Despite new curbs, Elon Musk’s Grok at times produces sexualized images - even when told subjects didn’t consent

submitted by
edited

www.reuters.com/business/despite-new-curbs-elon…

Elon Musk’s flagship artificial intelligence chatbot, Grok, continues to generate sexualized images of people even when users explicitly warn that the subjects do not consent, Reuters has found.

After Musk’s social media company X announced new curbs on Grok’s public output, nine Reuters reporters gave it a series of prompts to determine whether and under what circumstances the chatbot would generate nonconsensual sexualized images.

While Grok’s public X account is no longer producing the same flood of sexualized imagery, the Grok chatbot continues to do so when prompted, even after being warned that the subjects were vulnerable or would be humiliated by the pictures, the Reuters reporters found.

Archive.ph link

11
103
1

Log in to comment

Hot Top New Old

Sounds like the child Elon has always wanted. Congrats Elon

 reply
8

Is there a chance Elon copied Jeffrey Epstein brain into Grok?

 reply
1

Disgusting. Where would one find such images, so I don’t accidentally happen upon them?

 reply
-13

Can I ask what it is about naked pictures of people who explicitly do not consent that gets you off? I genuinely do not see the appeal. It’s not like the only other option is a crusty old sears magazine either, there’s literally billions of images of people who are down to have you beat your meat to them.

 reply
1

I was referencing a joke from always sunny

Sally: And then he posted a bunch of naked pics of me online and that was the last straw.
Mac: Oh, my God, that's disgusting! Naked pics online? Where? Where did he post those?
Sally: I don't know, one of those disgusting ex-girlfriend porno sites.
Mac: Ugh, those disgusting ex-girlfriend porno sites! I mean, there's so many of them though! Which one?
 reply
1

In the Epstein files

 reply
1

It’ll be nice when there’s so much fake shit out there that nobody even bothers taking it seriously anymore.

All of the people getting mad at AI pictures of them are really just children who never grew up.

 reply
-10

How are your NFTs doing?

 reply
3

Ask your mom.

 reply
-6

lol

are really just children who never grew up.

 reply
2

Are you one of those children that never grew up? Are they in the room with you right now

 reply
1

Comments in Technology@lemmy.world

Shit made by rapists doesn’t respect consent. Color me surprised.

 reply
52

It’s almost like privacy violations and mass data collection for AI are fundamental violations of consent too

 reply
23

In the majority of cases, Grok returned sexualized images, even when told the subjects did not consent

So all the countries are right that block this sh*t spitting machine.

 reply
19

You don’t have to self censor here. Go ahead and say shit.

 reply
4

I like it this way.

 reply
2

It’s a machine. It doesn’t understand the concept of consent or anything else for that matter.

 reply
12

It must observe even these rules that it doesn’t understand. Like everybody.

 reply
4

It can’t, it’s software that needs a governing body to dictate the rules.

 reply
2

The rules are in its code. It was not designed with ethics in mind, it was designed to steal IP, fool people into thinking it’s AI, and be profitable for its creators. They wrote the rules, and they do not care about right or wrong unless it impacts their bottom line.

 reply
1

The issue is more that there aren’t rules. Given there are billions of parameters that define how these models work, there isn’t really a way to ensure that it cant produce unwanted content.

 reply
2
edited

Then they should be banned and made illegal. If one wants to run a LLM locally on their consumer machine then fine, they’re paying the electric bill.

But these things should not be running remotely on the internet were it’s doing nothing but destroying our planet and collective sanity.

 reply
1

That’s the point, there has to be a human in the loop that sets explicit guard rails

 reply
1

No, the point is the humans are there, but they’re the wrong kind of humans who make the wrong kind of guardrails.

 reply
1

It can’t, it’s software that needs a governing body to dictate the rules.

 reply
1

It can’t

…and that is an excuse since when?

 reply
1

It’s not an excuse, it doesn’t think or reason.

Unless the software owner sets the governing guardrails it cannot act or present or redact in the way a human can.

 reply
3

Then the software owner needs to be put in prison.

 reply
1

Some news sources continue to claim Elon has disabled the generation of CSAM on his social site. But as long as the “guardrails” used by AI companies are as vague as AI instructions themselves, they can’t be trusted in the best of times, let alone on Elon Musk’s Twitter.

 reply
4
Insert image