Links Post - September
SSC already linked this article about knitting, so I’m going to assume you read it. It makes me sad I didn’t signal to boost it before it was cool, probably my favorite read for August.
OpenAI’s latest update around their approach to alignment, too high level for my taste, but citations are present and it’s much more common sensical, and less doomygloomy than I’d have expected.
This paper does a great job at inadvertently explaining why I think “antioxidants” are BS and why you should be very weary of anyone paddling a compound as such… by, explaining how this property is actually assesed.
Tebipenem seems like an amazing not-talked-about antibiotic and will become my go-to all-purpose antibiotic to take with me on my travels… once I figure out how to source a reasonable quantity for < 4,000$
I don’t think I’ve shelled out Christoph Molnar’s Interpretable Machine Learning (free) book. I should have, I recently remembered about it and re-read a few chapters with some work colleagues, it’s still holding up great.
Reverse engineering the NTK: towards first-principles architecture design — I’m not 100% sure I understand all the implications this has, and I don’t understand the limitations, so I’m very hedgy saying this, but, it seems like it could lead to a monumental paradigm shift.
DMT and Hyperbolic Geometry — If you occasionally read/watch QRI content and missed this, read it. If you don’t, it might be a good introduction if you’ve got time to follow the links.
Nintil’s take on AI Risk, very long but worth the read, made some things click for me.
VGRs take on AI Risk, very long but worth the read, made some things click for me.
The real covid death numbers can finally be estimated from excess mortality and happily enough somebody did it. Not sure it’s the best such analysis, but it’s the first one I stumbled upon, so it’s the one I’m linking.
I am usually unimpressed by VGRs musing on corpo-cultura-socio-web3-blah but I enjoyed this one.
Robbin Hanson write a lot of stuff that I’m “meh” about, in the last 3 months there are two articles that broke the mold, they were great.
Matt Lakeman went to Ukraine in the middle of a warzone to… you know, have a look around. Once again making me ashamed to call myself a real traveler or start a travel blog.
Rohit invoking VGR in: What we want is a hunter-gatherer lifestyle with space-age tools
I think nutritional epidemiology is mainly fictive. I previously mentioned a book describing this fantasy world, one I had never read, to people questioning if they should care about salt. I now feel bad about it and wil start reading any such thing before I mention it to others, lore nerds have essentially debunked it, it's not cannon and should be trusted.