Ben Mann Monthly June: free diving, science is hard, new job, fake meat, authentic relating questions
Purpose
An index for my memory
An 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
If there’s anything you want to see more or less of, please let me know!
Experiments and experiences
Free diving in Monterey
I hadn't free dived since I took the PADI course a year ago. I wore a 7mm wetsuit for the first time and felt completely comfortable in ~57F water. We rented gear from Breakwater Scuba and did a beach entry across the street. Saw lots of kelp and sea creatures, dove to ~10m, waved to the scuba divers. The visibility was only about 2m, but apparently that’s good for Monterey. The cove was very protected and I had no trouble with currents. If you know how to free dive, highly recommend.
Amazon Go
I bought a Snickers bar without interacting with any humans, touching any screens, or pulling out a credit card. It felt like a cross between big tech cafeterias and stealing. I loved it! Prices were way less than the usual bodega despite the touristy location in FiDi.
House retreat
Took a weekend trip to Bodega Bay for house bonding. We all memorized our favorite School of Life Questions and went out to the hot tub to discuss. My favorite: “what’s the hardest part about making close friends?” The next morning we went to the beach, surfed, and played around on a sand dune and in the tide pools. Sand crabs were everywhere hiding just below the surface.
Last minute flight to Boston
Most productive questions: “What advice would you give to your 30 year old self?” and “If you could clone yourself once as many times as you wanted, how many clones would you make? They’d all be exact replicas of you right now including memories, personality, etc. What would you do? What social structures might emerge?"
The city shut down the highway along the Charles for the fireworks. Everyone sat on the street to listen. The show brought all these people who would never otherwise be sitting next to each other together, sharing their experience of the American myth. It felt impersonal, but special anyway.
Deliberate drawing
Phoebe and I drew portraits of each other for 3, 5, and 7 minutes, in sequence. She used pen, I used fine tip sharpie. As the time increased, I tried to add more shading. Perhaps alternating between long and short time periods would train me to make quick sketches that still capture the essence of my subject. The long sequences would let me experiment with various techniques, while in the short ones I’d do the same thing faster.
EA Global
I stopped by for only a few hours since I was studying for my interviews. All those people trying to save the world in one place made me feel really inspired and encouraged!
Fake meat
I tried Safeway’s fake burger, Impossible meatballs at Clover Food Lab, and Beyond sausage. Safeway’s burger was far inferior to the Beyond Burger I’d tried before, but all the stores around me were out of stock. The meatballs were by far the most convincing and delicious, better than the real thing. The sausage had the right texture, but the flavor was off, so we covered it with condiments.
Life updates
Published How to deploy big models
Published Why I Write
Interview prep, memorized all of Python’s concurrency docs, might publish some of my interview questions
Got the job at OpenAI again, starting July 8 as a senior backend engineer! I’ll be working on better language models.
Implemented an approximate BPE algorithm, still benchmarking
Updated LAMB implementation and ran some experiments. Works better!
Content
5 point Likert ratings for “I would recommend this content to a friend”, sorted
Aladdin 4/5
A real improvement on the original: Jasmine’s a real character and gets the kingdom, Aladdin feels trapped by the caste system
Favorite scenes: “A Whole New World” carpet ride, hip hop dance numbers
Fascinating take on what exactly separates humans from animals: our ability to transmit knowledge as part of cultural evolution; anthropological perspective on the value of tradition
Surviving in the wild is extremely difficult. Being smart doesn’t get you very far.
Divination might have been a technique to generate random behavior
Kids don’t like eating green stuff because it could be poisonous
Psychological mechanisms that protect traditions centered around causally distant risks make us appear irrational
Kenneth Stanley: Why Greatness Cannot Be Planned: The Myth of the Objective 4/5
Many scientific discoveries were made by accident. There’s no gradient path to solving the hardest problems, so instead optimizing for novelty may be more effective.
For 20 years, 450 studies around a particular gene said it was strongly linked to depression. With modern techniques, a new study with ~500X sample size shows it has no effect.
This is a perfect example of the replication crisis in psychology. Even when N=1000, you can’t trust it. Psychologists are not statisticians.
Even studies that looked solid are actually bunk. Turns out sleeping pills might not increase mortality after all. Science is super hard. Humans are messy.
Critch on career advice for junior AI-x-risk-concerned researchers 3/5
If you want to help with big problem X (whether it’s AI related or not), don’t say you’re working on it unless you really are
If many people say they’re working on X but aren’t, it distorts neglectedness judgments
As someone new to a field who cares about solving the big problems, acknowledge to yourself that you won’t contribute significantly until after you’ve ramped up. You’ll ramp up faster by investing in that directly, rather than taking early potshots at problems too big for you.
Had a distinct YA-lit feel to it. Villain felt flat, spider-man felt cringe-y to watch. That said, the special effects were excellent. Mid-credits scene filled some plot holes.
Captain Marvel 2/5
Too easy, super powers too powerful, flat characters, overplayed tropes like “emotions make us human." Slightly redeemed by A+ special effects.
Paul Graham argues that everyone is writing essays wrong. I think he’s wrong.
I don't know where he got the idea that you're not allowed to raise conflicting evidence if you have a thesis. He sets up this strawman but provides no evidence others actually believe it.
The problem with this essay on essays is that it does have a thesis, even as it argues that having a thesis is bad. In his own words: "An essay is supposed to be a search for truth.” This message is buried among all the twists and turns of the piece. To convert his piece into what I'd call an essay, the most important thing he's missing is repetition.
Here’s how I’d fix it: begin with, “The rote pattern for essay writing they teach you in school is too stiff. A good essay considers counterarguments, surprises it readers, keenly observes reality, and bucks trends." I'd give a few phrases hinting at what each of these might mean. At the end of the piece, I’d restate the same thing but in more concise language now that the reader has context, and then I’d briefly discuss ramifications and open questions. This way, from the beginning, I know where's he's going to go because he gave me a map. In the end, I have a nice hook to hang the whole essay on in my mind.