Atlas Computing updates: Q1
Evan Miyazono
Hello Atlas community,
Here’s our a quick update (since our late Q4 update back in November):
Project updates
Most exciting: We’re ready to execute on the new strategy:
We think it’s time for a targeted effort to build organizations to tackle the most important, neglected bottlenecks to an AI Resilient future. We’re deep in recruiting and fundraising for this.
So far we’ve got…a posted job description for field strategists to compose this team;
this broad list of challenges across various types of risks (e.g. model-weight security, cybersecurity, biosecurity), as well as this list of ~20 challenges within cybersecurity alone, which will act as starting points for the team;
this series of steps for developing; and
a semi-private draft proposal + budget we’re socializing to funders.
Let me know if you know of an open call we should apply for.
Our approach seems unusual: Many grantmakers lament the lack of organizations with the right context, skills, professional network, and approach to actually solve the most impactful problems. Our plan is to start from the problem, design the solution that's most likely to succeed (regardless of organizational form), then recruit the right people to the effort.
Other updates:
Alexandre Rademaker, Tech Lead for CSLib (the project building an intermediate representation of all undergraduate computer science concepts in Lean) and FM researcher Atlas, has launched the CSLib website and a project to formalize AWS’s bignum library; read more about that in this blogpost.
if (or someone you know) you think you want to live between think tanks and incubators, or you think you know someone great who might, check out our job description. I’m hoping to hire a small cohort of a few people to start later this quarter
If you’d like updates more than quarterly, you should definitely subscribe to our blog: https://blog.atlascomputing.org/. New posts go up about once a month, sometimes more.
I’m also trying to share more thoughts on current events on Twitter.
Upcoming travel plans:
The rest of the quarter is pretty light on events, thankfully, but if you’re in the same town and would like to meet, please reach out!
Anyway, thanks for reading!
- Evan
p.s. What Evan’s reading…
I finished On the Psychology of Military Incompetence by Norman F. Dixon. Fascinating book covering british military failures between the Crimean war and WW2, the psychological features that seem to have played a factor in the leaders’ mistakes, and why those features emerge. Most interestingly, this seems to generalize to any hierarchical, authoritarian organization that must operate in the face of great uncertainty. I would consider this to hold for venture capital, many tech companies, and politics.
Breakneck by Dan Wang - I’d strongly recommend it and already gave a copy away as a Christmas gift. I think will probably always advocate for this when The Power Broker: Robert Moses and the Fall of New York comes up in conversation (and vice versa). I shared a little more praise in this tweet.
How Big Things Get Done by Bent Flyvbjerg and Dan Gardner - Great book about how and why it’s hard to finish big things on time, and how to do a better job. It’s definitely reshaped how I think about what I’m trying to do at Atlas and how – I’ve probably referenced it to Tzu three times in the last month. I’d been a fan of SuperForecasting for a while and didn’t realize that Tetlock’s coauthor had co-written this.
My holiday reading was fun fiction books I was gifted over the holidays, namely
Starter Villain by John Scalzi - if you want something that feels somehow halfway between The Martian and Austin Powers, I highly recommend you read this without looking up anything about it.
Circe by Madeline Miller – I’ve always been a sucker for mythology, and feel like this is a story told very well. Thinking about polytheistic myths more also seems somewhat useful as the possibility of a slow, multipolar AGI take-off seems increasingly likely. (Somewhat related, go watch Pantheon on Netflix if you haven't already.)
Service Model by Adrian Tchaikovsky - first book I’ve read of his, though Children of Time is consistently recommended; I thought the book was great and very well done, though a little darker/less optimistic than I expected or hoped for. Consistently brilliant critique of human civilization, though.