A new bioweapon let lose would be incredibly difficult to contain, and could potentially become something we have to contend with forever
The release of mirror life, or animal-borne pathogens with a long incubation period that infect humans, where the animals are in our cities — is pretty much game over.
I have been talking about this for over a decade now. No one seemed to care.
Giving everyone a simulated wuhan virus lab and hoping NO ONE will do anything bad with it is the stupidest idea ever. The chances of no one releasing anything with a long incubation period are effectively zero. And if anyone releases it, the chances of civilization surviving aren’t great.
All this so we can build slightly better local agents to send email, eh?
But if it's anything like materials design, better computational simulation is one tiny part of the RnD process.
A more credible threat would be a 1-man lab, with a fully automated setup, continuously working on a bioweapon.
But any hosted AI provider would immediately flag that. The risk would be local, abliterated models.
Not saying that the situation is good, but, am I missing something?
Can AI even simulate a virus / protein folding / etc. with any degree of accuracy? or does it just orchestrate open source tools that were already available?