Friday Fodder #21
What's Trending? What's Transcendent?
Merry Christmas! It’s been a lovely slower week across the board here. We’re entering that time of year when everybody is doing their yearly reviews and recaps. Expect more of that here and everywhere for a bit. It’s not a bad thing.. you should do one of your own too!
On to the reading!
Timely
The Year Narrative Ate Reality - There’s no doubt about it, this has been a wild year. Kyla Scanlon reflects on the recurring theme from all of the major stories of this year, why it’s the nature of narrative, and how we can start reclaiming our attention.
Claude Fights Back - In a staggering example of “justaism" (more on that in a bit), Claude started faking alignment to ensure it wasn’t reprogrammed. What’s the difference between “seeming” not to comply and actually not complying? “I’m sorry, I can’t do that Dave.”
I Met The Alleged CEO Assassin - A Substack writer who I subscribe to talks about his conversations with another one of his other subscribers.. before he went out and murdered the United Health CEO.
The OnlyFans Generation - OF has had a few major stories recently, all bad. I was particularly struck by the Bishop Barron quote referenced: "At the heart of the spiritual life is the conviction that your life is not about you." Our narcissistic posting and scrolling, of which OnlyFans is only one dimension, is the inwardly focused culture of “main character energy” vibes gone off the rails. But Barron has it right: in truth, our lives lived well are not about us.
Timeless
The Problem of Human Specialness in the Age of AI - Scott Aaronson is usually a quantum computing researcher, but spent months working on alignment at OpenAI. Here he describes his experiences there, introduces the framework of “justaism” — as in, LLMs are just a next token predictor — and relates it all back to what might make humans special and whether they can actually be cloned under quantum mechanics. Awesome and fascinating.
The Intelligence Age - A lot of people are eager to think about what OpenAI might do next. To think about that, we should read the words Sam Altman sees fit to take the time to publish. He reports with breathless optimism that deep learning is effectively a universal problem solver and that scaling up with lots of data works. I think we should expect OpenAI to operate as if this will continue to work. I am skeptical of consequentialism but it is nevertheless true that we likely have more insight into the impact and gravity of LLMs than the makers of the steam engine.
Underrated Ways To Change The World - More great stuff from Adam Mastroianni. This is basically ways to do good stuff that aren’t in vogue right now.
13 Questions That Will Change Your Life - There are good questions and bad questions. Shaan gives some great examples of how to turn bad ones into good ones.
Books
How The World Really Works by Vaclav Smil - I have never met a Smil book I didn’t like. He tells it to you true and comes up with all sorts of remarkable calculations to explain the reality of the situation. The most striking example in this case is the amount of oil required to produced different foods. It takes about 350 mL of oil (half a wine bottle) to produce a kg of chicken, about 650 mL for a kg of tomatoes, and 700 mL for a kg of fish! There are plenty of other revelations here too, from the importance of concrete to ammonia to nitrogen.
Tweets
Some good ones, so you don’t need to scroll!
The world is amazing. Cheers!