Discover more from Unruly Sun
1750 hours a year
I think the most interesting thing that happened to me this week my first meeting with the piano tutor. I’d been playing since the start of the pandemic but wanted for a while to get more focused. He taught showed me Bill Evans-esque rootless chord voicings of Autumn Leaves.
Questions about Governance. In which I summarize 6000 years of governing human civilization in two pages. Thoughts?
(Work in progress). Why open source machine learning is important. Comments highly welcome. Are there any lenses or questions that I’m missing?
Max Planck observed that new scientific ideas only spread as its opponents died. Longevity research promises to extend the lifespan of everyone — could that hold back scientific progress? When intense emotional interactions happen, it's often very helpful to take the time to debrief - holding the space to safely acknowledge what happened. It's taken me many years to realize that I should use three passes to read nonfiction. First to create the structure, context, and broad outline of information; second to digest the substance; third to recreate the internal logic.
Laws of Tech: Commoditize Your Complement, Gwern. I’m coming to realize that you can’t really comprehend the open source space without understanding business strategy. This piece by Gwern is a very helpful synthesis of one of the key principles within tech strategy. It explains, for instance, why IBM spends millions each year on open source software (hint: it’s not because they read the cyberpunk manifesto, nor is it for ‘good press’).
Moving beyond token governance, Vitalik Buterin. Focused on the crypto space, but applies to digital governance more generally. Vitalik basically explains why giving each member of a group a token to trade and vote with is a dangerous way to govern. He provides good examples of why formal decision theoretic frameworks fail in real life in group settings. The more automated a voting system, the more values have to be explicitly specified. Yet at a very high level, our values are not just high dimensional, but nebulous/conceptually fuzzy enough that any attempt to formalize them fails shortly. See Goodhart’s Law, the principal-agent problem etc.
Benford’s Law. Leading digits do not occur uniformly in real life (‘1’ is far more common than any other digit). Political scientists used this to detect faked votes in the Iranian election.
Island, Aldous Huxley. A countermovement to the dystopia of Brave New World. In his foreword to the last edition of BNW, he writes, “If I were now to rewrite the book, I would offer the Savage a third alternative. Between the Utopian and primitive horns of his dilemma would lie the possibility of sanity... In this community economics would be decentralist and Henry-Georgian, politics Kropotkinesque and co-operative.”
Understanding Institutional Diversity, Elinor Ostrom. This book is where she really elaborates on her Institutional Analysis and Development framework. As an Ostrom stan I need to have read this already. Read it before I fell asleep on Thursday.
Get statistics from awesome-ml + LFAI repos. Do analysis. (not done - got the statistics, but am only beginning analysis)
Governance essay (done)
Construct dependency graph using statistics
Collect literature review into spreadsheet. (done - you can see my literature review sheet here)
Talked to Ostrom-workshop researcher about using action arenas as a lens of analysis (done - very fun conversation)
Next Two Weeks
Quantitative analysis - what are the most popular ML repos in my sample? what percentage of them are active? how many contributors do they have? how many projects depend on them?
Write essay on said contribution
Get in contact with people who know what they’re doing (G, S, T, H,…)
Finish notes on Understanding Institutional Diversity.
Create a timeline
Check in with IRB