youzicha

In 1925, Einstein went on a walk with a young student named Esther Salaman. As they wandered, he shared his core guiding intellectual principle: "I want to know how God created this world. I'm not interested in this or that phenomenon, in the spectrum of this or that element. I want to know His thoughts; the rest are just details."

that's what I remember from when I tried to find a source for this quote. It's not something Einstein wrote, it's someone's recollection that he said soemthing more or less like this

Anyway, I'm interested in the spectrum of this or that element. That's always been metaphorically true of me, and for the last few years since I switched to quantum chemistry, it's literally true... well, not elements, but it's literally true that I'm interested in the spectrum of this or that molecule. Remembered this quote this morning while reading a paper in the cafe, after writing "50 nm shift between isomers!" on a graph of spectra.

"He made pantries Of Vaux and Arden And the village gentry's Kitchen-garden. Fruits within yards Were his staples; He drank whole vineyards From Rome to Naples, Then went to Brittany For the cider. He could sit any Horse, a rider Outstripping Cheiron's Canter and gallop. Pau's environs The pubs of Salop, Wells and Bath inns Shared his pleasure With taverns of Athens..."

And there's an even more technically brilliant variation of this, which is exemplified in McGinley's "Lesson for Beginners." I'm going to transcribe the whole poem, because (though it's a bit glib), I have rarely found its equal for marrying verve and musicality to keeping on topic and not wandering into rhyme for rhyme's sake.

LESSON FOR BEGINNERS by Phyllis McGinley

Martin of Tours, When he earned his shilling Trooping the flags Of the Roman Guard, Came on a poor, Aching and chilling Beggar in rags By the barracks yard. Blind to his lack, The Guard went riding. But Martin a moment Paused and drew The coat from his back, His sword from hiding, And sabered his raiment Into two.

these responses from Yudkowsky in this argument from 2021 are a fun read, especially the idea that once "self-improvement" is reached the end of the world happens shortly afterwards (days, hours).

have coding agents reached the point of self-improvement? because people are using LLMs to optimise other LLMs, like the nanoGPT challenge, and the results are decent but not earth shattering: they quickly reach the limits of the improvements they can make.

to make further improvements they would need fundamentally better algorithms, better chips, better fabs, and it's not clear if those improvements would automatically catalyse further improvement.

basically this is software engineers being paranoid about how they'll destroy the world by creating a computer-based world zipbomb without thinking about the logistical limitations that we're not really talking about turing machines and theoretical computers so much as we're talking about highly abstracted electrical circuits

even if you do reach the point of self-improvement and it's exponential and it gives us 10 days or whatever, the ability to expand infinitely is limited by the resource capacity of whatever machine it's plugged in; you need to be able to produce More Computers because they physically degrade, and even then there's benefits to all kinds of automation and specialization

Grok: twincest

Me: what

Grok: they’re Tories because they’re overcompensating for twincest

Me: hmm maybe