InquilineKea's Shortform

What are some of your favorite search terms to use on LessWrong (or any site with lots of "interesting" content) to find important/interesting things/or things whose patterns self-recur at the most fascinating intervals?

(the best terms are terms that often mix the technical in soulful contexts). Most technical work completely lacks soul/vitality and that's why many of the most enlightened people hate academic papers with vicious intensity

I'll start with some: 

higher specificity [you know it's good when it's usdd, common enough to try w/o frustration, and they're not common enough to be incorrectly used too often, but rare in most corpuses]: deconvolute, perplexity, Lyapunov time, fractal, td-idt, wavelet, Markov boundary/blanket, friston, Hamiltonian

maybes: Perlin Noise, Nyquist sampling, hypergeometric, postsynaptic density, calmodulin, neuregulin

just for LW: Hierarchical temporal memory

higher-sensitivity [above-average chance of being interesting]:

above-average chance at detecting neurodivergence:  Laurens Gunnarsen, gwern, nickcammarata, Adderall, 2CB, 2CT7 [the latter 3 are interesting on "dry nerd" forums, but low-signal on reddit]. Searching for specific MBTI or Enneagram types can be interesting on LW or high-signal discords/FB groups

maybes: speculative decoding

sites to mine for good neologisms: qualiacomputing, cyborg wiki, Stephen Wolfram's blog, Anders Sandberg, Taleb (eg fat tails), Robin Calhart Harris. information theory terms like suspense/surprisal/KL divergence/mutual information, the spin glasses literature (eg ising model), some models of game theory [carl bergstrom has some cool neologisms..] also John Conway

Theoretically: compiler engineering, functional programming, category theory, quantum computing, and systems people. It is not common for people to come up with the most important neologisms from these fields, but someone skilled could do it really well (especially in capturing common patterns in data access patterns as more and more of the Internet is made of human-free compute)

specific to LW: https://en.wikipedia.org/wiki/Hierarchical_temporal_memory

https://twitter.com/nickcammarata/status/1769890642652803460