I don’t understand this. Train Grok to do WHAT?
1 Like
Training simply means it scraped some images or other content from the cloud. These LLMs are not training for any specific purpose. Perhaps that is an unfortunate term to use but it is the term of art in the LLM world. Quote from the Ars Technica article, which in turn quotes from the complaint:
“Because Grok’s terms treat public X posts and Grok’s own outputs as training data by default, publicly posting an image does not just expose it to viewers, but also feeds [it] directly into the pipeline xAI uses to train and improve its model and thereby generate further images,” Doe’s complaint said.
2 Likes
So it’s using it to make child porn?
1 Like
“It” doesn’t make anything on “Its” own. Apparently, “It” scraped some images of this victim from the cloud and then humans are able to prompt “It” to make more images. Presumably, these prompts describe whatever messed up things these (sub)humans imagine. The responsibility for these images rests squarely on the shoulders of the individuals who cause their creation. The owners of these LLMs are accountable only inasmuch as they have not been diligent in removing these images and in installing guardrails to make it hard for users to create these illegal images.
Quoting from the complaint:
Generative AI models are now able to generate photorealistic images based on simple user prompts. In conjunction with text prompts, users can also feed models pre-existing images or videos to re-publish, alter, or emulate.
Do not make the mistake of anthropomorphizing this computer code and database, which is all that LLMs are.
3 Likes