I trained character-level n-gram models to play hangman to explore LLM concepts in a tractable setting. Check out the paper and experiments.
https://t.co/C7zJGANvF4
Tldr: trained on 370k-word corpus, trigram + linear interpolation ~ 8 wrong guesses per word.
@Ludvicus It’s Sucker Punch. Check out their breakdown of Ghost of Tsushima at GDC, 4 yrs ago. For most other studios, there isn’t a commercial culture or economic incentive for purpose-built engines and genuine problem solving. Hence the over-reliance on full-suite general-purpose tools .
@Creative_Math_ Join the manim community edition discord server, and read the docs! The community and maintainers are incredibly helpful. With their help I was able to create stochastic simulations of binary fission under antibiotic treatment.
How is AI reshaping human thought?
Terence Tao and Tanya Klowden propose a "Copernican View" of intelligence where humans are no longer the sole center. The practical takeaway: stop treating AI as a flawless oracle, and start managing it like a brilliant but inept grad student.
Garry's Mod creator says programmers over-relying on AI is like using too much porn: 'You lose your ability to ejaculate using your imagination' https://t.co/oGoFBvfPQO
Wrote a small paper on a simple but foundational stochastic model of binary fission (Bogdanov, Kevei, Szalai, et al) and an accompanying simulation using Manim.
Paper & code here: https://t.co/D4m8DAa1j7
Sim of 40 gens:
@MathType I first got exposed to Jordan normal forms and indecomposable representations of matrices from Michael Barot’s ‘Representation Theory of Algebras’. Great book 👍