3 ms·
AI Bests Virus Experts, Raising Biohazard Fears
- Workaccount2 2y agoThe best way to combat this now is to probably not talk about it. Just like vulnerabilities in white hat scenarios, let the developers know and then have a lead time before releasing the information publicly. Ironically this study comes from two safety oriented organizations, so I question their reasoning for running to make it public that you can use SOTA models right now to do the knowledge legwork for creating deadly bioweapons.
- throwawaymaths 2y agowho is the developers that need to patch thr vuln? mother nature?
- gus_massa 2y agoThe title may be confusing. They claim that the AIs can be a good advisors to fix problems that someone find doing experiments and giving hints like: fake quote> You are overcooking the virus, remember to keep the temperature below 37°C (100°F) As far as I read, the AIs don't propose new virus variants.
- Workaccount2 2y agoLLM providers so they can add in hard filters, at least for now.
- mistrial9 2y agoDoes this mean that engineering a virus at a multi-government lab and then it escapes, causing draconian lockdown and forced medications across the world, might really happen now ?
- EdwardDiego 2y agoDon't need to hitch your little weird obsession to any passing car mate.
- candiddevmike 2y agoWhile I don't necessarily agree with everything you said, it's sad that we didn't learn doing boring, human researcher GoF research is bad. Now we are going to do AI assisted GoF research.
- klabb3 2y agoThey only really claimed to believe in the conspiracy theory of lab leak, which many do and isn’t at all outrageous imo. But to your point, I 100% agree. Even if Covid was zoonotic, it absolutely could have been a human fuckup. The amount of lab leaks over the years, including things like Anthrax, Polio, SARS and even Covid (yes, in Taiwan after the initial outbreak) from BSL3 and 4, are just waaay to common. After a plane crash, there are almost always changes to protocol, equipment or even the aircraft. Here, we had millions of casualties globally, and this risk is as real now as yesterday. I get that we have pandemic fatigue, and nobody wants to think about it. But if keep playing with fire, it will happen again, sooner or later.
- _mlbt 2y agoThe lab leak hypothesis is by far the most likely explanation. The CCP is not a trustworthy actor, so their version of the events need to be taken with massive skepticism. Once you ignore their narrative and look at the actual facts, it becomes obvious.
- lawlessone 2y agoThat GoF research saves us time by figuring out the most probable routes or ways a virus could mutate to something worse.
- _mlbt 2y ago
- EdwardDiego 2y agoSure thing Jan. I'll just read the paper when they actually publish it.
- gus_massa 2y agoThe preprint is in https://www.virologytest.ai/vct_paper.pdf https://www.virologytest.ai/vct_paper.pdf , but I can't find if they submitted it to peer review.
- lawlessone 2y ago> AI models like ChatGPT and Claude now outperform PhD-level virologists in problem-solving in wet labs OK, if you said an AI made specifically for generating virus genes i might think maybe.. I find it very hard to believe the chatbots can.
- lubujackson 2y ago"Oh, you're right, that virus would wipe out humanity. I have reconfigured the virus to be safe for humans. Deploy?"
- yawnxyz 2y agoif anyone's ever worked in a lab, it's abundantly clear even when you know Every single step / have a well-defined protocol for any assay, most ppl including PhDs and Postdocs regularly mess it up. There are so many things go wrong for even the most basic things it's amazing how anyone gets microbiology done in the first place
- Alifatisk 2y agoSo you mean to tell me that LLMs, who scramble together words, manage to outperform PhDs in virology? How is this possible? I can’t comprehend this.
- josefritzishere 2y agoIt does read like it was written by Sam Altman.
- alganet 2y agoI'm proud of you, Anakin. You brought hope to those who had none.
- AIPedant 2y agoI was outraged to learn that the test is basically a combination of multiple-choice and true-false. What a misleading pile of crap this study is! a) Doing well on multiple-choice tests does not really imply anything about doing well in a real lab setting, especially for an AI. b) Using multiple-choice tests to compare LLMs to humans has an obvious flaw: LLMs are probably superhuman at guessing multiple-choice answers based on superficial statistics! We all learned as children that there are ways to game multiple-choice questions even if you have no idea what the answer is. c) It is just unacceptably lazy for AI researchers to do this. They aren't teachers on a tight deadline. There is no justification whatsoever for AI researchers to use Scantrons. I am truly disgusted with the poor scientific rigor in AI research. So depressing.