Ben Mann Monthly Dec 2021
Reflection, Canada, hiking, donation, cold takes, Omicron
An index for my memory
A menu of topics for my next conversation with you
A faster way to share what I’m excited about without the barrier of writing a complete blog entry on it
A skimmable way to spread content I found valuable
Experiments and experiences
Andy led a discussion of donating to charity. I learned some important tips. If you donate less than $6k, you can't deduct it, so you're better off batching your donations every other year. If you donate stock, you don't pay long term capital gains tax AND you can claim the full value as a deduction, so compared to selling the stock to have as cash, you get a big bonus. For the curious, here's a great post on how it works. Soon after the discussion, I saw Todor's Facebook post that he'd donated stock to charity. I decided to jump on the bandwagon and try to inspire more folks to do the same. A nice rule of thumb for choosing how much to donate: 10% of your income. This number is stolen from the ancient religious tithing practices. To me it feels like enough to be meaningful, but not enough to impact quality of life. That said, $110k (100 shares of TSLA) is not 10% of my income; instead it's what I would've gotten in cash had I sold the stock I donated, plus some extra for not donating much last year. I tried to not overthink where to send the money. After a few weeks of waffling, I chose GiveWell's Maximum Impact Fund. My runner up was EA's donor lottery, but the amount I donated didn't nicely fit into any of the lots that were available and I was afraid of winning. Some people commented that they'd donated or publicized their donation because they saw my post on Facebook! I'd love to normalize this kind of thing not as showing off, but instead celebrating doing good and sharing our methods.
Driving to Canada
Diana and I road tripped up to Canada from San Francisco. Along the way, we met up with old friends. We went hiking in the rain in Portland, ate Thai food and hung out fireside in Seattle. On the way into Canada, we had to time our COVID tests perfectly since the border required it to be within 72 hours of crossing, but Kaiser only promised to return the test within 72 hours. Luckily it came through at midnight on the second day, so we got through on day 3 with 2 hours to spare. We were both selected for "random testing" and had to swab at a pharmacy in Canada. For contrast, on the way back to America, we didn't have to fill out an application or contact tracing form, didn't have to show our vaccination history or PCR test, or really do anything. We just showed our passports and drove through!
3 weeks in British Columbia
Diana's parents were excellent hosts. They constantly stuffed us with delicious fruits and Korean food. They told stories from Diana's childhood, their past across continents, compared our cultures, and more. On weekends Diana, Roong-ji and I hiked in moist mossy forests. On Christmas the sky dumped snow, so we shod ourselves with crampons and trekked through the snow with our intrepid 8%-wolf dog. At first I was worried her bare paws would freeze on the snow and ice, but she seemed more at home than balmy San Francisco. We walked in all conditions in lieu of all other forms of exercise until I noticed all the holiday food starting to catch up to my waistline. Thirty situps and pushups every morning quickly dispelled my excess flesh. The highlight: a 12-mile hike to a frozen waterfall in a meringue wonderland of snow-blanketed trees and bridges. Afterwards, we sipped london fog and hot chocolate and nibbled a chocolate chip cookie.
This year, I switched from my usual 8760 review to YearCompass on Stacey's suggestion. It was much less work, but also less useful. It seemed too focused on the positive side, but trying to take a "yes and" approach to planning felt nice while doing it. I might do some bonus reflection. My main conclusion was that last year was particularly great. On the personal front, I got engaged and got a dog. At work, I joined a brand new co as the first employee and published a paper. In health, I finished my first half-Ironman, ran my longest running race (21 miles), climbed a V5, "graduated" from therapy, and avoided COVID. Socially, I'm pretty happy with my newsletters, but I miss the depth of thought cultivated by the medium posts I used to write. I spent too long reading the news and social media. I'll write less about content I didn't like. Focus areas for next year:
work: hire enough people to wear only 1-2 hats, publish a major alignment result
health: find a running/yoga routine to improve cardio and flexibility
📜 Anthropic published paper #2
🎂 Dad's 70th birthday
📚 Almost done Perhaps the Stars
🦠 COVID Omicron variant peaking
5 point Likert ratings for “I would recommend this content to a friend”, sorted. Anything < 4
Cogent arguments for why this may be the most important century.
Exploration of Steven Pinker arguments that now is the best time ever. Probably it is, but that doesn't mean there aren't areas we're doing poorly at, like community and substance abuse.
Trustworthy, practical updates on the current state of the pandemic and how to behave accordingly.
Seems like in Feb/March, once hospital load is down and if you're boosted, it's probably better to just get infected and move on with your life.
Huge paper from DeepMind incorporating approximate nearest neighbors search into language models. Unlike REALM, uses a static retriever, so engineering is easier. Gets extremely good performance on a wide range of benchmarks.
I'm particularly excited about this to improve factual accuracy, citing sources, and incorporating new or domain-specific knowledge without additional training.
The real story is even more unbelievable than the film, where the lead dog Togo breaks the rope trying to pull a broken ice flow to shore. He jumps back in the water, grabs the rope, ties it around himself, swims to shore, and saves the whole team. 🤯
He had a laser printer in the Arctic Circle?!
Don't Look Up 3/5
"People need to be able to say things to each other!" Yes.
The Rescue 3/5
Insanely huge international undertaking to rescue a couple of kids. Didn't seem effective use of resources, but impressive technical achievement.
300's production style, for cro-magnons.