Edward Yu

7 followers · 14 following · 265 views

on the atlas — 57

highlights — 9

Nintendo's counsel, John Kirby, countered that Universal had argued in a previous case that King Kong's scenario and characters were in the public domain (by way of the novelization of the 1933 film). Sweet ruled in Nintendo's favor, awarding them Universal's profits from Tiger's game ($56,689.41), damages and attorney's fees.[36]: 217 Universal appealed, trying to prove consumer confusion by presenting the results of a telephone survey and examples from print media where people had allegedly assumed a connection between the two Kongs.[118]: 118 On October 4, 1984, the court upheld the previou…
Donkey Kong (1981 video game)
Here we represent each burger recipe solely by its ingredients and their associated weights. Accordingly, a recipe x0 = {m, w}, consists of a binary ingredient mask m ∈ {0, 1}K that indicates presence or absence of each ingredient and the ingredient weight . We adopt a two-stage diffusion-based framework that decouples ingredient selection from ingredient quantification: Specifically, we integrate a multinomial diffusion model41 to generate ingredient masks with a score-based generative model42 to generate ingredient weights conditional on a given mask
Generative artificial intelligence creates delicious, sustainable, and nutritious burgers | npj Science of Food
The big mistake in this pattern of failure is projecting your subjective lack of comprehension onto the object you are looking at, as "irrationality." We make this mistake because we are tempted by a desire for legibility.
A Big Little Idea Called Legibility — Ribbonfarm
Johnson tells the researcher that perhaps they should start by examining his brain's baseline reaction to meaningless stimuli. He naively suggests a white-noise pattern as the right starter image. The researcher patiently informs him that subjects' brains tend to go crazy when a white noise (high Shannon entropy) pattern is presented. The brain goes nuts trying to find order in the chaos. Instead, the researcher says, they usually start with something like a black-and-white checkerboard pattern.
A Big Little Idea Called Legibility — Ribbonfarm
Women, despite being discouraged, were involved as participants, proxies for absent crusaders, or victims.
Crusading_movement
Chinese state propaganda outlets have latched onto the narrative of the Thucydides Trap in order to promote a set of power relations that favors China
Thucydides_Trap
China's main interest has always been its regional position, and if that is the case, then there are strong arguments for it to show patience, as its economic pull becomes progressively stronger
Thucydides_Trap
a number of pressing issues the two nations are at odds over that increase the likelihood of the two powers falling into the Thucydides trap
Thucydides_Trap
This is the punchline of the geometric analysis. RL is sparse but unbiased. SFT is dense and biased but diffuse. Same-family OPD inherits SFT's diffuseness because the teacher distribution is calibrated to the student's family. OPSD is the unique case where you get density, bias, and concentration at once — which is why it ships with explicit defenses (KL clipping, fixing the teacher to the initial policy) that the other methods don't need.
will brown on X: "On SFT, RL, and on-policy distillation" / X