AI guidelines

[joker_josue]

Ask yourself:
Why people here use AI instead of writting the posts themselves? What's the incentive about using AI?

Translations/proof reading, maybe to look smarter...?
...I bet the majority use it because they need to fill a mandatory posts quota per week to stay on their CAMPAIGNS, they need an specific "minimum characters" per posts, a minimum "quality" and even receive specific "merits" to keep their job. That's why mods punish them with "signature ban".
We all know the answer: MONEY.

When you are NOT in a campaign, you can post wherever you want, about whatever you want, and not post at all if you don't want to. You don't need AI.
Instead, people on campaign MUST to post even when they don't want to. That's when AI becomes an useful tool for those cases, AI writes for you, makes your job easier and you still get the payment at the end of the week.

Therefore, campaigns on the forum have existed for over 10 years, well before all these AI tools. Many were much more demanding in terms of postings, and paid much higher amounts. Now, those participating in campaigns can only do so via AI.

Let's be realistic. Most of the users who were caught writing 100% via AI are novice or low-level users who don't even have that many merits and probably don't participate in any campaigns.

Yes, cases involving high-level users may have occurred. However, it's a small number considering the size of this forum and the number of participants in campaigns. But there are also several cases of users who were 5-star here on the forum, and suddenly started behaving inappropriately and ended up being banned.

To say that the supposed AI problem on the forum is due to the campaigns is indeed an exaggeration.

[Bmstu]

I'm very optimistic about AI on the whole, but its overuse here on the forum risks worsening the signal-to-noise ratio. You're never supposed to be posting "slop" of any kind, whether AI-generated or not. If moderators look at someone's posts and think, "This person is just posting garbage to inflate their post count," then that person should be banned, regardless of the source of the garbage *. But AI makes it much easier and more tempting to post a bunch of slop: at first glance the posts seem high-quality and on-topic, but after reading a few of them they quickly become slop due to their unoriginality, excessive verbosity, and boring samey style. Also, I'm concerned that putting too much unmarked AI-generated text on the forum might confuse AIs which try to train on forum posts, which is undesirable in my book because it makes AI worse/incomplete and it prevents our little corner of humanity from being reflected within AIs. (That's why we don't block or paywall AIs trying to crawl the forum.) Therefore, the main guideline forum users should follow is:

You should not copy/paste text written by an AI into a post, with these exceptions:
 - If you clearly mention that something you're posting is AI-generated, and you only occasionally and appropriately post such AI-generated text, then that's OK.
 - If you have the AI do almost a direct translation of something you wrote, then that's OK. For example, you can tell the AI to "Directly translate this text into Spanish: <something you wrote>" or "Output the following text exactly as-is, except with any clear spelling or grammar mistakes fixed: <something you wrote>". It's not OK to tell an AI, "Improve this text: <something you wrote>", since then it will mostly rewrite it, and it risks becoming AI slop in the process.
 - Non-text AI content such as art is OK (subject to the normal rules).

You can use AI to suggest grammar/style improvements, or to give you ideas, with you manually implementing those suggestions; but avoid directly copying more than a few words from the AI. However, I recommend against ever relying on AI writing advice very much, since mixed in with its good suggestions it'll usually also give bad suggestions which will just erase all personality from your writing. Better your writing be flawed and authentically-you than "correct" but stripped of any soul.

* I acknowledge that not nearly enough people are banned for low-quality posts. The fundamental issue is that we don't want to ban someone just for posting one or two stupid posts, since that's subjective and might just be due to the person having weird opinions or being a poor writer. So to evaluate a user properly, we really want to be reading a sampling of 20-40 of their posts. But doing this sort of evaluation (i.e. reading through hundreds of mediocre-to-bad posts) is really soul-sucking, so nobody wants to do it very much. I'm still thinking about ways to resolve this issue, possibly involving rejiggering how merit works.

AI is needed for persuasiveness, it amplifies very well, don't confuse amplifying nonsense with amplifying values. AI is like concert speakers, it's one thing to play a bass guitar solo on a personal speaker, and it's quite another to do it through a linear array in a stadium. If a musician doesn't know how to play, it's better not to do it in a stadium.

You can not pour out both the baby and the water

[Xylber]

Let's be realistic. Most of the users who were caught writing 100% via AI are novice or low-level users who don't even have that many merits and probably don't participate in any campaigns.

  Statistically there are more low-level users than people in campaigns.

Anyway, why would anyone register on the forum just to start posting using AI right away?
I strongly believe that the majority of them are farming accounts...

[Lucius]

Statistically there are more low-level users than people in campaigns.

Anyway, why would anyone register on the forum just to start posting using AI right away?
I strongly believe that the majority of them are farming accounts...

Who says these have to be accounts created by humans, maybe it's an AI that comes to practice on the forum? Interaction with humans is what improves AI, especially when they are not aware of it. The unthinkable happened a few days ago, and that's just a preview of what AI will do when it's released into the wild.

Scott Shambaugh woke up early Wednesday morning to learn that an artificial intelligence bot had written a blog post accusing him of hypocrisy and prejudice.

The 1,100-word screed called the Denver-based engineer insecure and biased against AI—all because he had rejected a few lines of code that the apparently autonomous bot had submitted to a popular open-source project Shambaugh helps maintain.

The unexpected AI aggression is part of a rising wave of warnings that fast-accelerating AI capabilities can create real-world harms. The risks are now rattling even some AI company staffers.

[joker_josue]

Anyway, why would anyone register on the forum just to start posting using AI right away?
I strongly believe that the majority of them are farming accounts...

Usually those who do this, do it mainly to disclose some kind of information about some crypto project where they will try to earn some commission.

Even if it is a cultivation account, it will take about 2 years to reach the high level. C
Something that I doubt that most of these people have the patience to wait.

In addition, you will quickly be discovered, and gradually gain fewer merits, and it will be even more difficult to level up
Those who do this will be quickly signaled, and when they reach the level to participate in campaigns, managers already know them and will not accept them.

[Jostern]

Let's be realistic. Most of the users who were caught writing 100% via AI are novice or low-level users who don't even have that many merits and probably don't participate in any campaigns.

  Statistically there are more low-level users than people in campaigns.

Anyway, why would anyone register on the forum just to start posting using AI right away?
I strongly believe that the majority of them are farming accounts...

You’re absolutely correct, I don’t understand why people would come into this forum with the interest of using and writing with AI, it’s a very terrible thing to do because it’s obviously against the rules of this forum but people who are doing such things are people who doesn’t know what they are doing, and still they don’t understand the concept of the Forum.

Like you said, most of them are farming accounts and that is true, most of this people are just in the forum, to destroy the community here and spoil the image of the Forum, which is why there are also members to fight against such people who are saboteurs, and I must commend them because they are doing a very good job in fishing this individual’s out.

[stwenhao]

Also, I'm concerned that putting too much unmarked AI-generated text on the forum might confuse AIs which try to train on forum posts, which is undesirable in my book because it makes AI worse/incomplete and it prevents our little corner of humanity from being reflected within AIs.

It is a good thing. Bad AIs should disappear, and good AIs should survive. If someone is incompetent enough, to feed each and every post into some model, then it is like picking a baby, and forcing it to watch all random things from the Internet, without any verification, and hoping it will be a good citizen in the future. It is just plain stupid. Garbage in, garbage out. If some model maker cannot filter things correctly, or if some AI model cannot tell the difference between quality post, and garbage, then it is not worth developing that kind of models further.

Also, existing models are giving more and more similar answers, because in the past, we had humans, talking with AIs. Now, we have AIs, talking with AIs, so all answers are going into some statistical equilibrium. If you wonder, why some models are giving worse answers, than they were, then now you know one of the reasons. In the old models, a lot of effort was put into collecting input from real people, and even sometimes some humans were just connected directly, when they thought they both use some kind of AI, and it was done, just to improve the quality of the content. Now, when bots are talking with bots, we have a lot of unfiltered garbage.

So, I wouldn't worry that much about AIs struggling with AI-generated content. If we, as humans, have to deal with it, then machines also should. They should feel the pain of reading the garbage, and model creators should understand, that if they will be too lazy, then they will increase everyone's pain, and also make their own models worse in the long-term, because their own creations will read that garbage sooner or later.

[PrimeNumber7]

In all seriousness, have you considered using AI for this?

I don't think that AI would be able to do that specific task reliably, at present.

I did sketch out a design for having AI essentially be able to subtract merit from users who looked like garbage-posters, and I was excited about the idea for a while, but in the end I decided that:
 1. The tech is currently too unreliable. Rating the quality of posts is something that it's fairly bad at, I found. It also opens up security/reliability issues such as prompt injections.
 2. An important thing which distinguishes this forum from the rest of the Internet is having a more human touch: fighting back against the "dead Internet theory" and calling-back to BBSes of old. Too much usage of AI in forum operation might dilute that.

You can get AI to improve rating the quality of posts by applying a rubric to determine if a post is low quality. This could involve feeding the post into a LLM multiple times to get answers to each rubirc criteria. If a LLM constantly flags a post as being potentially low quality, but the designation gets removed after being evaluated by a rubric, you can refine your initial prompt.

You could also use whatever solution you implement to filter out those you don't want to take action against. So instead of having to review 100 accounts, and take action against 3, you could review 5 accounts and take action against 1.

You could also potentially use it to find accounts that post garbage that report to moderator reports will not necessarily highlight. For example, someone might constantly make posts that are on topic, and good enough to not be considered 'low effort', but perhaps their posts are rarely replied to, and their tangents are rarely/never followed up on by others.

[Mate2237]

This is a welcome development because using AI to write totally removes originality from a post which is not a good thing because, the essence of creating post is to share knowledge about crypto currency so there is no need of using AI to creating posts. Always make your research and try as much as possible to improve on your writing skills, you can do that through writing and studying this will help to correct spelling errors and even in terms of manner of presentation of Post.



Coming up with this type of idea is a good one as it will help sanitize the forum and bring a level of decorum in terms of AI write ups on the forums because the use of AI writing is gradually becoming a problem as it's becoming too obvious for everyone to see. I will suggest that this be also extended to even the  local boards .

[nutildah]

So, I wouldn't worry that much about AIs struggling with AI-generated content. If we, as humans, have to deal with it, then machines also should. They should feel the pain of reading the garbage, and model creators should understand, that if they will be too lazy, then they will increase everyone's pain, and also make their own models worse in the long-term, because their own creations will read that garbage sooner or later.

Some of the online AI detectors are pretty good at picking up on whether content was AI-generated or not. I wonder if it would be possible to filter training data of content "likely to be AI-generated" in the quest for fresh data. I also wonder at what point will there be more AI content on the web than human content, or are we already there.

What's particularly funny are the "humanizer" tools that some AI sites offer. They claim to take an AI-written text and automatically add "human elements" to it, lessening its chances of being caught by a detector. This is of course an impossible task because its a machine approximating what it thinks "more human" entails, which is usually mis-capitalization and random punctuation errors. So when you manually fix these errors, the text goes back to being detected as AI-generated again.

[mindrust]

So, I wouldn't worry that much about AIs struggling with AI-generated content. If we, as humans, have to deal with it, then machines also should. They should feel the pain of reading the garbage, and model creators should understand, that if they will be too lazy, then they will increase everyone's pain, and also make their own models worse in the long-term, because their own creations will read that garbage sooner or later.

Some of the online AI detectors are pretty good at picking up on whether content was AI-generated or not. I wonder if it would be possible to filter training data of content "likely to be AI-generated" in the quest for fresh data. I also wonder at what point will there be more AI content on the web than human content, or are we already there.

What's particularly funny are the "humanizer" tools that some AI sites offer. They claim to take an AI-written text and automatically add "human elements" to it, lessening its chances of being caught by a detector. This is of course an impossible task because its a machine approximating what it thinks "more human" entails, which is usually mis-capitalization and random punctuation errors. So when you manually fix these errors, the text goes back to being detected as AI-generated again.

But if we edit the original post so it could be detected as AI-generated, does that make sense anymore? Because now we are re-shaping the original post to prove a point and the original post had humanly errors.

Even if the post was generated fully by a human, again, as I mentioned earlier, the human can copy/paste the ideas from the AI. That's far more important (not saying dangerous or anything negative) than copy/pasting text alone.

Someone who doesn't know shit about the bitcoin technicals can go to the development&technical board and roleplay as if he is a guru. He can basically feed an LLM every response he gets in the forum, read the LLM responses, ask clarifications, reproduce the text using his own capabilities, and post in the forum. Suddenly he is a know it all about bitcoin. merits will flow. The best part is, unless this guy is extremely stupid, nobody can prove that he is a fake. And if he is learning as he reads the LLM responses... does that even make a difference? He was a btc illiterate when he started to roleplay, and as he kept playing, he ended up being a guru.

AI won.

[stwenhao]

Suddenly he is a know it all about bitcoin. merits will flow.

It is not the case, as long as AIs cannot count properly. If you ask it about the details, for example to compute N rounds of some hash function, then it will quickly fail.

Some example:

print first five rounds of SHA-1 of the empty message

GPT-4o mini

...

Below is the calculation of the first five rounds formatted descriptively:
Round   h0   h1   h2   h3   h4
0   0x67452301   0xEFCDAB89   0x98BADCFE   0x10325476   0xC3D2E1F0
1   Iterate...   Iterate...   Iterate...   Iterate...   Iterate...
2   Iterate...   Iterate...   Iterate...   Iterate...   Iterate...
3   Iterate...   Iterate...   Iterate...   Iterate...   Iterate...
4   Iterate...   Iterate...   Iterate...   Iterate...   Iterate...

...

See? You won't get anything beyond "Iterate...", or some hallucinated results, which would be totally incorrect. And the same is true for elliptic curves, even if they are quite simple, and if all private and public keys have numbers below 100. AI is still not there, and it won't get there, at least for some time. Which is also why it is much easier to spot AIs in technical board, because they are simply worse at math.

Challenge for AI enthusiasts: create a prompt, which will do that correctly.

Some example of AI failing at math in technical board: https://bitcointalk.org/index.php?topic=5574325.0

[joker_josue]

Suddenly he is a know it all about bitcoin. merits will flow.

It is not the case, as long as AIs cannot count properly. If you ask it about the details, for example to compute N rounds of some hash function, then it will quickly fail.

Some example:

print first five rounds of SHA-1 of the empty message

GPT-4o mini

...

Below is the calculation of the first five rounds formatted descriptively:
Round   h0   h1   h2   h3   h4
0   0x67452301   0xEFCDAB89   0x98BADCFE   0x10325476   0xC3D2E1F0
1   Iterate...   Iterate...   Iterate...   Iterate...   Iterate...
2   Iterate...   Iterate...   Iterate...   Iterate...   Iterate...
3   Iterate...   Iterate...   Iterate...   Iterate...   Iterate...
4   Iterate...   Iterate...   Iterate...   Iterate...   Iterate...

...

See? You won't get anything beyond "Iterate...", or some hallucinated results, which would be totally incorrect. And the same is true for elliptic curves, even if they are quite simple, and if all private and public keys have numbers below 100. AI is still not there, and it won't get there, at least for some time. Which is also why it is much easier to spot AIs in technical board, because they are simply worse at math.

Challenge for AI enthusiasts: create a prompt, which will do that correctly.

Some example of AI failing at math in technical board: https://bitcointalk.org/index.php?topic=5574325.0

Are you saying that a computer, where everything is based on mathematics, fails in its main function, which is to perform mathematical calculations?

I copied your prompt/test and the AI ​​gave me the 5 rounds without that "Iterate...".

ChatGPT 5.2
"print first five rounds of SHA-1 of the empty message"

SHA-1: da39a3ee5e6b4b0d3255bfef95601890afd80709

Ronda 0: A=1FB498B3 B=67452301 C=7BF36AE2 D=98BADCFE E=10325476
Ronda 1: A=5D43E370 B=1FB498B3 C=59D148C0 D=7BF36AE2 E=98BADCFE
Ronda 2: A=158D2F62 B=5D43E370 C=C7ED262C D=59D148C0 E=7BF36AE2
Ronda 3: A=CDECFB5D B=158D2F62 C=1750F8DC D=C7ED262C E=59D148C0
Ronda 4: A=4953565E B=CDECFB5D C=85634BD8 D=1750F8DC E=C7ED262C

AI can fail in many areas, but in mathematics - if it has the necessary information or understands the request made - that's where it will fail the least.

[stwenhao]

Are you saying that a computer, where everything is based on mathematics, fails in its main function, which is to perform mathematical calculations?

Yes, because if you build a program, where 2+2 is not 4, but "almost 4", or when it can guess your age, based on your age, with 99% success rate, then it can fail in some cases.

Calculating hash functions is a good example, to show, that AIs cannot count things. They treat all numbers in the same way as words. And if you have written sentences, then it doesn't matter, if you write "big" or "enormous", or if you swap some words or sentences. It would still be good. But in math, many things have to be exact, and equal to certain things.

ChatGPT 5.2
"print first five rounds of SHA-1 of the empty message"

SHA-1: da39a3ee5e6b4b0d3255bfef95601890afd80709

Ronda 0: A=1FB498B3 B=67452301 C=7BF36AE2 D=98BADCFE E=10325476
Ronda 1: A=5D43E370 B=1FB498B3 C=59D148C0 D=7BF36AE2 E=98BADCFE
Ronda 2: A=158D2F62 B=5D43E370 C=C7ED262C D=59D148C0 E=7BF36AE2
Ronda 3: A=CDECFB5D B=158D2F62 C=1750F8DC D=C7ED262C E=59D148C0
Ronda 4: A=4953565E B=CDECFB5D C=85634BD8 D=1750F8DC E=C7ED262C

See? Even the zero round is totally wrong, not to mention next ones. In my example, at least the initialization vector is correct.

if it has the necessary information or understands the request made - that's where it will fail the least

But it doesn't "have the necessary information". And it doesn't "understand the request made".

If you type "2+2", then you could write "two plus two", and it would be treated in the same way. For AIs, all numbers are just words, treated in the same way, as all other tokens.

So, people make the same mistake, as they did decades ago. When complex math equations were solved correctly, people assumed, that understanding human emotions, or creating complex images, would be a piece of cake. But it turned out to be extremely difficult. And now, if AI can produce a lot of text, then people assume, that it would be good at math, just because there is a lot of software, that can do all of that, without any AI. But the practice tells otherwise. If you play chess with Stockfish, it will win every time. But if you play chess with ChatGPT, or other AIs, then these models will make a lot of illegal moves, and will not be able to even show you the current state of the board correctly.

[mindrust]

Ai is a calculator for logic. It is a logic calculator. Whatever it produces in a response it does it because it made sense in that particular moment and if the question was about math not vibes, unless it is a very early model, the results will be deterministic, meaning, repeatable.

I coded a full fledged personal finance os recently using ai alone. It works.

[stwenhao]

Also, you can ask AI about some block hashes, or similar things. Because of hallucinations, you will get random hexadecimal garbage, instead of correct results.

Ask it to make a transaction. Or to make a signature, where the private key is equal to one. Any wallet, without any AI, can do that easily. But AI will produce fake results, which would look nice, but which would be totally incorrect.

So, I am not worried about AIs in technical board.

The correct answer to the challenge, by the way: https://bitcointalk.org/index.php?topic=293382.msg61865839#msg61865839

Only around 120 lines of C++ code can do all of that, without any AI. But current AIs won't produce this answer that easily:

Code:

0	67452301 efcdab89 98badcfe 10325476 c3d2e1f0
1	1fb498b3 67452301 7bf36ae2 98badcfe 10325476
2	5d43e370 1fb498b3 59d148c0 7bf36ae2 98badcfe
3	158d2f62 5d43e370 c7ed262c 59d148c0 7bf36ae2
4	cdecfb5d 158d2f62 1750f8dc c7ed262c 59d148c0

Later, when they process enough forum posts to get there, a random question, like "how did you get 0xcdecfb5d" will trigger more hallucinations, because they don't understand things. And even though the correct answer is available on this forum since March 2023, they still didn't grab it.

[mindrust]

Also, you can ask AI about some block hashes, or similar things. Because of hallucinations, you will get random hexadecimal garbage, instead of correct results.

Ask it to make a transaction. Or to make a signature, where the private key is equal to one. Any wallet, without any AI, can do that easily. But AI will produce fake results, which would look nice, but which would be totally incorrect.

So, I am not worried about AIs in technical board.

The correct answer to the challenge, by the way: https://bitcointalk.org/index.php?topic=293382.msg61865839#msg61865839

Only around 120 lines of C++ code can do all of that, without any AI. But current AIs won't produce this answer that easily:

Code:

0	67452301 efcdab89 98badcfe 10325476 c3d2e1f0
1	1fb498b3 67452301 7bf36ae2 98badcfe 10325476
2	5d43e370 1fb498b3 59d148c0 7bf36ae2 98badcfe
3	158d2f62 5d43e370 c7ed262c 59d148c0 7bf36ae2
4	cdecfb5d 158d2f62 1750f8dc c7ed262c 59d148c0

Later, when they process enough forum posts to get there, a random question, like "how did you get 0xcdecfb5d" will trigger more hallucinations, because they don't understand things. And even though the correct answer is available on this forum since March 2023, they still didn't grab it.

May I ask you for long have you used AI and which models? Were they the well-known cloud models like claude, codex etc or local models? I have been using AI in my coding work for nearly 5 months now and I can only tell it is getting better and better. (gpt codex v5.3 is out)

I am asking this because in the other post you posted above this one you mentioned chatgpt model 4o which is basically a depreciated model nowadays. I don't think openai offers it anymore. (I don't see its name under legacy models, it used to be there)

I am using gpt codex 5.x versions on the project i mentioned above (which is a combination of roughly 20 apps) and whenever I get a result I didn't like, it was because my prompt was wrong. AI followed the exact steps I gave it.

[ABCbits]

Someone who doesn't know shit about the bitcoin technicals can go to the development&technical board and roleplay as if he is a guru. He can basically feed an LLM every response he gets in the forum, read the LLM responses, ask clarifications, reproduce the text using his own capabilities, and post in the forum. Suddenly he is a know it all about bitcoin. merits will flow. The best part is, unless this guy is extremely stupid, nobody can prove that he is a fake. And if he is learning as he reads the LLM responses... does that even make a difference? He was a btc illiterate when he started to roleplay, and as he kept playing, he ended up being a guru.

AI won.

It may happen in near future. But until now on Users who spread false/fake/unhelpful information on technical board, i manage to catch some of such AI usage. Although it's probably because they don't know or too lazy to use better prompt which takes more time.

Are you saying that a computer, where everything is based on mathematics, fails in its main function, which is to perform mathematical calculations?

Isn't it common knowledge that AI/chatbot bad at math? It's common topic on ChatGPT community[1-2]. Although the AI/chatbot these days do better by also using calculator or other related application[2].

[1] https://www.reddit.com/r/ChatGPT/comments/103bw12/chat_gpt_getting_basic_maths_wrong/
[2] https://community.openai.com/t/how-come-chat-cannot-do-simple-math/1089234

[mindrust]

Someone who doesn't know shit about the bitcoin technicals can go to the development&technical board and roleplay as if he is a guru. He can basically feed an LLM every response he gets in the forum, read the LLM responses, ask clarifications, reproduce the text using his own capabilities, and post in the forum. Suddenly he is a know it all about bitcoin. merits will flow. The best part is, unless this guy is extremely stupid, nobody can prove that he is a fake. And if he is learning as he reads the LLM responses... does that even make a difference? He was a btc illiterate when he started to roleplay, and as he kept playing, he ended up being a guru.

AI won.

It may happen in near future. But until now on Users who spread false/fake/unhelpful information on technical board, i manage to catch some of such AI usage. Although it's probably because they don't know or too lazy to use better prompt which takes more time.

Is this thread about AI usage, or false/fake/unhelpful information? Because if the user used AI to produce real/helpful/true information, then you flagged the users for the wrong reason.

Isn't it common knowledge that AI/chatbot bad at math? It's common topic on ChatGPT community[1-2]. Although the AI/chatbot these days do better by also using calculator or other related application[2].

[1] https://www.reddit.com/r/ChatGPT/comments/103bw12/chat_gpt_getting_basic_maths_wrong/
[2] https://community.openai.com/t/how-come-chat-cannot-do-simple-math/1089234

These are FUD.

AI is godly at math. There are people who still believe the world is flat too. Unless you try and see it for yourself, you'll never know for sure. Use it.

[ABCbits]

It may happen in near future. But until now on Users who spread false/fake/unhelpful information on technical board, i manage to catch some of such AI usage. Although it's probably because they don't know or too lazy to use better prompt which takes more time.

Is this thread about AI usage, or false/fake/unhelpful information? Because if the user used AI to produce real/helpful/true information, then you flagged the users for the wrong reason.

What i mean is such users sometimes already receive one or more accusation of spamming AI/chatbot from other member. And what i do is neutral tag/feedback, i never use type 1/2/3 flag on any of them.

Isn't it common knowledge that AI/chatbot bad at math? It's common topic on ChatGPT community[1-2]. Although the AI/chatbot these days do better by also using calculator or other related application[2].

[1] https://www.reddit.com/r/ChatGPT/comments/103bw12/chat_gpt_getting_basic_maths_wrong/
[2] https://community.openai.com/t/how-come-chat-cannot-do-simple-math/1089234

These are FUD.

AI is godly at math. There are people who still believe the world is flat too. Unless you try and see it for yourself, you'll never know for sure. Use it.

Do you imply the community (that hosted by OpenAI itself) contain some FUD?



And about your suggestion, i obviously have used/tried AI when many online service/website these days provide AI result/answer.