Do other generative AI services also do this? Or do they have functioning safeguards?
Usually there are safeguards against nsfw content on public services.
Asking for a friend?
No. I was just interested because I’ve seen a lot of headlines about how Grok creates all these problematic images of women, minors etc., but not about other similar generative AI software. But my understanding so far has been that there’s no real way to safeguard generative AI at all, so I was wondering whether this was a Grok-only problem, and if so, how others were avoiding it.
🤨
These people and journalists don’t see to remember celebrity photoshops. Or maybe they do and want to continue to feed the outrage machine.
Pardon me if I duck out of my Two Minutes Hate for today.
You might be interested in 404 media’s coverage. They’ve been reporting on it for a while – it’s been part of Sam Cole’s beat since the vice days.
4chan has been doing this for like a decade, why only now is it getting attention?
I assume it’s for yet another push to restrict access to even more of the internet.
Probably bc you had to have the barest modicum of photoshop skills back in the day to pull it off, but now any idiot can upload a pic and click the nekkid button
Photoshop was for 20 years ago. There were automated tools by the late 2010s. Results were mixed, I threw a picture of Trump in and then threw up.
4chan is a relatively sequestered part of the internet, and doing this required a level of skill.
There’s a reason people freaked out when deepfakes - first photos then videos - started appearing as it lowered the level of entry for creeps. No longer did you need to learn to use a tool through hundreds of hours of not well documented features to achieve nudes, but even the dumbest dipshit with minimal tech skills and money could download the right models, and either rent or buy the right hardware and get lifelike deepfakes, photos or videos, within minutes or hours…
That quickly passed fortunately, but then came the combined agentic models that could use input images, LLMs, and other generative AI to replace the previously lengthy processes with a much quicker approach through cloud hardware.
Aka grok. Which, let’s be honest, is an even more idiot-proof solution as all it requires is a Twitter account and marginal prompt engineering skills (which is quite well documented already), so even the worst trolls can easily get going (and here by worst I mean the wannabe trolls who aren’t even good enough to be basic trolls).
Yet again the level of entrance was lowered significantly.
On a sidenote: I fucking LOVE !degenerate@lemmynsfw.com 🤩
As if being naked is unhuman…
