[021] ChatGPT Spam Filter [Deleted]

[021]

021 submitted a new resource:

[021] ChatGPT Spam Filter Bot - Spam filter bot using ChatGPT.

The ChatGPT Spam Filter Bot is a highly efficient solution for forum administrators who want to automatically filter spam messages using advanced AI technology. The bot is powered by ChatGPT's natural language processing capabilities, which enable it to accurately identify and filter a wide range of spam messages.

With its powerful spam-filtering capabilities, the ChatGPT Spam Filter Bot add-on is an essential tool for forum administrators who want to maintain a clean and welcoming...

Read more about this resource...

 

[duderuud]

Interesting to see how effective this will be.

 

[021]

Interesting to see how effective this will be.

Feels much more effective than existing XenForo spam filters. I also plan to add more filter context over time as needed, which should improve spam detection.

 

[Alpha1]

It would be useful to send the content to moderation queue and/or create a report so that the moderators can act upon it and spam clean / ban the user if needed.

Functionality to detect and report spam and other serious rule breaches (child porn, drug dealing, stalking, etc) in conversations would also be very useful. Maybe in a separate addon.

 

[Alpha1]

@021 I don't know if its feasible at this stage of chatGTP, but if its possible then the following would be an interesting use case:

1. Feed the community rules to chat GPT
2. Maybe also feed to ChatGTP: the warnings given. So that it can learn in what situations which warnings have been given.
3. Scan new posts for rule breaches and when detected, then optionally send to moderation queue and/or create report citing the rule breach. This would allow moderators to handle the report from there.

 

[021]

@021 I don't know if its feasible at this stage of chatGTP, but if its possible then the following would be an interesting use case:

1. Feed the community rules to chat GPT
2. Maybe also feed to ChatGTP: the warnings given. So that it can learn in what situations which warnings have been given.
3. Scan new posts for rule breaches and when detected, then optionally send to moderation queue and/or create report citing the rule breach. This would allow moderators to handle the report from there.

Easy. Prepare your money, hahah

 

[ProCom]

I've been NECK DEEP in ChatGPT having it help me write SQL queries and PHP pages that are otherwise WAY over my head. I'm SO glad / fascinated to come to XF and search for ChatGPT to find these addons you've been working on. BRILLIANT!!!!

Regarding this spam filter addon:

It would be useful to send the content to moderation queue and/or create a report so that the moderators can act upon it and spam clean / ban the user if needed.

Yup, I 100% agree with this. I don't want spammers getting feedback in real-time that their posts are being blocked. I want them to feel like they might get through, just to find that the posts are in moderation and then the accounts are spaminated!

Feed the community rules to chat GPT

Yup, this too! In the API videos I've watched, I believe it's possible to tell the ChatGPT API who it is, and give it pre-prompts, like: "You are a discussion forum moderator. Check the content below to see if it violates these rules, or otherwise looks like SPAM:

Rule 1:
Rule 2...
etc."

 

[021]

Yup, I 100% agree with this. I don't want spammers getting feedback in real-time that their posts are being blocked. I want them to feel like they might get through, just to find that the posts are in moderation and then the accounts are spaminated!

Ok, I will add an option to select an action when spam is detected)

 

[ProCom]

Ok, I will add an option to select an action when spam is detected)

BRILLIANT!

I'm watching this thread to see when it has that feature, and then I'll be first in line to purchase!

So, this (and your other ChatGPT addons) require the API key, which requires an OpenAI paid account, correct? Seems like their prices have dropped a LOT with the recent updates, and now using ChatGPT 3.5 Turbo.

Any idea how many tokens (on average) this addon requires every time it reviews a post for spam (assuming average 1st post length of about 1,000 characters)?

Oh... and I just realized: I didn't see anything about how many / which messages are scanned? Hopefully not EVERY message/post... even for very old/active/vetted members? Ideally it would scan the same # of messages in this XF setting, or have it's own "scan the first xy posts of each new member" setting.

 

[021]

So, this (and your other ChatGPT addons) require the API key, which requires an OpenAI paid account, correct? Seems like their prices have dropped a LOT with the recent updates, and now using ChatGPT 3.5 Turbo.

Yes.
Just a question to find out what I'm doing wrong: many clients ask me a question, the answers to which are clearly stated in the description, such as in this case. Why did you choose to ask this question?

Any idea how many tokens (on average) this addon requires every time it reviews a post for spam (assuming average 1st post length of about 1,000 characters)?

It all depends on the content of the posts. You can see how OpenAI calculates tokens here: https://platform.openai.com/tokenizer

Oh... and I just realized: I didn't see anything about how many / which messages are scanned? Hopefully not EVERY message/post... even for very old/active/vetted members? Ideally it would scan the same # of messages in this XF setting, or have it's own "scan the first xy posts of each new member" setting.

The add-on embeds the checker into the standard XenForo system in its entirety and will check messages like any other checker, like Akismet, for example.

 

[ProCom]

Just a question to find out what I'm doing wrong: many clients ask me a question, the answers to which are clearly stated in the description, such as in this case. Why did you choose to ask this question?

Glad you asked! Your text says "may charge"... but from what I understand, the API is only available for people that pay. So, I didn't know if you knew something I didn't know regarding their pricing structure, etc. Hopefully that makes sense?

I'd probably word it with something like: "This addon requires an OpenAI / ChatGPT API Key that is pay-per-use. You can learn more about the key and how to setup an account here <link>."

(I'd also add a direct link to your other addon required... just to make things clear and easy)

The add-on embeds the checker into the standard XenForo system in its entirety and will check messages like any other checker, like Akismet, for example.

For a site that gets thousands of posts per day, many of them long-and-wordy, this could get really expensive really quickly, no? In full honesty: I wouldn't use this addon unless I could control which messages it scans. No reason (for my forum) for it to scan every message from members that have been posting tens-of-thousands of times for 15 years.

 

[ProCom]

It all depends on the content of the posts. You can see how OpenAI calculates tokens here: https://platform.openai.com/tokenizer

This was helpful, thanks! (I didn't know about that tool)

For fun, I ran some basic numbers at the pricing here, which is currently: gpt-3.5-turbo $0.002 / 1K tokens

I took the average post length, average number of posts, and cost, and I'm at about $3/day during my busy-season. Not the end of the world, but I'm guessing it would be only pennies if I only scanned the first 10 or even 100 posts for each member.

 

[ProCom]

Question: I wonder if any of the other models would be adequate and less expensive for this task (spam-checking)?

OpenAI API

An API for accessing new AI models developed by OpenAI

platform.openai.com

(sorry if I'm now monopolizing this thread, but I've been DEEP immersed in ChatGPT for the last few months... and LOVE it, and watch hundreds of videos about it, the CEO, etc. etc. etc.)

 

[021]

Glad you asked! Your text says "may charge"... but from what I understand, the API is only available for people that pay. So, I didn't know if you knew something I didn't know regarding their pricing structure, etc. Hopefully that makes sense?

They provide free trial access, as far as I understand.

For a site that gets thousands of posts per day, many of them long-and-wordy, this could get really expensive really quickly, no?

Unfortunately, I cannot give you a definitive answer to your question. OpenAI's pricing policy is quite individual and in order to answer your question, you need to test the add-on in practice. I recommend that you wait for feedback from other buyers who have large forums where this add-on is installed.

No reason (for my forum) for it to scan every message from members that have been posting tens-of-thousands of times for 15 years.

As I mentioned earlier, this add-on is fully integrated into the XenForo system, which means that options such as "Maximum messages to check for spam" are applicable to this add-on

 

[rdn]

They provide free trial access, as far as I understand.

Correct.

 

[021]

Question: I wonder if any of the other models would be adequate and less expensive for this task (spam-checking)?

Unfortunately not, they are not smart enough to make the add-on work stable.

 

[021]

By the way, unfortunately, my PayPal account has been restricted from accepting payments... Therefore, I'm looking for a partner who can accept payments from the my store to PayPal and send me funds in cryptocurrency. Ready to consider any terms of cooperation.

 

[Alpha1]

Here are the ways OpenAI allows you not to go over your budget:

You can configure a usage hard limit in your billing settings, after which we’ll stop serving your requests. You may also configure a soft limit to receive an email alert once you pass a certain usage threshold. There may be a delay in enforcing the limit, and you are responsible for any overage incurred. We recommend checking your usage tracking dashboard regularly to monitor your spend.

When you sign up, you’ll be granted an initial spend limit, or quota, and we’ll increase that limit over time as you build a track record with your application. If you need more tokens, you can always request a quota increase.

This method could potentially be implemented into the addon settings:

You can limit costs by reducing prompt length or maximum response length, limiting usage of best_of/n, adding appropriate stop sequences, or using engines with lower per-token costs.

Pricing

Simple and flexible. Only pay for what you use.

openai.com

 

[021]

Ok, here is an example of how to calculate the approximate cost for your forum:
Let's say we check all users who have less than 50 posts. Let's take an average message length of 1800 characters or 390 tokens and let's assume that users post 200 such messages per day. So

• Total tokens per day = 390 tokens per post * 200 posts per day = 78,000 tokens per day
• The total number of tokens used over a 30-day period would be 78,000 tokens per day * 30 days = 2,340,000 tokens
• Based on the current pricing for the ChatGPT GPT-3.5 Turbo model of $0.002 per 1K tokens, the estimated cost for 2,340,000 tokens would be = (2,340,000 / 1000) * $0.002 = $4.68 for 30 days

Spoiler: Sample message for 389 tokens

ChatGPT[a] is an artificial intelligence chatbot developed by OpenAI and launched in November 2022. It is built on top of OpenAI's GPT-3 family of large language models and has been fine-tuned (an approach to transfer learning) using both supervised and reinforcement learning techniques.

ChatGPT was launched as a prototype on November 30, 2022, and quickly garnered attention for its detailed responses and articulate answers across many domains of knowledge.[3] Its uneven factual accuracy, however, has been identified as a significant drawback.[4] Following the release of ChatGPT, OpenAI's valuation was estimated at US$29 billion in 2023.[5]

ChatGPT is a member of the generative pre-trained transformer (GPT) family of language models. It was fine-tuned (an approach to transfer learning[6]) over an improved version of OpenAI's GPT-3 known as "GPT 3.5".[7] The fine-tuning process leveraged both supervised learning as well as reinforcement learning in a process called reinforcement learning from human feedback (RLHF).[8][9] Both approaches used human trainers to improve the model's performance. In the case of supervised learning, the model was provided with conversations in which the trainers played both sides: the user and the AI assistant. In the reinforcement learning step, human trainers first ranked responses that the model had created in a previous conversation. These rankings were used to create 'reward models' that the model was further fine-tuned on using several iterations of Proximal Policy Optimization (PPO).[8][10] Proximal Policy Optimization algorithms present a cost-effective benefit to trust region policy optimization algorithms; they negate many of the computationally expensive operations with faster performance.[11][12] The models were trained in collaboration with Microsoft on their Azure supercomputing infrastructure.

 

[eva2000]

Ok, here is an example of how to calculate the approximate cost for your forum:
Let's say we check all users who have less than 50 posts. Let's take an average message length of 1800 characters or 390 tokens and let's assume that users post 200 such messages per day. That is 390 * 200 * $0.002 = $156 per day.

I think your maths are a bit off as $0.002 is per 1000 tokens not per token

• Total tokens per day = 390 tokens per post * 200 posts per day = 78,000 tokens per day
• The total number of tokens used over a 30-day period would be 78,000 tokens per day * 30 days = 2,340,000 tokens
• Based on the current pricing for the ChatGPT GPT-3.5 Turbo model of $0.002 per 1K tokens, the estimated cost for 2,340,000 tokens would be = (2,340,000 / 1000) * $0.002 = $4.68 for 30 days