Newsletter

Hello,

Here’s everything since my last little missive to you:

New Articles:

Memory and imagination both use the same architecture

Excerpt: Memory, like many things in the brain, is a bit of a mysterious function, but it’s also one of the first cognitive functions people think might be worth improving. However, the way we typically think about memory makes that quite difficult. Memory seems like it can be broken into some number of different kinds, but this ‘multi-storage’ model misses important things. Instead, the architecture of the brain gives us a clue as to the way memory works that lets us get a handle on it.

Main idea: Memory takes the form of neural maps in the brain, tying our experiences and perceptions together. These maps are the same maps we use to process the world, and imagine the future. Mapping memories to old memories is the way to think about it, not storing memories in a bank.

New Marginalia: My links and notes on interesting content from around the web:

On hypersonic arms, and hypersonic arms control. Link

–

Some new, some old, observations about the mismanagement of depression. Good just for the highlights. Full paper here:

Depression is neither disease nor disorder rather an adaptation that evolved to serve a purpose

Depression is so much more prevalent than currently recognized that it is “species typical”

Antidepressants drive neurotransmitter levels so high that homeostatic regulation kicks in

Antidepressants may suppress symptoms in a manner that increases risk for subsequent relapse

Cognitive therapy works by making rumination more efficient and “unsticking” self-blame

Adding antidepressants may interfere with any enduring effect that cognitive therapy may have Link

–

How to extract insights with seemingly limited resources. On reshaping data for better understanding Link

–

Are men still more influential? I mean obviously, and unfortunately, yes. But encouraging changes, most striking when woman are a 2:1 majority. Link

–

LLM persuasiveness is capped. I told you.

model persuasiveness is characterized by sharply diminishing returns, such that current frontier models are barely more persuasive than models smaller in size by an order of magnitude or more. Second, mere task completion (coherence, staying on topic) appears to account for larger models’ persuasive advantage. These findings suggest that further scaling model size will not much increase the persuasiveness of static LLM-generated messages. Link

–

I hope you found something interesting.

You can find links to all my previous missives here.

Warm regards,

Dorian | btrmt.