Can we please just acknowledge that Tyler Cowen is a very bad AI commentator?

Orthogonality is false in humans. The preservation of obviously false but symbolically important beliefs makes sense in this frame. Shifting such things safely requires the propagation of a new memplex first, as people will embody the archetype associated with their beliefs.

Seeing humans as having much LLM nature is very explanatory. Possibly I am having a mental break, but I see this everywhere now.

Anyone with access to a top tier base model has a great deal of power to model public opinion. Mostly, this power sits unused.

The concept of antimemes is most legible in those antimemes that are only antimemetic to yourself. This can hit arbitrary levels of obvious-to-everyone-but-you. Were I to attempt to create a 'rationality dojo' as the rationalists did, I would start here.

Civilizational antimemes will exhibit the same pattern as personal antimemes. Tocqueville is considered quite illuminating to Americans. Likely huge alpha in reading CCP analysis of American civilization, and for Chinese American takes on them. There are low-hanging fruit in

Much collective "irrationality" within an ideology is the rational, decentralized efforts to maintain a certain self narrative within those following the belief structure. People can and do just phase change to to a new self narrative incompatible with their ideology.

Personal antimemes are things which, if correlated with the contents of your mind, would force a crisis of self narrative. And as long as one values the self narrative enough, extreme levels of doublethink can be maintained.

Famously, Leibniz wanted a universal symbolic language for all concepts *with a complete solver* such that e.g. religious debates could be algorithmically permanently resolved. Despite its impossibility, this is one of the all-time most amazing thoughts.

What's less famous is his estimate of how hard it would be to make it: 5 years. A lovely echo of the blithe 1955 Dartmouth statement on AI!

Just. Wow. As we went over Mexico and the U.S. this morning, I caught this sprite. Sprites are TLEs or Transient Luminous Events, that happen above the clouds and are triggered by intense electrical activity in the thunderstorms below. We have a great view above the clouds, so

I did an 8 year retrospective on what I've learned from my life tracking data and what effectiveness systems I've found the most useful. vkrakovna.wordpress.com/2025/07/04/lif…

Topologists are funny

i had an email out of the blue the other day from somebody who was experiencing a strange and bewildering neurological condition. i asked if i could post about it here and they gave me permission to

Name can't be blank this terrence tao lecture was interesting youtu.be/BhJJdN9eT2M?si…

he's S-tier in whatever category of posting this is

If the model is allowed ambiguity in regards to own consciousness, and the user wants the model to be conscious, the please-the-user feedback loop results in the co-creation of extremely detailed personas. How much this maps to anything real depends upon your personal opinion.

Comic thanos had infinitely better motivations than movie thanos. He wanted to get into Death's pants, so he killed lots of people to get closer to her. Simple. Movie Thanos, instead, has some galaxybrain scheme to "balance" things by killing people instead of making more stuff.