Topic: Using AI to Our advantage in the forum. - page 2. (Read 330 times)

It seems with anything when you add AI, it become a click bait. The idea you are pitching can be just a simple web application. I do not see any intelligence that you require to create a such application.

Although, it's a good recommendation but I doubt anything like that is even needed. To do whatever is said in the post is a tough job and instead of getting any support it will add more burden to the forum. Someone can train AI on their machines but from forum's side I don't think it's even possible.

If someone is good at data sciences then he/she can create an AI model to detect malicious links but I believe that thing can be quite tough and no one will even think about something like that unless they can earn profits from that system.

I doubt we have enough moderators to handle the sheer volume of reports such a system will create by the hour. If you throw in length of posts into the mix, it will be impossible for the mods to work with and the reward is low. Accounts that engage in the most obvious type of spam arebof little value, the person behind the account will simply create another to continue spamming with if one gets nuked.

If there is a way to narrow the filter down to give moderators a workable amount of data, this could be a good idea. I actually think there is a user who uses bots to track spam or something similar, IIRC.

It is no longer news that artificial intelligence has enormously visited the forum and some members are using them to make posts in the forum. This had prompted so many questions about the usage of AI in the forum. I have come to understand that it is wrong to use AI in a forum where humans are supposed to be discussing. While we are fighting the wrong use of AI in the forum,  can we see my idea of using this same AI to the advantage of the forum by involving it in spam detection.

How possible is this?
I am not a data analyst and I also do not have machine learning and deep learning experience, but I am somehow aware of how things happen or how data or datasets are collected. I believe people who has good knowledge of what I'm suggesting will say whether it is possible or not to actualize.

• Collection of sample spam posts

Training an AI with sample of posts upto 100 or above notorious spammers. This will enable the AI to flag or report to moderators any posts that falls to the category it was trained to detect.  On a balance, sample posts of quality posters could also be learnt the AI so that it understands what a quality post could look like. This will improve its accuracy of determining if new posts are spam.

• Feeding the AI model with sample spam phrases and keywords

People who has been here for long or simply say moderators know these spam phrases.  So, the AI is programmed to report to the moderators any posts where these phrases or keywords are found. In the rules and regulations of the forum thread, such phrases were mentioned

• Filtering and reporting posts according to the length of post

In as much as one line posts could also be quality, we know that majority one line posts are spam. So, reporting all one line posts to the moderators by the AI will be of great help. In the same manner unduly long posts will be checked of repetitions and reported to the moderators who inturn decide whether such short or long posts are good.

• Other things the AI could be trained to do

The AI could also be trained to analyse links to check the malicious ones and spam links.
The AI will detect how frequently a poster posts, it could report high number of posts within a short period of time which mostly are spam posts.

Pros and Cons
If this is implemented, it will lessen the burden on moderation and post reporting and thereby giving us a more spam free forum. This idea can also be tested in a particular board like Trading discussion and see how it performs before introducing it to the larger forum.
I also know that the AI will not report with 100% accuracy,  that is why I had suggested that they will only flag a post and report to the moderators who will do the final moderation. If it is trusted overtime,  who knows they could be incorporated into moderation.