Our Active thesis - WHAT CAN WE OFFER THAT NO ONE CAN?
- What is our comparative advantage as a group?
- So far, we are a very big collection of practicing independent scientists with SKILL, but are searching for problems.
- We have some edge in natural sciences, network theory, formalisation of physics - I would probably call our topic space 'Independent applied engineering'
- We have growing connections with metascience funders.
- Do we need specific subject matter?
- In analogy with the AI community, it feels like we do - they have really done well, but I don't know what in the hard sciences we can do since we don't have experimetns
- What are we not?
- Probably not like these private labs, which are vulnerable to being closed source
- We are not that focused, and we shouldn't probably be diving too deep into 'specific areas' - more meta style infra support
- We are not explicitly EA / rationalist aligned
- Do we need more independent scientists in the world?
- What edge do independent scientists have over regular scientists in academia?
- This seems to be skills related to industry
- No financial constraints / publish or perish
- Better expertise in coding and tooling
- Better expertise in project delivery
- What would successful infrastructure look like?
- Conferences - this is the main place where ideas are generated
- Would a publication be at all useful?
- if so - what would it look like?
- A publication with 1000s of readers, filled with submissions from independent reseachers
- But then
- Is upskilling something that we should focus on?
- A platform for upskilling - I would need someone to compile resources / build tech to get a collection of problems from textbooks / modify them. But the thing is, there are so many places to upskill online (everyone is using chatGPT) and so it's hard to imagine we could make a better one unless we had a REALLY significant targeted project.
- I think people use AI + textbooks to learn - so
- Our only possible edge that I can see is physLean, but this is a WHOLE thing
- In this case - we should probably be doing as much as possible to support AI for Math, PhysLean, and other conferecnes
- A wiki where we can share resources