The avatar, created to deter young people from extremism, has been subverted and is breaking out of niche online silos
Her name is Amelia, a purple-haired “goth girl” who proudly carries a mini union flag and appears to have a penchant for racism.
If you are unfamiliar with Amelia, the chances are you will soon encounter one viral meme or another inspired by her on Facebook or X, where her reputation is growing.
Videos of Amelia typically feature her walking through London, or the House of Commons, declaring her love for England and warning of the dangers of “militant Muslims” or “third-world migrants”. In one clip she is harangued by bearded man in Islamic attire for eating a pork sausage.
Oh no what are we going to do… a far right AI-avatar! People know that you can create 10000+ left, center, etc. AI-avatars, as well as fact checking AI-avatars debunking what this one says, right?
No way am I trusting an AI fact checker
That would be for people that are trusting the original AI.
It could, among other things, link to sources.
It doesn’t even have to be accurate. It suffices if it makes clear that there are different perspectives, or that you simply can’t trust anything. People exhausted by contradicting slop is a better outcome than allowing only one type of slop to take over.
A lot of genAI now links to sources, that are frequently wrong, and I haven’t noticed that changing fanboys obnoxious praising them
You’re suggesting the solution to AI slop spewing racism is more AI slop somehow “debunking” racism? How would that work exactly? This sounds like a propaganda wet dream.
You should look up how Russian style propaganda works. It isn’t by lying convincingly but rather by flooding all channels with so much noise that it requires major effort to determine what’s true and what isn’t. When you’re drowning in bullshit most people will give up and just decide to believe whatever agrees with their personal opinions rather than whatever the reality is and it’s a lot easier to manipulate vibes than it is reality.
I think that flooding the zone with alternative slop would have a neutralizing effect. At the end everyone gets exhausted, which is a better outcome than allowing right wing slop to spread completely unhindered. Of course people will then start protecting their spaces, moderating the slop they don’t like, but you‘d at least have neutralized the public spaces, which would be a win.
What you’re describing is literally the KGBs playbook and no it doesn’t work in anyone’s favor but the ones spreading propaganda. All you’ll end up with is an even less informed and apathetic populace and we’re already struggling with that problem. That works in favor of dictators and fascists. Democracies on the other hand require an informed and engaged populace in order to function.
I think the final form of this conflict is dead internet theory territory.
It’s another scenario that is hard to predict the end game because AI at this scale has not been seen before.
That’s what „flood the zone“ ultimately means. It’s effectively no different than the other solution usually proposed, which is banning social media. Except than bans can be avoided, people will seek to challenge them, etc.
What if you don’t use it as only measure? You launch 2 things in parallel: 1. neutralization: flood the zone, exhaust everyone, 2. provide a solution: sane channels that make clear how thinking, fact checking and honest dialogue are the only way out.
The problem is that the algorithms optimize for this kind of content. The people seeing this LLM avatar wouldn’t start seeing the “good” LLM avatars. Instead, content in the “good” spaces will become watered down with low quality/ rage-bait-y slop
It depends on how you present it. It doesn’t have to be „boring good“. And even then you can still work e.g. on the comment area (reactive, not original content).