Most EA effort on pandemic preparedness seems to be focused on pathogen agnostic approaches such as meta-genomic sequencing or PPE stockpiling. Defenses in Depth explicitly advocates against approaches that require categorizing known pathogen types and defending against those specifically. My question is why are we so sure that future pandemics will not belong to known pathogen types. This Metaculus question has only a 6% chance that a pathogen in an unknown family will cause the next pandemic. My understanding is that this would indicate more of a focus on the kinds of interventions that would be effective against known pathogens like a universal flu or coronavirus vaccine instead of putting all of our chips into other pathogen agnostic interventions.
The two reasons I can think of why most EA effort goe toward pathogen agnostic approaches are first EAs tend to view extinction level pandemics as hundreds of thousands of times worse than "ordinary" pandemics and second they believe that catastrophic pandemics are more neglected so the marginal dollar or talent goes further here than toward a universal coronavirus vaccine for example. Is this an accurate characterization or am I wrong about the tractability of pathogen specific approaches?
I think the standard argument here is:
Most of the risk comes from intentional release
If there's an adversary they can aim for the gaps in your system
Reducing attacker options still helps, but helps less against more sophisticated adversaries
Advances in AI are making adversaries more sophisticated across the board
For example, in biosurveillance, where I'm most familiar with the field (but speaking for myself), if the world built out something really good globally that could sensitively detect the top 20 most likely attacks, adversaries would be very interested in figuring out how to create pathogens the system wouldn't detect. If the system was using PCR, which is essentially exact match detection, bypassing the detection system could be as simple as 1-2 codon substitutions.
A separate piece of this is neglectedness, similar to what you raised: there are actually a lot of pathogen specific efforts already happening, and a lot of EA-ish biosecurity work is about shifting the specific/agnostic mix by adding more on the agnostic side. Back to biosurveillance again, traditional biosurveillance is all relatively pathogen specific, whether it's tracking norovirus outbreaks or wastewater PCR for specific known things. There's still effort needed here (ex: once something has been detected you want sensitive and cheap pathogen specific testing deployed widely and rapidly) but we need relatively much more effort in detecting pathogens existing systems won't pick up.