#resplendent

i admit im kind of shy re: thinking and talking about OCs interacting with canon characters BUT this gives me an excuse to fill out a little character meme I'd wanted to do with Scout for a while, soooo >:3c rubs my paws together deviously

since this meme has 9 spaces im throwing Scout's Windy, Ai, and KusanagiKolter opinions into the ring as well! in general, I picture all of these relationships specifically with the offline versions of these people, Scout included--I definitely think she has some large level of agoraphobia IRL, and it's only in Link VRAINS as Resplendent she lets herself not be strangled by fear and anxiety. So actually meeting, getting to know the other Lost Incident victims and Revolver properly, she's kinda afraid of all of them! Takes a while for them to not be turbo on edge even with the sweeter people like Soba and Miyu ;_;

Gonna elaborate more on these individually under the cut in case I start Rambling

Obsidian Lobster, to celebrate legendary resplendent Ryoma <3

The black looks great on him! This is what I want in a resplendent, a different colour scheme / unexpected theme.

put Resplendent on our tomolife island and made Windy for her and she had the gall to think it wasnt a good match her GIRL THAT THING CAME OUT OF YOUR BRAIN. THATS YOUR TRAUMA BOOGER.

Hi all! I am taking part in @cdramazine's Resplendent - a zine dedicated to Chinese Dramas.

Not unexpectedly, I have written a fun Nirvana in Fire Lin Chen and Mei Changsu jianghu romp, complete with hidden knives.

Preorders open on April 30th and all of the art and fic look amazing! Plus there is some brilliant merch, so do come and check it out!

[...]One of the questions most often brought to Langya Hall is Chief Mei’s travel itinerary. Lin Chen does not shy from answering it (though he does ensure that the price is suitably high). Langya Hall is neutral, after all. Furthermore, if Mei Changsu can’t deal with such a minor inconvenience as having a public schedule, then he does not deserve to top the Scholars list. “Lin Chen,” Mei Changsu replies, giving a polite bow and settling himself primly into the abandoned seat. “By now I would have expected you to realise that I take an interest in everything.” “Everything?” Lin Chen echoes, snapping open his fan and bestowing his most lascivious grin upon his friend. He can hear mutters from the crowd behind them, none of whom would know subtle if it walked up and hit them over the head with a tea-brick. Lin Chen despairs of the state of the jianghu, he truly does. He consoles himself, as he’s often done before, with the thought that at least their collective stupidity ensures his Hall remains relevant.

AI deployment is a double-edged sword. The allure of artificial intelligence is undeniable, promising to revolutionize industries and solve complex problems with unprecedented efficiency. Yet, beneath this resplendent facade lies a mathematical conundrum: the impossibility of fairness.

At the heart of this issue is the realization that no classifier can simultaneously satisfy multiple fairness constraints. This isn’t just a philosophical dilemma; it’s a mathematical certainty, as demonstrated by Choquet’s theorem and various impossibility theorems. Consider the competing fairness metrics like equalized odds, demographic parity, and calibration. Each aims to ensure fairness in different ways, but they are often at odds with one another. For instance, a classifier that achieves equalized odds (ensuring equal true positive and false positive rates across groups) may fail to meet demographic parity (ensuring equal selection rates across groups), and vice versa.

To understand why, let’s dive into the statistics of disparate impact using confusion matrices, precision-recall curves, and ROC-AUC scores across demographic groups. These tools reveal how classifiers perform differently across populations, often due to inherent base rate differences. A Bayes-optimal classifier, which is theoretically the best possible model given the data, will naturally reflect these base rate disparities. It optimizes for accuracy, not fairness, and in doing so, perpetuates existing inequalities.

Algorithmic fairness interventions, while well-intentioned, often merely shift discrimination from one metric to another. For example, adjusting a model to improve demographic parity might inadvertently worsen equalized odds. This is not just a theoretical concern; it’s a practical one, as seen in recent AI-related stories where overpromised capabilities led to failed projects and funding bubbles. (Remember the hype around AI systems that promised unbiased hiring decisions, only to be scrapped after perpetuating biases?)

The confusion matrix, a staple in evaluating classifier performance, illustrates this trade-off. It breaks down predictions into true positives, false positives, true negatives, and false negatives, providing a detailed view of where a model succeeds and fails. Precision-recall curves and ROC-AUC scores further quantify these successes and failures, highlighting disparities across demographic groups. Yet, even with these tools, achieving fairness remains elusive.