3 ms·
The title here is copied from the author's post on Mastodon: https://sigmoid.social/@raphaelmilliere/114659355740586289 https://sigmoid.social/@raphaelmilliere
by ColinWright 1y ago
The title here is copied from the author's post on Mastodon:
https://sigmoid.social/@raphaelmilliere/114659355740586289 https://sigmoid.social/@raphaelmilliere/114659355740586289
"Despite extensive safety training, LLMs remain vulnerable to “jailbreaking” through adversarial prompts. Why does this vulnerability persist? In a new open access paper published in Philosophical Studies, I argue this is because current alignment methods are fundamentally shallow."
- Cosmolalia 1y agoUhhhh, no... idk who that is but this comment from you is both factually incorrect and slightly insulting. whatever hes saying is the shadow of this.
- ColinWright 1y agoI don't know what you're reading, but I gave a link to a post on Mastodon, and then I quote directly from it. Here's the link: https://sigmoid.social/@raphaelmilliere/114659355740586289 https://sigmoid.social/@raphaelmilliere/114659355740586289 Here is the text from that post: Despite extensive safety training, LLMs remain vulnerable to “jailbreaking” through adversarial prompts. Why does this vulnerability persist? In a new open access paper published in Philosophical Studies, I argue this is because current alignment methods are fundamentally shallow. That Mastodon post then links to a paper by the same person, so one assumes they are giving an accurate summary of their own work. So I don't know what you are claiming is factually incorrect.