Start of Inkhaven!
https://lydia.ml/ is now live!
Hi! Inkhaven starts today, & I figured I should pull together a quick post explaining what that means for the blog, plus setting intentions for the month.
All posts will be crossposted to lydia.ml.
Embrace conversational writing!
When I was scoping out ‘GiveWell of AI Safety’ for Manifund earlier this year, one of the first things I did was write a blogpost.
The first draft took me ~2 days. I showed it to Austin, whose feedback was “minor edits, but otherwise good to go!”
I wasn’t satisfied and went back to the drawing board. One week later, I had a significantly more polished draft. I showed this to Austin and his feedback was “hmm…I think I liked the first version better!”
Now, when I reread, I too prefer the first version. Readers do get a lot of value out of half-formed notes—see e.g. representative Guzey drop.
Since then, I’ve been trying to publish my initial forward-pass more often.
I want to do lots of technical writing…
I read lots of ML papers, and I want to write about these—ideally including notes from follow-up experiments. I get lots of value from reading others’ paper summaries, so I’m happy to put more out into the world.
These will live under the AI/ML tab (I considered setting up another publication, but I think tags work for now. The content is what matters).
Also, since Beware the Delmore Effect, I’ve become more…open-minded. This means there might be indulgent detours into other technical topics as well:
How much compute would it take to simulate humans at varying levels of granularity? (inspired by Type-0 risks in VWH)
What actually is going on in neuromodulation? What’s the state of injecting messages directly into the brain?
What would putting longevity on the Gap Map look like?
The future of human labor: how I changed my mind about humanity’s ‘skill issue’ (ft. Kulveit et al., Korinek and Suh)
…the Delmore Effect strikes again.
I’ll go to Constellation (& Boston, Toronto)
I plan to spend 17-21 Nov at Constellation. I’ll offer to scribe people’s thoughts that haven’t made their way into the world yet. If you know anyone there who might appreciate or benefit from this, please let me know!
I’ll spend 6-10 Nov in the Boston area and 10-11 Nov in Toronto. I may yet be persuaded to visit more of Canada (discovering Canadian AI research (Mila, Vector Institute, UAlberta(!)) has been exciting recently).
On slow days, there’ll be more personal posts
Here are some in the works:
Real life doesn’t come with dependency conflict warnings
Two types of homeostasis
Do you love your field or yourself?
Reducing your susceptibility to infohazards
Three weird, replicable social moves
Tier list of things postponed to post-singularity
Why people resist specializing
The Nonperishable Agenda
You can email me at hello@lydia.ml to vote for any of these :) I have also set up a market where you can vote on which is most likely to happen (and, by doing so, increase the likelihood that it does).
(Betting ‘yes’ on all, then holding me to it, is probably easy money!)
…the Delmore Effect strikes again.
See you more often :)