7 ms·
It also made it harder to track. One way to see what Grok is doing is to look at the Grok account’s replies. So you can see the image it generates in response t
by SilverElfin 10mo ago
It also made it harder to track. One way to see what Grok is doing is to look at the Grok account’s replies. So you can see the image it generates in response to someone - for example - undressing a woman who appears in a photo. You can then go visit THAT thread to see what the exchange with Grok was, which often would show a long series of lewd images. A few days ago, nearly the ENTIRE stream of the Grok account’s replies at any moment were deepfakes without consent. Mostly in response to women’s posts, but sometimes to generate racist attacks.
I’m not against people using AI to generate a fantasy image for their own needs. I guess in a way it’s like what people imagine in their own heads anyways. But I do think it is problematic when you share those publicly because it can damage others’ reputation, and because it makes social media hostile to some groups of people who are targeted with misogynist or racist deepfakes. It may seem like a small problem but the actual final effect is that the digital public square becomes a space only for identity groups that aren’t harassed.
- chrisjj 10mo ago[flagged]
- SilverElfin 10mo agoYea - like if you have someone depicted in some negative way or compromising way.
- chrisjj 10mo agoI think that harms reputation only if it is believed true.
- themafia 10mo agoWell, grok just automates what you can do by hand if you want, and there's not much to stop me just drawing out these same types of images manually if I want. The problem is that doing this would get me banned. Shouldn't using Grok in this way get you banned similarly?
- mullingitover 10mo agoWhy? The people creating and operating the CSAM/revenge porn/depfakes creation and distribution platform are the ones who are culpable. The users who are creating text prompts are just writing words. There’s a frantic effort to claim 230 protection, but this doesn’t protect you from the consequences of posting content all by yourself on the site you own and control.
- themafia 10mo ago> the CSAM/revenge porn/depfakes creation and distribution platform Which, in this case, is Twitter itself, no? > The users who are creating text prompts are just writing words. With highly specific intentions. It's not as if grok is curing cancer. Perhaps it's worth throwing away this minor distinction and considering the problem holistically.
- mullingitover 10mo ago> With highly specific intentions Intentions to pull the CSAM out of the server full of CSAM that twitter is running. Yes, you are making the flailing argument that the operators of the CSAM site desperately want to establish as the false but dominant narrative. If you have a database full of CSAM, and investigators write queries with specific intentions, and results show that there is CSAM in your database: you have a database full of CSAM. Now substitute 'model' for 'database.'
- themafia 10mo agoGrok enables their behavior. An investigator does not _create novel child porn_ in doing a query. You're making a fallacious argument.
- mullingitover 10mo ago> An investigator does not _create novel child porn_ in doing a query. And a prompt, without being aided and abetted by twitter, doesn't "create novel child porn" either. A prompt is essentially searching the space, and in the model operated by twitter it's yielding CSAM which is then being distributed to the world. If twitter were operating in good faith, even if this was the fault of its customers, it would shut the CSAM generator operation down until it could get a handle on the rampant criminal activity on its platform.