As of this week, my Substack is six months old! š„³
This was and still is entirely an exercise in āpublic learningā as a personal project, without me ever expecting (or looking) to get followers ā but thanks to all of you who do follow this newsletter, and especially to Stephen D. Turner who says nice things about me in his own excellent Substack!
It is also time to take a two-week break; there will be no āFive Thingsā newsletter until after mid-April, at the earliest. But, given that this project was only ever a personal learning exercise to help myself get up to speed on the fast-moving world of AI safety and biosecurity, I think I might also be at the point where there is little (personal) value added from just keeping up with news stories and journal preprints. I want to make sure to also focus on better analysis, broader contexts, and deeper understanding of these issues. Unfortunately I havenāt had the time to do the kind of longform writing that I was hoping to do here, and so over the next few months Iād instead like to publish some of the many articles I have in my drafts.
Hereās a sampling of the articles I want to write which Iām publicizing to (1) help keep myself accountable and (2) offer anyone interested in giving draft feedback to get in touch:
Are LLMs good at biology? (Benchmarks vs. wet lab uplift studies)
Can we get a METR graph for bio?
The structure of scientific evolutions (the boring, non-revolutionary work of professional scientists)
What is biosecurity, and why Iām a bit more worried about it (in three parts)
Maybe AI x Biosecurity is actually a bad idea
How to invent bespoke biological superweapons (just kidding)
Tacit knowledge is overrated
The Visions of Vanneavar Bush
Complexity science by many other names
Bimodal solutions for science, public health, and democracy
Who funds American science and why it matters
Can we get plants to fertilize themselves?
Counterfactuals, causality, and other hard problems
Just look at the thing (microscopy for molecular dynamics)