coolstuffyouwilllike

[coolstuffyouwilllike]

Continuing my ongoing investigation into traditional forms of wrestling and my current fascination with the aesthetics of muscular human bodies, here is Indian mud wrestling. It seems like the industry is pretty exploitative, which is sad but not surprising. The physicality itself, though, is very impressive, and the aesthetics of these guys covered in red mud is fantastic.

[coolstuffyouwilllike]

Actually rather than writing a TTRPG itself in toki pona, I think something that could be interesting to do would be to make a TTRPG that literally just uses toki pona as its spellcasting system.

Like. You know how Dungeon Master (the 1987 CRPG) had a spellcasting system where you constructed spells by combining runes to describe the form and function of the spell (e.g. to make a fireball you combine the fore rune with the wing rune). It would be like that but with the freedom and flexibility that comes from being run on a tabletop game instead of on a computer in 1987. And the rules for how you construct spells would just be regular toki pona grammar.

[coolstuffyouwilllike]

im preparing to defend my thesis on therapeutically applied ttrpgs tomorrow and im sitting here and pondering ttrpgs and how unique they are as a genre and how you might describe them as texts and objects to someone who's not too familiar

because like ttrpgs (the texts) are very creative and theyre very artistic and inspiring and beautiful and moving and they're also ... instructional manuals. lmao. yknow? theyre Technical Writing by and large unless youre looking at like a lyric game or something. the main purpose of a ttrpg text is to instruct you how to carry out a series of tasks and rituals in order to produce a desired outcome. theyre instructions on how to facilitate a specific experience

and theyre pretty unique in that as a genre (in the writing studies sense of the term - a specific type of creation with shared features, uses, and intentions)

like kind of the only way i can think to compare them to something else is if you had a play script, right

and in the play script you have a description of all the characters. and you had a description of the setting. you know how many acts there are, maybe those acts have titles. the play might give you suggestions on how to design the stage and what props to include. and then as you're flipping through it all the stage directions are there and you know which order things are supposed to happen in

but all the character dialogue is blank

you cant really look at that and say "wow ive been moved by this story" because like. well. the story isn't there yet. it's potentially easy to tell what the intent of the play is - what kind of story it's likely to tell. you can look at the set of characters and reason out what might happen if you put them all together. you might even know some of the actions that are meant to take place in this play because of the stage directions (does someone shoot a gun? do the characters cry? does everyone fall over dead or do they dance?)

but at that point, even though it's creative and maybe even beautiful, this is really mostly an instructional manual. the story hasn't been told yet. there's a lot of room to imagine and generate and create - within the confines of the instructions they've given you. surely this is art, right? but what kind of art is it?

ttrpg texts largely aren't telling a story, they are instructional manuals on how to generate a specific story in the confines of that specific game. beautiful, lovely, creative instructional manuals that are really hard to explain to anyone who hasn't been involved in the experience of generating from one before

but ... yea :) thats the best metaphor i could come up with. thesis defense tomorrow !!

[coolstuffyouwilllike]

Actually rather than writing a TTRPG itself in toki pona, I think something that could be interesting to do would be to make a TTRPG that literally just uses toki pona as its spellcasting system.

Like. You know how Dungeon Master (the 1987 CRPG) had a spellcasting system where you constructed spells by combining runes to describe the form and function of the spell (e.g. to make a fireball you combine the fore rune with the wing rune). It would be like that but with the freedom and flexibility that comes from being run on a tabletop game instead of on a computer in 1987. And the rules for how you construct spells would just be regular toki pona grammar.

[coolstuffyouwilllike]

The Valley of the Ghosts (Kutkhiny Baty) on Kunashir Island, Russia. The soft volcanic ash has been sculpted over centuries by wind and rain

[coolstuffyouwilllike]

This little one acts completely feral unless you make the "mommy greeting kittens" noise at them. When my phone isn't making them unsure they will come charging to the front of the cage for attention by the second trill.

You can see the body language change in the ears with the first and second trill real well

[coolstuffyouwilllike]

okay so if you need more veggies/fruit, protein or fibre (bc most people do NOT eat enough) in your diet but you struggle to do so, hear me out:

look up recipes (especially snack recipes) that are child/toddler/baby-friendly

i can guarantee there is a woman with a cooking blog out there who has found away to pack a bunch of vegetables into a surprisingly delicious little snack for her kids. this process has never failed me when i feel like i am not eating enough fruits and veggies. my entire flat is eating spinach muffins at the moment, which doesn’t sounding particularly appealing to most people and yet somehow. they’re delicious.

[coolstuffyouwilllike]

Collage #1 (Blue Suede) - James Tenney (1961)

An early electronic composition made from manipulated samples of Elvis Presley’s version of “Blue Suede Shoes.”

[coolstuffyouwilllike]

The photo is not very good and I still need to wash and iron it, but I couldn't wait to show my first ever big cross stitch project;

Love how it turned out, especially with the green embroidery thread (cotton thread on linen)

[coolstuffyouwilllike]

Ch👏

[coolstuffyouwilllike]

I believe one of the most underrated math/coding websites of all time is projecteuler.net. As one of the top 2000 ranked in this website, I feel like I can say just how amazing this website is. Here is a list of everything I love about it.

Well thought out idea: This website is full of math problems in which (almost) all of them can only be solved by code, but you cannot just use brute force. You need to use the best of math and CS to solve the problems.

Good problems, and not just a couple. This website has >900 well thought out problems that can be solved in under a minute even on incredibly slow computers

Variance in difficulty: There are problems that will take you under 5 minutes to complete, and others will take you weeks without making a lick of progress (looking at you #177). Even if you don't know anything bout number theory or CS, there will be problems you can solve and learn new concepts with. Just dost skip to the recent problems, they're VERY hard.

Price: I won't sugarcoat it. It's free. Completely. No need to pay for problems. No missing features. No ads. Anyone can make a free account and have the exact same experience. This might be it's biggest selling point in my opinion.

These problems have made me spend hundreds of hours confused and dumbfounded until I find a solutions. Now, there are definitely sites you can find that will give you the answers But don't use them for solutions or else you'll be cursed for life. They're good places if you want to see how others solved it, but it is 233168% better to just skip on these websites and solve them yourselves. Even after all of this time, I have never copy-pasted an answer, purely because I prefer the struggle, and I believe that I learn more by finding out the necessary info myself.

Overall, stop reading this and go to the website (projecteuler.net). It deserves all the praise it can get. I hope you enjoy the problems like I do :)

[coolstuffyouwilllike]

sarah cahill is one of my favorite living pianists. The piece is by Leo Ornstein

[coolstuffyouwilllike]

big find for me tonight, can’t recommend it highly enough, kinda freaking out. this one’s a collab between the “A” Trio, the ‘oldest free improv group from Lebanon.’ They’re fabulous, a lot of their albums still have the meditative quality that the opening of this one does but also a lot of fun, all worth hearing (I really liked “A Tower With No Imam” off their album Folk). They’re playing here with AMM (Eddie Provost & John Tilbury) who were one of the first and greatest British free improv groups (ridiculous body of work amongst each of them and the AMM itself, Tilbury is one of the best performers of Feldman and definitely Cardew in my opinion.)

anyway that’s the nerd history pitch mostly this stuff is beautiful, really leaves you with something. If you like noise when it’s ‘quieter’ and don’t mind instruments (that’s a joke), if you’re into stuff like this at all, and maybe even if you’re not—-I think a thing like this could turn people onto it. wow. cool shit.

[coolstuffyouwilllike]

Old-school planning vs new-school learning is a false dichotomy

I wanted to follow up on this discussion I was having with @metamatar, because this was getting from the original point and justified its own thread. In particular, I want to dig into this point

rule based planners, old school search and control still outperform learning in many domains with guarantees because end to end learning is fragile and dependent on training distribution. Lydia Kavraki's lab recently did SIMD vectorisation to RRT based search and saw like a several hundred times magnitude jump for performance on robot arms – suddenly severely hurting the case for doing end to end learning if you can do requerying in ms. It needs no signal except robot start, goal configuration and collisions. Meanwhile RL in my lab needs retraining and swings wildly in performance when using a slightly different end effector.

In general, the more I learn about machine learning and robotics, the less I believe that the dichotomies we learn early on actually hold up to close scrutiny. Early on we learn about how support vector machines are non-parametric kernel methods, while neural nets are parametric methods that update their parameters by gradient descent. And this is true, until you realize that kernel methods can be made more efficient by making them parametric, and large neural networks generalize because they approximate non-parametric kernel methods with stationary parameters. Early on we learn that model-based RL learns a model that it uses for planning, while model free methods just learn the policy. Except that it's possible to learn what future states a policy will visit and use this to plan without learning an explicit transition function, using the TD learning update normally used in model-free RL. And similar ideas by the same authors are the current state-of-the-art in offline RL and imitation learning for manipulation Is this model-free? model-based? Both? Neither? does it matter?

In my physics education, one thing that came up a lot is duality, the idea that there are typically two or more equivalent representations of a problem. One based on forces, newtonian dynamics, etc, and one as a minimization* problem. You can find the path that light will take by knowing that the incoming angle is always the same as the outgoing angle, or you can use the fact that light always follows the fastest* path between two points.

I'd like to argue that there's a similar but underappreciated analog in AI research. Almost all problems come down to optimization. And in this regard, there are two things that matter -- what you're trying to optimize, and how you're trying to optimize it. And different methods that optimize approximately the same objective see approximately similar performance, unless one is much better than the other at doing that optimization. A lot of classical planners can be seen as approximately performing optimization on a specific objective.

Let me take a specific example: MCTS and policy optimization. You can show that the Upper Confidence Bound algorithm used by MCTS is approximately equal to regularized policy optimization. You can choose to guide the tree search with UCB (a classical bandit algorithm) or policy optimization (a reinforcement learning algorithm), but the choice doesn't matter much because they're optimizing basically the same thing. Similarly, you can add a state occupancy measure regularization to MCTS. If you do, MCTS reduces to RRT in the case with no rewards. And if you do this, then the state-regularized MCTS searches much more like a sampling-based motion planner instead of like the traditional UCB-based MCTS planner. What matters is really the objective that the planner was trying to optimize, not the specific way it was trying to optimize it.

For robotics, the punchline is that I don't think it's really the distinction of new RL method vs old planner that matters. RL methods that attempt to optimize the same objective as the planner will perform similarly to the planner. RL methods that attempt to optimize different objectives will perform differently from each other, and planners that attempt to optimize different objectives will perform differently from each other. So I'd argue that the brittleness and unpredictability of RL in your lab isn't because it's RL persay, but because standard RL algorithms don't have long-horizon exploration term in their loss functions that would make them behave similarly to RRT. If we find a way to minimize the state occupancy measure loss described in the above paper other theory papers, I think we'll see the same performance and stability as RRT, but for a much more general set of problems. This is one of the big breakthroughs I'm expecting to see in the next 10 years in RL.

*okay yes technically not always minimization, the physical path can can also be an inflection point or local maxima, but cmon, we still call it the Principle of Least Action.

[coolstuffyouwilllike]

Polynesians did also rely on a form of a physical map called a stick chart, illustrating the specific wave and swell patterns surrounding different island chains. These were particularly helpful during cloudy conditions when the sun and stars were less useful. To navigate the Marshall Islands, the Marshallese represented ocean swell patterns using parts of coconut fronds and shells as islands. Like a subway map, they don’t so much represent distances as they do relationships. The complex and decorative stick charts were often only understood by the person who made them. They were memorised before a voyage by the pilot who would lie on the floor of a canoe to get a sense of swell movement and often lead a squadron of 15 or more boats.

[coolstuffyouwilllike]

NICOS!

CRT-TV Drums. The sound is produced by catching static electricity from a CRT with the hands and sending electrical signals through the body to a guitar amplifier📺

ー

ブラウン管ドラム。ブラウン管テレビの静電気を手でキャッチして身体を通して電気信号をギターアンプに送ることで音を鳴らす📺

Played by Ei Wada

Created by Ei Wada x Rinichi Washimi x Nicos Orchest-Lab

NEW SINGLE / VIDEO ALBUM

[coolstuffyouwilllike]

The Australian Ballet is doing Alice in Wonderland again and on one hand I’ve seen it before, and on the other, their Queen of Hearts has my favourite costume in anything every