I was sitting on the bus scrolling Reddit when I found multiple AI-generated posts, all coming from accounts that are known to be AI. Since Reddit and other platforms don’t want to fix the problem, let’s do it ourselves!
My idea is to create a web browser plugin that works like uBlock or SponsorBlock, where users can submit known usernames of AI-generated content, then it would filter them out. Who would be interested?
It sounds incredibly hard to maintain.
Kellen said:
It sounds incredibly hard to maintain.
It is, and you’d have to discern between fake/wrong submissions, users partially using AI, fake positives, etc. Too tedious.
Kellen said:
It sounds incredibly hard to maintain.
It is, and you’d have to discern between fake/wrong submissions, users partially using AI, fake positives, etc. Too tedious.
And AI that serves a specific purpose, like bot posts that filtering out is kind of missing the point.
Kellen said:
It sounds incredibly hard to maintain.
It is, and you’d have to discern between fake/wrong submissions, users partially using AI, fake positives, etc. Too tedious.
Let the users do it?
Not hard to set a benchmark where the system flags enough user flags as AI.
@Dane
I’ll be honest, I wouldn’t trust my content to be filtered by a majority of users, a select group of users, or anyone in general.
Also, think of it as karma here; a lot will just submit based on emotion, benefit, or Dunning-Kruger… 
@Vann
Agreed, not to mention the bots that dilute the result. People profiting on such content would 100% screw user-maintained results.
@Dane
I am worried about how to protect it from fake AI accounts that would downvote good submissions. Maybe you have to earn karma to spend it or always have a ‘positive KD ratio’ to downvote.
Cassian said:
@Dane
I am worried about how to protect it from fake AI accounts that would downvote good submissions. Maybe you have to earn karma to spend it or always have a ‘positive KD ratio’ to downvote.
Can’t have it without trying. I think you have a good idea going.
Kellen said:
It sounds incredibly hard to maintain.
It is, and you’d have to discern between fake/wrong submissions, users partially using AI, fake positives, etc. Too tedious.
Well, you would have ranks; people with lots of rep would have more decision power, maybe a ledger crypto network.
Of course, it’s harder to gain rep than to lose it, just like IRL.
@Cassian
You had me until crypto network, lmao. I think it’s a good idea, honestly. It’s not hard to insert a basic flag system into the site - and then just track the results.
Dane said:
@Cassian
You had me until crypto network, lmao. I think it’s a good idea, honestly. It’s not hard to insert a basic flag system into the site - and then just track the results.
By crypto I don’t mean those shitcoins but the idea of an immutable ledger, but it might fall face-down as the ledger will be gigabytes and tera and peta in size as it ages. I am fishing for ideas, hoping to spark someone to create their own vision too.
@Cassian
You gain no benefit from a blockchain here. Falsified history is not one of your threats. You want locally stored filter lists, similar to what uBlock does.
Hadi said:
@Cassian
You gain no benefit from a blockchain here. Falsified history is not one of your threats. You want locally stored filter lists, similar to what uBlock does.
I just want to know the Venn diagram of people who hate AI but like crypto. OP and like 3 other people?
Kellen said:
It sounds incredibly hard to maintain.
We use AI to clean up submissions to the AI filtering app.
Kellen said:
It sounds incredibly hard to maintain.
No pain, no gain; AI is taking over, the world has to react. What do you think would be hard to maintain?
Kellen said:
It sounds incredibly hard to maintain.
No pain, no gain; AI is taking over, the world has to react. What do you think would be hard to maintain?
If you’ve got the technical capability to build this, then I’d recommend you start it off and make a commitment to maintain it by yourself for a full year.
If you don’t have the technical knowledge or capability to build and maintain it yourself, then I’d recommend listening to people who likely know what they’re talking about, instead of saying they’re wrong and that they should do something about it.