AI is great. it is the most transformative technology we will ever build.
AI will accelerate everything: science, code, thinking, everything downstream. this is both good and bad.
people will use it to produce things much faster than ever: papers, code, analysis. humanity cannot keep up with that pace.
if we cannot make sense of what is being produced, we cannot figure out what is true. and if we cannot figure out what is true, we cannot decide things well: which drug is safe, which policy works, what to believe.
our epistemic infrastructure, the tools and institutions we use to figure out what is true, was not built for this. peer review, policy debate, code review, community notes: they all assume things arrive at human speed and can be checked by a small number of humans. both assumptions are now breaking.
everyone will talk to reality through AI. it will be everywhere.
we need to defensively and differentially accelerate our epistemic capabilities, before AI gets far ahead and we are stuck with the old systems when we can least afford to be. that is a global catastrophe waiting to happen.
AI also gives us an opportunity to rethink and expand our design spaces of the epistemic infra. we can use AI for epistemics. we can build new infra with AI, that keeps humans in the loop but uses AI to make it better.
we are building this. we start with science and cybersecurity.