derek kwan關志豪🦦🐈🍛🐧🍮 @mrufrufin@post.lurk.org

I've been swapping microsds amongst pis and am now booted from my broken laptop's ssd in an external enclosure through a desktop tower and it's like interacting with my broken laptop but not and it's kinda throwing me down some existential paths

use cases would be maybe gaming lol, but audio live coding hopefully again one of these days, but more importantly ai/ml/cv stuff. i guess self-repairable would be good too but doing computer surgery today reminds me of why i barely do any hardware projects (i have clumsy, half-broken hands despite being a percussionist...)

that being said, what should my next laptop be? linux-friendly a must. a dedicated gpu would be nice (but i never quite got it to work on my old machine outside of nouveau drivers), intel preferable and some sort of decent warranty plan (bc i break things). having physical ports would be nice too. a somewhat nice, sturdy case would be nice too. i had a dell xps 9550 that was ok (15.4 inch, metal top and bottom, some sort of carbon fiberish plastic body), before today the touchpad didn't click anymore, some keyboard keys got stuck, and i replaced the fans once, and the whole thing once due to battery swell, but it lasted 5-ish years. should i get another dell xps? go lenovo something (are they still good)? something else like asus or something? have the heartbleed/meltdown things been sorted out yet? is it a right time to buy or is some new architecture gonna drop soon? i have a windows tower so i'm ok for now but by next february-ish i need something.

lol spilled water on my laptop (and a lot of other things) and it has the flashing orange light of doom (even after taking a hairdryer to the parts). ordered a usb nvme enclosure so hopefully i can mount the old ssd on an old linux-enabled machine (and luks won't complain) and recover stuff from it (like projects due wednesday oops).

i feel like i'm doing the right thing but the autograder says no

my vim sessions are starting to mirror my internet browsing sessions in that i open way too many tabs so i'm finally learning about :mksession and now i'm wondering if there's a tmux equivalent

does the existence of newtypes in haskell make haskell canon in the gundam uc universe

fun fact: you often import numpy as np and guess what, the np in np-complete is the very same so any np-complete problem can be completely solved by numpy

AI pretty much done, next is Computer Vision and (mb) Computer Networks

AI was fun and also mb way larger of a field than I was expecting. It feels like different sorts of strategies to solve intractable problems "well enough" for "most of the time". Which is cool and also I guess more like how the "real world" works?

CV is to continue on my ML/AI learning road and back on the art side, I've encountered OpenCV way too often and not known enough about how to use it properly that taking a class for me is pretty much a must-do. Also supposed to be one of the harder classes in the programs so that makes me a little nervous...

CN hasn't gotten the best of reviews overall in terms of the program (which is why I'm thinking of switching out) but also I've never taking a networking course and I find the topic interesting, I'm interested in learning how computers talk to each other and things get routed every which way and the knowledge might come in handy someday, although not directly applicable to all the ML/AI stuff I've been learning and want to get out of this program but also I'm kinda afraid if I take a heftier class (than CN) that it'll take away from the time I can dedicate towards CV,.. so mb I'll ending up sticking with it, ugh, I guess I have a week to decide.

is boosting in ai/ml the algorithmic equivalent of "throwing sh*t at the wall and seeing what sticks" and then how come it works so well

just made a "can't see the (random) forest for the (decision) trees" joke in the forum for my ai class and umm mb should take a break for a bit

gave up linux/ble stuff for now and watching lecture vids on probability and bayes nets for ai class. probability was def one of my weaker subjects back in my math degree days (alongside group theory and multivar calc but i blame that on sleeping through 90% of the classes) but actually bayes nets seem pretty neat and mb i would've liked probability better had i been exposed to bayes nets earlier, they're like a structured way of figuring out chances of things given potentially really messy dependencies and relations.

seems to connect ok with my pixel 3a so maybe bluetooth is borked on my computer somehow? Or I can try reinstalling bluez for the 3rd time lol. hmm, mb will bust out the pi to see if it's just my compy

currently lamenting how all of the python libraries i've tried can detect my arduino nano ble sense but none of them can actually connect to it, despite blueZ upgrade attempts =(

quick link to model: https://derekxkwan.github.io/talea-vis/

#NIME website is updated so I guess I can share my work with Anne-Sophie Andersen that's part of this year's conference. http://nime2021.org/program/#/paper/76 It's a web-based 3D model based on analysis done on Gerard Grisey's composition for chamber ensemble Talea that you can explore, mouse-over for analysis details, and press P to play relevant snippets from a performance. I used #threejs for all the 3D bits of the model (generated on system load), #p5js and #processing to generate their textures, and #python for a bit of data conversion here and there.

and also the MC is basically playing a twin-stick shooter on their phablet for half the anime and when it comes to controlling a real robot, prefers to use the same interface? which makes me wonder how it that works if he isn't staring at the screen. do the twin sticks transpose on the screen to whether his fingers press? how error-tolerant is the device to him accidentally lifting his fingers off the screen then momentarily (distinguishing between transposing controls vs pushing a stick to the extremes of a range)? Or maybe it's just muscle memory or mb there's some sort of haptics involved? How does a touch-based interface manage to be expressive (capable of controlling various dimensions and triggering various techniques) yet error-tolerant?

also i find myself thinking about maybe what sorts of connections the phablet controller has with the physical robot (probably bluetooth? in that case, how does it manage 10+ feet ranges, does real bluetooth do that?)

also i should stop procrastinating and get back to class readings

i've been watching robotics;notes (which isn't about my ai for robotics notes from last semester thankfully) and everybody has weatherproof phablets with nifty AR features that they all keep side-holstered and maybe it's noteworthy how their digital existences seem tied to them but i'm more intrigued by how the batteries seem to last forever and how they seem to be a legit competitive gaming platform.

side-effects galore! very clean code 👍

it seems like throughout my life over and over again i end up having to do some sort of osc to web browser bridge via websockets and i used to do it in node but i'd figure i'd learn async programming a little more and try it in python where it isn't so automatic so here is my hacky, baby-deer-in-async-land, wip python version https://github.com/derekxkwan/osctowebskt-py