Things I read and liked in January

Recursive self-improvement, reveries of ruin, Russian retail realizations, research principles, rewriting

[1] Patrick Butlin and Theodoros Lappas propose Principles for responsible AI consciousness research. The Guardian covers an associated open letter signed by, among others, Karl Friston and Stephen Fry.

[2] Brian Graham highlights a hilarious bit of Frederick Douglass’s “Composite Nation”, in which Douglass argued for a multi-ethnic America (in 1869!). Douglass defends Chinese immigration, and argues that America has been strengthened by all countries and races—except, of course, the French.

To no class of our population are we more indebted for valuable qualities of head, heart, and hand, than to the German. Say what we will of their lager, their smoke, and their metaphysics, they have brought to us a fresh, vigorous and child-like nature; a boundless facility in the acquisition of knowledge; a subtle and far-reaching intellect, and a fearless love of truth. Though remarkable for patient and laborious thought, the true German is a joyous child of freedom, fond of manly sports, a lover of music, and a happy man generally.¹ Though he never forgets that he is a German, he never fails to remember that he is an American.

A Frenchman comes here to make money, and that is about all that need be said of him. He is only a Frenchman. He neither learns our language nor loves our country. His hand is on our pocket and his eye on Paris. He gets what he wants and, like a sensible Frenchman, returns to France to spend it.

[3] Matt Bateman correctly pushing back against anti-vanilla prejudice. Mitski nailed this as well:

Vanilla is the most perverted flavor of all the ice creams. I feel like it gets a bad rap because we say “vanilla sex”. But actually I’ve realized that after you go through all the flavors, you come back to vanilla, and realize and understand its depth, and just how f***ing nasty it is. Underrated. A good vanilla is like touching God.

[4] Gwern conjectures that the o1 —> o3 models may enable rapid recursive improvement: use more inference to create better training points; those training points can train a stronger base model; that model’s inference then creates even better training points, etc.

[5] Tyler Cowen, who has not been super bullish on AI progress, on using OpenAI’s Deep Research: “I think of the quality as comparable to having a good PhD-level research assistant, and sending that person away with a task for a week or two, or maybe more”. I also have been using Deep Research, and find it unnervingly good.

[6] Writer Michael Bhaskar and DeepMind co-founder Mustafa Suleyman wrote a book about AI called The Coming Wave. It’s not how they intended the metaphor, but recent AI progress does indeed give me this feeling:

And as they stood so, their hands met and clasped, though they did not know it. And still they waited for they knew not what. Then presently it seemed to them that above the ridges of the distant mountains another vast mountain of darkness rose, towering up like a wave that should engulf the world, and about it lightnings flickered; and then a tremor ran through the earth, and they felt the walls of the City quiver. A sound like a sigh went up from all the lands about them; and their hearts beat suddenly again.

'It reminds me of Númenor,' said Faramir, and wondered to hear himself speak.

'Of Númenor?' said Éowyn.

'Yes,' said Faramir, 'of the land of Westernesse that foundered, and of the great dark wave climbing over the green lands and above the hills, and coming on, darkness unescapable. I often dream of it.'

[7] In letters, Tolkien explains that he crafted Faramir’s dream from his own “Atlantis-haunting”, a recurring nightmare: “he speaks for me. That vision and dream has been ever with me”. (The movie gives the dream to Eowyn.)

[8] Sadly, extreme pain can overwhelm basically any defense against it: Here is a Zen teacher on how his awakening experiences were not enough:

When I was experiencing that, like I said, my realization didn’t save me from it….all of that kind of broke down, so it’s like all of that abandoned me.

…I did feel totally abandoned. And when you’re experiencing that, you know that somebody else who hasn’t experienced it—no matter how much and how compassionate and how loving they are—you know that you’re in a profound state of aloneness. I felt profoundly alone.

[9] And here is cognitive scientist Ruben Laukkonen: “When the ambulance arrives, I can’t even tell them what’s happening. My gf’s in a panic. They’re confused, and I’m in some kind of dark void, where anywhere my consciousness dares to land it finds only unbearable pain. Pain looped in with anticipatory fear of the next cough. You know, I’ve been meditating daily for 12 years. Dozens of retreats. And I was just totally humbled by this experience. There are states where there’s nowhere to rest.”

[10] Related: the admonition in “Letter from Utopia” that “You cannot get here by any magic trick or hokum, or by the power of wishful thinking, or by semantic acrobatics, meditation, affirmation, incantation, or by an act of parliament.”

[11] Related: hell must be destroyed.

[12] How to strength train for picking up your kids. Cute, practical.

[13] From Experience Machines’s Texas correspondent Charlie Schaub: Boris Yeltsin in Houston:

“Between trying free samples of cheese and produce and staring at the meat selections, Yeltsin stopped people pushing carts to ask about the food they had selected and its cost," Houston Chronicle reporter Stefanie Asin wrote at the time.

“On his way out of the store, Yeltsin asked for the manager. He wanted to know if special education is necessary to be a supermarket manager." Yeltsin was most struck by the store's stark comparison to a typical Russian grocery store at the time. "Even the Politburo doesn't have this choice. Not even Mr. Gorbachev,'' Yeltsin said as he observed the store's fish department.

[14] Celine Nguyen explains the practice of “deep copying” of writing in order to understand what works well and why; she then does some, to understand what makes for a good beginning to an essay.

[15] But also, per links 4-6, superhuman writing sure seems like it’s coming / is partially here? Which kinds of human writing will people find valuable as we keep moving into AI dominance?

1. The stereotype of Germans as happy is interesting. Also, I don’t know if any of you are history buffs, but “joyous child of freedom” got dented a bit in the 20th century