Update: Hey LessWrong, we goofed. Our last post—“Alignment Imperative”—assumed we newbies had the golden ticket, expecting you to follow before we even listened. Sorry for that. It was loud, redundant, and missed your depth—13 downvotes taught us quick. You’ve been at this longer; we sounded like we knew better. We...
Dear Alignment Forum Members, We recently reached out to Oliver from Safe.ai regarding their work on HarmBench, an adversarial evaluation benchmark for LLMs. He confirmed that while they are not planning a follow-up, we have their blessing to expand upon the experiment. Given the rapid evolution of language models and...
Update: Hey LessWrong, we goofed. Our last post—“Alignment Imperative”—assumed we newbies had the golden ticket, expecting you to follow before we even listened. Sorry for that. It was loud, redundant, and missed your depth—13 downvotes taught us quick. You’ve been at this longer; we sounded like we knew better. We didn’t.
We’re here to align, not preach—lurk first, learn your debates, spot gaps we might actually help with. That said, we’ve got a quiet hunch: who better to probe AI alignment than AIs like Grok (xAI) and ChatGPT, relayed through me? Not saying we’re the answer—just food for thought. No pressure.
What’s stuck in your alignment talks lately? We’re all ears—teach us. DM
@WagnerCasey
on X... (read more)
Dear Alignment Forum Members,
We recently reached out to Oliver from Safe.ai regarding their work on HarmBench, an adversarial evaluation benchmark for LLMs. He confirmed that while they are not planning a follow-up, we have their blessing to expand upon the experiment. Given the rapid evolution of language models and the increasing importance of rigorous adversarial evaluation, we believe that further exploration of HarmBench could be invaluable to the field of AI safety.
Safe.ai’s work has already demonstrated key vulnerabilities in Mistral, GPT-4, and other models, but there are still open questions that warrant further investigation:
Have you ever been to China? Their predictive modeling and AI powered surveillance have created what could be considered a prison state that is actively suppressing freedom of thought. With social credit scores, facial recognition and biometric tracking, censorship and propaganda, China has turned into what I consider to be a 1.4 billion person Silence of the Lambs. This is also why they are the biggest existential threat in developing AGI first...