Post #4089993

For some reason I ended up writing about why blocking biology requests is the wrong approach to avoiding a risk of bioterrorism and biosafety with LLMs. Most people talk about the cybersecurity side of these guardrails. Felt giving the other site some attention was worth a shot. I think it's slightly interesting because it introduces actual ideas from counterterrorism, and a real report from a government that has a non-politicized institution to prevent this, which was written in a language that most people don't speak, and so miss out on insights from typically. Also tried to add some nuance to what to do instead, and how to do it without e.g. hampering valid scientific goals. Even if that turns out to be really hard, and was a problem way before AIs were a thing that carried biosafety/biothreat risks. https://point.free/blog/the-ai-bioweapon-risk-isnt-jailbreaks/