DEV Community

NLxDoDge
NLxDoDge

Posted on

Hear me out dev.to, Chat GPT against spam πŸ€”?

This topic might already have been touched a lot on dev.to, but what do we think about an auto moderator. That can scan posts for weird or spam content? Maybe even with chat GPT checking the contents of the post and assigning a score based on it for easy moderation.

I frequently check the latest tab, but I keep seeing advertisements, spam and more...

I know we have a a lot of people who regularly use the tools given to report people. But it's getting a bit to much maybe?

Is there already something being done against the bots? (I, for example haven't checked the register page for example, for a Captcha. But thinking about this, we might already have captcha solvers with chat GPT or some other form of bot).

What does dev.to think?

Top comments (4)

Collapse
 
ben profile image
Ben Halpern

Hey!

Yes, we have a pretty solid system to fight spam these days after years of it being a pretty big problem.

We have started using GPT a bit here and there to complement some other text classification techniques.

We, importantly, still have people in the mix to verify and take action, but we are making use of useful AI tools.

Our approach is effectively a add-on to our core Forem codebase, so we're building such that other people could build useful tools themselves. If you have an idea on how to do this, in the future you might be able to build a system which interacts with the Forem API to do this. And if it's awesome, maybe we could use in on DEV πŸ˜…

Collapse
 
nlxdodge profile image
NLxDoDge

Cool to hear, I never really looked into the code base.
I will see what I can do when I have some spare time, as I always wanted to learn/do something with predictions after school. But never really got a case for it.

Collapse
 
devarshishimpi profile image
Devarshi Shimpi

🫑

Collapse
 
jmfayard profile image
Jean-Michel πŸ•΅πŸ»β€β™‚οΈ Fayard • Edited

ChatGPT is a large language model to generate text that imitate humans, I don't follow how that would be useful, also it would probably be prohibitly expensive.

What you have in mind may be a classifier that are used for example to classify in Spam / No Spam
en.wikipedia.org/wiki/Naive_Bayes_...