AI Bests Virus Experts, Raising Biohazard Fears | Hacker News Reader

AI Bests Virus Experts, Raising Biohazard Fears

time.com

31 pointsrbanffy28 comments

if 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

> 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.

"Oh, you're right, that virus would wipe out humanity. I have reconfigured the virus to be safe for humans. Deploy?"

I 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.

The 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.

who is the developers that need to patch thr vuln? mother nature?

The 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.

LLM providers so they can add in hard filters, at least for now.

Sure thing Jan. I'll just read the paper when they actually publish it.

The preprint is in https://www.virologytest.ai/vct_paper.pdf , but I can't find if they submitted it to peer review.

Does 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 ?

Don't need to hitch your little weird obsession to any passing car mate.

While 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.

They 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.

The 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.

That GoF research saves us time by figuring out the most probable routes or ways a virus could mutate to something worse.

In theory that is true but it all goes out the window when one of these viruses escapes lab containment and causes a global pandemic.

> multi-government lab

What are you talking about? The Wuhan Institute of Virology is firmly located on sovereign Chinese territory. In fact, the main reason why any serious scientist should still give the lab leak theory the time of day is due to the fact that China was so cagey with what was going on at the WIV. If this lab was multi-governmental we would have the data China withheld and the lab leak theory wouldn't have any shadow to lie in.

Furthermore, the real problem with the lab leak theory is that it does not matter whether or not it is true. If COVID-19 originated from animal sources, then we should regulate wet markets, crack down on the endangered animal trade, and stop doing gain-of-function research. If COVID-19 originated from a lab, then we should stop doing gain-of-function research, regulate wet markets, and crack down on the endangered animal trade. All of those things are very prudent things to do if you want to slow down the introduction of novel pathogens.

Other than politics why would you believe any other scenario to be more likely than the lab leak hypothesis at this point? It’s far too much of a coincidence that patient zero was in such close proximity to the WIV that studies and researches viruses exactly like COVID-19. The actions of the CCP and those aligned with their interests just further cements the likelihood of this hypothesis.

Because the closest relative of the sars-cov-2 virus is found in animals in Thailand, and just before the pandemic, exotic animals from Thailand were imported directly to the Wuhan wet market, which was a regional hub for importing animals. The WIV lab did not have a copy of that virus. The original sars-cov-1 virus emerged at an animal farm upriver from Hong Kong. The MERS virus emerged at a camel farm. Seeing a pattern yet?

> The WIV lab did not have a copy of that virus.

According to the CCP, whose version of the events cannot be trusted. The reality is that they had the time, means, and motivation to destroy any samples and records related to COVID.

Also speaking of patterns, the WIV has a history of safety lapses and was known to be working with SARS family viruses in BSL-2 labs, which are wholly inadequate.

According to the Thai virus researchers who did not even discover the related virus for the first time until the year AFTER tbe pandemic started. So no, unless someone has a time machine the WIV did not have a sample in 2019. The consensus of the international virus research community is that the WIV did NOT have a sample of the covid-precursor virus.

Animals infected with the Thai virus were imported from Thailand to the Wuhan wet market. People on the north side of the river within 100 meters of the wet market got sick. People on the south side of the river within 15km of the WIV did not get sick. We are not relying on the CCP for the information that people on the south side of the river did not get sick. There were foreign citizens living south of the river and we got our information from them.

The evidence favors spread from market:

Here is a review paper of all the evidence so far:

https://www.annualreviews.org/content/journals/10.1146/annur...

Not true, there is significant evidence that actually favors the lab leak hypothesis…

https://www.realclearscience.com/articles/2025/03/28/covid-1...

Additionally the Biden administration and other world governments suppressed the lab leak hypothesis which means that researchers seeking government funding had an incentive to support the official wet market narrative instead…

https://thehill.com/opinion/healthcare/5058339-biden-adminis...

The only paper they cite was the bat virus paper in cell, which isn’t evidence for lab leak.

The review paper I mentioned shows dozens of papers showing evidence of market as initial spread.

So 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.

It does read like it was written by Sam Altman.

I'm proud of you, Anakin. You brought hope to those who had none.

Reply on news.ycombinator.com