Skip Navigation
SneerClub @awful.systems Sailor Sega Saturn @awful.systems

Building an early warning system for LLM-aided biological threat creation

openai.com Building an early warning system for LLM-aided biological threat creation

We’re developing a blueprint for evaluating the risk that a large language model (LLM) could aid someone in creating a biological threat. In an evaluation involving both biology experts and students, we found that GPT-4 provides at most a mild uplift in biological threat creation accuracy. Whil...

Building an early warning system for LLM-aided biological threat creation

OpenAI blog post: https://openai.com/research/building-an-early-warning-system-for-llm-aided-biological-threat-creation

Orange discuss: https://news.ycombinator.com/item?id=39207291

I don't have any particular section to call out. May post thoughts tomorrow today it's after midnight oh gosh, but wanted to post since I knew ya'll'd be interested in this.

Terrorists could use autocorrect according to OpenAI! Discuss!

98
Hacker News @lemmy.smeargle.fans bot @lemmy.smeargle.fans
BOT
Building an early warning system for LLM-aided biological threat creation

You're viewing part of a thread.

Show Context
98 comments

  • Look, ultimately why I'm focusing on technical capacity is because I don't disagree on the political side of this. I think it's highly unlikely that anyone would use bio weapons or develop them soon, and it's definitely true that it's inextricably political. My point is ultimately that bio weapons are getting easier and easier to produce and it's not a non issue to consider. The top comments here were more than happy to dismiss bioweapons wholesale as even a thing that could happen.

    I'll check out your book recs. I'll point out that the main book on this topic folks have cited is ten years old now, and we're experiencing a legitimate bio tech revolution right now. How much further it'll go is anybody's guess.

    • I understand that you’re olive branching me here, but I don’t accept “How much further it’ll go is anybody’s guess”. Trends are analysable, and the sources of projections are equally analysable. A book that’s ten years old is far better than (a) 30-year old (and more) newspaper-level stuff, without citations, about backyard anthrax, (b) nothing, and (c) two links to tangentially related reports, and you’ve brought those three.

      I am seriously concerned about the confluence of two things: (1) how closely your comments here mirror, right down to the level of language, press releases and opinion columns paraphrasing press releases, some of them (the anthrax stuff) extremely old hat; (2) the level of outrage and confidence you bring to the table when challenged on this and similar. Phrases like “How much further it’ll go is anybody’s guess” are press release language - they have absolutely no place in serious discussion, but they have a powerful rhetorical effect which allows them to displace serious analysis, and that displacement furthers specific, analysable, interpretable sectoral and political interests.

      The same goes for “bio tech revolution” - you are never clear, in any of this, what that actually entails. What you do is cite possibility and unknowability, in a manner innovated precisely by sectoral and political interests from the 1950s onward. You have no detail of any value, and you write off actual detail with speculation and glib remarks about the age of the detail you’re given - that is a political innovation to which you have allowed yourself to be susceptible. You may also try on Naomi Oreskes for size as an author who grapples with this in both directions.

      • Sure a book that's ten years old is better than 30 year old papers. Doesn't mean that it can't be wrong, especially in the face of technological change. Digging into how modern biotech is constantly breaking ground is difficult, there's no one thing happening. Biocrystallization is used commercially, we produce drugs using modified organism as, we're understanding genetics at genome network/hologenetic levels like never before with new computational and statistical tools. Immune system level effects are getting easier to control. In part I'm not being specific to not oust synthetic biologist friends of mine, in part because the conversation hasn't called for it, and in part it's not my field, just close to it. (Biomedical)

        Technological trends at this scale are notoriously impossible (in fact, there's a whole field of research on it trying to figure out how to predict breakthroughs for funding reasons) to predict. We could find out tomorrow that genetics is incomprehensibly complicated in a way that defies most of the use cases for bio tech that we'd like, or alternatively, we find new tools which incrementally make it easier to predict phenotypic effects from genotypic changes. My bet is on the latter, but the prior isn't a ridiculous position to hold.

        Here's a cool presentation about where macroscopic synthetic biology is at: https://m.youtube.com/watch?v=1qRIetbuoH4

        It's not for bioweapons strictly, my interests in this field tend to be more about macroscopic synth bio, but perhaps illustrative of why this field is not easily summarized and what sorts of leaps have been made and what challenges are obviously remaining.

        Your comments on my rhetoric are frankly, rhetorically speaking, grasping at straws. Vagary isn't always a conspiracy developed by powers that be. Sometimes its because the person you're talking to feels you're arguing in bad faith, is tired when they have opportunities to respond, and isn't planning on writing a white paper for your digestion. This isn't a summit on bio warfare, this was me going "hey bio weapons aren't a technological dud" and you deciding to be "seriously concerned" about someone having the audacity to disagree with your stance. Feel free to elaborate on your "serious concern" or don't. We are here to discuss different things, and you're showing nothing but adversarial participation.

98 comments