All posts
Why biological weapons are scary, and what we can do about it
Comparing conventional, chemical, nuclear, and biological weapons on cost per death and mass per death puts biological weapons in a class of their own.
The AI Industrial Explosion — Part 5: Given AGI, automating physical production is probably not that hard
Once we have the minds, building bodies capable enough to automate physical labor is not that hard.
The easiest pathway to control is through executive power
Power is already centralised in the US President and Chinese General Secretary. Whoever seeks control during an AI transition, the easiest route runs through the executive.
Freeing Thucydides
Wars happen because states cannot credibly commit to the agreements they would both prefer. AIs could supply the missing enforcer, ending competition for control of the future.
Linkpost: How Fast is Post-AGI Growth?
(Don't fear) the strangelet
Nuclear physics allowed the creation of the atomic and hydrogen bombs. Does new physics provide even more destructive possibilities?
Destroying the universe: How hard can it be?
The universe is probably not stable, but intentionally triggering its decay might still be impossible.
The AI Industrial Explosion — Part 4: Cheap power
What's the cheapest energy system we could plausibly build, and could it keep pace with explosive post-AGI growth?
The AI Industrial Explosion — Part 3: Going faster
How much faster could an AI-automated economy grow without inventing new technology? Reoptimizing across today's frontier plants and considering historical labor-intensive methods cuts the doubling time roughly in half, to four-to-eight months
How I think about catastrophic biological risk (part II): risk breakdown by type of prevention
How I think about catastrophic biological risk (part I): risk breakdown by type of response
This series of posts outlines how I think about the most extreme types of risks. My goal here is to share my worldview in a straightforward and compressed form rather than trying to persuade a skeptical audience, although I do share some of my reasoning.
The AI Industrial Explosion — Part 2: Transition Dynamics
How fast could an AI-automated economy actually start growing? Today's economy doesn't produce enough of the stuff that makes stuff. Restructuring takes a few years — but then the second doubling comes in half the time, and the economy is many times its current size within a decade.
10 big projects for reducing bio x-risk
The field of people working on reducing bio x-risk is distressingly small. I sketch out 10 big, urgent projects I'd be excited for new people to come work on and own.
Prioritizing Environment-to-Human Biological Threats
Pathogens that replicate in the environment and transmit to humans pose a uniquely direct existential risk, far more so than those that spread person-to-person or can't grow outside a host. Of the possible exposure routes, airborne transmission is by far the hardest to defend against.
It May Be Possible to Improvise A High Grade Bioshelter
Surviving an environment-to-human pathogen would require widespread protection from airborne exposure, indoors and out. We think this may be achievable using improvised bioshelters and PPE made from household materials, though this hypothesis still needs more testing.
Considerations for PPE Strategy
A misaligned AI or human-AI group could attempt takeover by releasing a highly transmissible engineered pathogen. I discuss what a PPE strategy aimed at this threat model needs to get right.
The AI Industrial Explosion — Part 1: Maximum growth rates with current production methods
How fast can an AI-driven economy grow? Economists expect a few percentage points; at best those closer to AI development imagine Dyson spheres within years. Who is correct?
Linkpost: Day in the Life: Global Catastrophic Risks Grantmakers
How I currently use AIs for research
I thought I'd document how I'm currently using AI tools, both because it's changed a lot in the past month and also in the hope that someone can look at this, be horrified, and tell me all the things I'm doing that could be better.
Will whole brain emulation matter for the AI transition?
Whole brain emulation gets suggested as a potential AI-safety lever. Here we investigate two questions: when could we get it relative to superintelligence, and would it matter if we did?
On far-UVC and air filtration
Discussions of built-environment pandemic defense tend to treat far-UVC as the standout technology and air filtration as an afterthought. I think this gets the ranking wrong.
Could humanity survive a total loss of agriculture?
I evaluate whether non-agricultural food technologies could be scaled up fast enough to feed humanity if there was an extreme attack on the crop production.
Linkpost: The Low-tech Plan to Patch Humanity's Greatest Weakness
The Four Pillars: A Hypothesis for Countering Catastrophic Biological Risk
Biological risks are more severe than has been widely appreciated. Here we outline a hypothesis for ‘four pillars’ of biodefense that should work against even the most sophisticated engineered pathogens or ‘unknown unknowns’.