I'm a (new) biomedical engineering MS student, so this is based on a few years of undergraduate and graduate-level study in the realm of biology.

For example, could there be some kind of engineered virus that contaminates all food sources on the planet? I don't know and would be interested to hear from folks about that.

Viruses are typically specific to a particular organism or set of related organisms. Some have very narrow host specificity, while others are more broad. The same is true of bacteria and fungi.

So your question can be refined slightly by asking, "how hard would it be to engineer, culture, and deploy a set of viri, bacteria, and/or fungi capable of killing all food sources on the planet, and what prevention or mitigation strategies exist to combat this outcome?"

Understand that the engineering, culturing, and deployment of each pathogen within this set of bioweapons would at present be extremely difficult, requiring a high level of technical expertise, time, money, and government support. The efficacy of each pathogen would need to be confirmed, and then it would need to be stored or maintained until the attacker had accumulated a set of bioweapons with efficacy against a sufficiently broad set of food sources to threaten civilizational collapse or human extinction.

This process would be coming against a backdrop of improving ability to rapidly develop and roll out vaccines, by a scientific apparatus that is vastly bigger than the resources available to the attacker.

Furthermore, as you may know, we had an mRNA vaccine within days of sequencing COVID. Delay in vaccine rollout was for manufacturing and testing. With animals facing a virus, we could develop an mRNA vaccine, test it with RCTs (with no particular concern for safety, only efficacy), and then start vaccinating immediately.

Bacteria, fungi, and parasites are potentially harder targets, but these pathogens are also harder to engineer and spread because their biology is more complex and they often require insect or animal vectors to spread. These vectors can be killed or restricted geographically.

We also have far fewer ethical barriers to re-engineering non-human animals in ways that might make them more resistant to a particular pathogen. We can also use chemical and physical barriers to screen them to arbitrary degrees. If necessary, food sources could even be grown in sterilized airtight containers. This would be extremely logistically complicated and energetically expensive, but it could be done if necessary to safeguard the survival of humanity.

An extreme nuclear winter scenario could wipe out all photosynthetic organisms, eliminating the base of the food chain. Even in this scenario, however, it's plausible that we could use nuclear energy or fossil fuels to produce artificial light in adequate quantities to maintain some agriculture. There have been visions that adequate supply of solar energy and an "energy explosion" would allow us to transition to high-productivity vertical farming just to enhance the abundance and quality of crops in a "world as usual" scenario.

I don't know any details of this proposal, but this would potentially protect us against nuclear winter and harden us significantly against biorisk all at once. It seems likely to me that vertical farms would have some sort of shield surrounding them to screen out pests even in a world that's not facing an unusually severe biorisk. Shielded vertical farms seem like a great mitigation strategy that would not only protect humanity against such a bioweapon, but also disincentivize an attacker to pursue such a strategy in the first place.

Bottom Line:

Insofar as bioweapons are an existential risk, the vast bulk of that risk will be from pathogens that attack the human body.

Further Thoughts:

Your endpoint is that humans would likely survive even an extremely severe bioweapon attack, and could then rebuild civilization. Yet this world-state appears fragile.

Unlike AI safety or "grey goo," an outcome like this would almost certainly be due to a deliberate human attack on other humans, with apocalyptic or world-controlling intentions.

The attacker would know in advance about the expected outcome. They might conceivably harden themselves to survive it, perhaps by preparing their own bunkers.

That attacker could lay advance plans for how to track down the survivors and either dominate them or eliminate them using more conventional means. An attacker with the resources and coordination capacity to carry out an attack like this might conceivably be expected to also be able to "finish the job."

So we should consider biorisk, and other deliberate attacks on humanity, in terms of two components:

• Initial attack level success (to varying degrees, such as "causing chaos," "X-disaster," and human extinction)
• Given attack success level, the difficulty of "finishing the job"

The "finishing the job" aspect has both positive and negative aspects. The negative aspect is that our calculations for the X-risk of a bioweapons attack can't stop with, "well, they probably can't infect everybody." The positive aspect is that it provides a whole new opportunity for preventing or mitigating the threat.

I haven't seen EA deal very much with post-X-disaster scenarios. I know that there's some attention from ALLFED on how we might improve the food supply in the case of a nuclear winter. Perhaps there is some thought out there on how to resist a "finishing the job" post-X-disaster scenario.

This post would benefit from an analysis of the relative likelihood of a biorisk and malevolent AI risk.