are neutrally tagged, but we should ask ourselves, is that enough? It is not, because what's the point of giving someone a neutral tag or even multiple tags if they are still allowed to keep doing that? A tag wouldn't stop that from happening, and in the future, as @Lucius said, we will see more AI-generated posts than human-written posts, with all of them having a few neutral tags on their profiles, serving no purpose other than a mere warning that they are making AI-generated posts.

I respect everyone finding such users and making reports in the AI Spam Report Reference Thread, and those who are tagging those accounts mentioned there. I do that myself whenever I find someone making AI-generated posts. We report the posts as well, and they usually get removed too, but the users are barely banned. However, the point is, after all that hard work, we barely see the users facing any consequences for their actions.

So, I think, if we don't want to see the forum turn into a place full of AI-generated content, we need a rule that would get users banned, first temporrily, and then permanent for repeated offenses. Also, some will say it's not easy to always identify AI-generated content, but, whenever it's done, as we can see in the thread I mentioned above, at least then some action should be taken.

[Obim34]

This is one thing that I have never tried and wouldn't think of using AI for anything. I don't like the concept of using AI, because it doesn't help you to grow in knowledge and writing. It will make you lazy and unable to use your reasoning faculty to think outside the box. However, the world is not seeing it this way because they feel AI is making things more easier for them.

If you try to understand the post from OP side of prejudice, you will have to continue hating on AI.

What is the difference between use of AI and Search engines? The only notice is that AI goes deeper into details unlike normal engine search, similarly both tools are comprehensively used to find data or information that the human brain can't tell, if AI can be infiltrated, so can the search engines be poisoned in the future.

[Btcdeybodi]

"AI" is mostly used by people who are too lazy (or dumb) to write something on their own,

You can add that people who uses AI to write their posts for them are unintelligent too because anyone who is intelligent wouldn't ask any help from an AI to write their post for them. How many minutes does it even take to write a comprehensive post, and technically, if someone is too lazy to write a post by themselves that means they will literally be lazy at doing everything because they will always ask for help and assistance from others and indirectly, they are also trying to say that those things or people they depend on, are way smarter than them.

[Ambatman]

I don't really have issue with AI perse just people that are over reliant on them
And take their response as their own.

I biasly believe that over reliance has killed creativity
Checking certain blogs you would realised that it seem like they were all cut from same material.

I believe AI generated post should be stated by their poster forum wise.
Like a form of reference.

There was a time around last month that we were having discussion where personal POV was needed
And someone posted an answer from gpt and stated specifically I quote "in my opinion "
Nobody was happy after his action was found out
It felt like we that were giving our opinion were stupid.