Instagram story viewer> @shegoeslocofoco> Posts

Pi🌈🦊🎧

@shegoeslocofoco

523

followers

447

following

Sound Design @guerrillagames
Curious person. Soft unless provoked (or hungry). Here cos nobody is leaving to pixelfed😞 || www.pinartemiz.com

POSTS STORIES REELS TAGGED

Cooking something finally!

Unbearable Lightness of Being Optimized
-- searching for desirable difficulties in the time of 'ai' --

What is it?

A series of , initially online 'gatherings' for frustrated makers, designers, educators, and technologists who want to get curious about their relationship to their work, their tools, and each other.

We will meet in small groups to confront a few increasingly nagging questions:

- Which parts of our thinking and making could stay effortful, intimate and which parts can we hand off without losing what we care about?
- What alternatives to corporate cloud AI exist, and how far can a few people get using them?
- How do we keep our reliance on each-other intact and deliberate, rather than letting it be replaced by monopolized dependency?

It will be half conversational, philosophical and half hands-on, focused on the tangible.
Over several sessions, we will (at least):

- Build personal maps of which relations we care to keep intimate and protect from default modes of automation.
- Redesign some of our active workflows/pipelines to keep learning when it matters.
- Run local models on our own machines and discover what to look for in capacities and limitations.
- Explore the possibility - and reality - of pooling compute or resources with our chosen communities.

Details
Size: Max 8-10 participants (mixed backgrounds hopefully).
Format: Five bi-weekly 90-minute sessions (open for discussion)
Place: Online for now on Signal + Jitsi (hope to create local pockets later if it sticks)
Time: Start date will be decided with the group in discussion. (Expect this year tho)
Cost: is commitment. This will need active participation, presence, and a shared agreement to keep conversations private.

If you are curious, a little stuck, and want to explore practical alternatives with a small group, send me a message to join to ride.

Peace out!😊

#toolsforconviviality #stayingwiththetrouble #communitybuilding
#creativetechnology
#aiandcreativity #artificialintimacy
#workingtogether #interdependence by @shegoeslocofoco

12

1

3 months ago

Circadian rhythm got wonky so here is a late nite doodle/cover of harvest moon. Feels like a lullaby tbh.🌕

#somaterra #harvestmoon #lullaby
#coversongs #singingcover by @shegoeslocofoco

43

31

10 months ago

LUCIER'S ROOM

Alvin Lucier fed his voice into a room 32 times until only the room's resonant frequencies remained - the room itself becoming the instrument, its acoustic signature made audible through the destruction of speech.

This is the LLM analog. The seed - Lucier's original text - is fed to open-weights transformers running locally. In chain mode each output becomes the next input, the seed appearing only once. In other modes outputs accumulate as context, or tokens survive and decay weighted by the model's own attention. The internal activations are captured at every iteration using TransformerLens hooks: residual stream drift per layer, attention patterns, KV cache state, logit distributions over the vocabulary. The text transforming iteration by iteration is shown alongside the drift of the internal representations through vector space.

A. Lucier's room revealed something specific to that space - its geometry, its materials, its physics. This room reveals something specific to its training - what text was included, in what proportions, toward what objective. Different models produce different attractors. The room changes depending on which weights you load.

It is not always clear, however, whether what stabilises is meaningful - a fingerprint of the training data - or just a small model breaking down. Alvin Lucier's room had no such ambiguity. This one does.

More on: https://www.pinartemiz.com/thinking/luciers-room-research/

#machinelearning #mechanisticinterpretability #alvinlucier #localllm #mediaart #experimental #artisticresearch by @shegoeslocofoco

10

0

5 months ago

guess what I named this project? 😊

(there is sound in this post, and 2 videos.
also; the last boop in the first video is too cute imo ahaha)

still fooling around with fitness functions, spatial relations and starting conditions, nn architectures (very simple ones obviously) and how all these create different patterns in the process of "training/learning" over generations... by mapping velocity, rotation, position etc values to simple synthesis parameters I get to hear these "patterns" not only see them...

Next idea is to grade them against certain rhythmic input (by user) and see if they can approximate the "beat" by controlling their flight-speed between targets over time... atm they dont have any awareness (big word) of that type of sequential memory... altho, even now if I change target positions this results in very uncontrolled rhythmic or arpeggiated motifs as well, at least until they converge and start kamikazeeing all at once lol, but giving them some temporal inputs and adjusting the fitness function to score for that might make it more interesting... will see.

what I like about this mini project is that what's really interesting happens in the messy middle =) process focused much?

if you'd like to experiment with RL yourself check out: https://gymnasium.farama.org/

#creativecoding #reinforcementlearning #stochasticsequencer #evolutionaryalgorithms #artificialneuralnetworks #whenshithitsthefan #foolingaround😜 by @shegoeslocofoco

15

5

8 months ago

should give trypophobia warning with this... sorry.

in the beginning there was the warped noise fields.... an invisible mathematical "terrain" that defines where the digital tissue grows and how it twists.

the simulation then morphs between "clustering" (tubular growths, and tiny holes that, quiet frankly, look disgusting) and "pulsing" (excitable waves you see towards the second half). this is the reaction-diffusion part.

to give it that 3d-ish look (a bit of depth), heightmap shading is used. essentially mapping the reaction density to height and using it for normal calculation. this recreates light and shadows on the surface, making that "digital goo" look imo very latexy...

I obviously hacked and slashed most of this together. beyond reaction-diffusion algo everything is new to me.
but will share the shadertoy link when am done with cleaning it up

#reactiondiffusion #shadertoy #noisefield #creativecoding #proceduralgeneration by @shegoeslocofoco

12

0

8 months ago

slicing through a morphing "slime-putty" to observe "interference patterns" of the warped noise field that's flowing through it.

in other words: taking the signed distance field (SDF), which normally creates perfect concentric circles, and adding the gradient noise to it before calculating the sine wave lines.

I still barely know what am doing but learn through play is the method here =P

watch this for more education:

https://www.youtube.com/watch?v=f4s1h2YETNY

#shadertoy #creativecoding #proceduralgeneration #playground by @shegoeslocofoco

11

0

8 months ago

Snapshots from ongoing particle simulation explorations (and by that I mean I barely know what am doing😂)
4th one is my favourite tho, so far, i think...

#threejs #reactiondiffusion #minimalsurface #creativecoding #audiovisual #proceduralgeneration by @shegoeslocofoco

32

6

8 months ago

I know we are all supposed to use touchdesigner these days (blob tracking for days as the canon aesthetic) BUT am just hacking and slashing threejs particle simulation examples to create a frankenstein of an audiovisual thingy majiggy. Outer boundaries are relatively easy to figure out, but not sure if the surfaces (patterns inside those volumes, think of em like magnetic fields attracting the particles) are implemented correctly; considering I have no idea how to code the required math (PDE) for them or the patience to learn, I am relying on my ignoramus copy-paste skills(!) and only will find out once I show it to people who know better lol. Still. FUN.

Oh and no, audio part is not there yet. So no sound here.

Go check yourselves what else people do with it: https://threejs.org/examples/

#particlesimulation #threejs #creativecoding #trialanderror by @shegoeslocofoco

17

3

8 months ago

WIP as uje😬 now practicing the vox lines. Lower register folx in my socials, who might like to lend their voice ping me please. (espesh my queer fam; extra espesh trans folx 🙏🏻) Not time limit for reaching out as all my personal creations take forever to be called 'finished' lol #songwriting by @shegoeslocofoco

12

3

10 months ago

NOT a review just dopamine dump after recieving the delayed package 📦 I hope I won't be late the queer market cos I just wanna play with this now actually 😂

#somaterra by @shegoeslocofoco

33

21

10 months ago

From the 7 year old first post on fb, before Aya joined me in life, to latest shots captured by friends at work, and selfies on the floor of our studio where I could catch her occasional goofy side. She is truly the sweetest part of my days. @aya.the.shiba 🤍 by @shegoeslocofoco

40

6

a year ago

This is such an odd ball post historically (no dogs lol), but I don't wanna go through my archive to find it again so it deserves a front drop.
During the visit this year, my aunt as usual allowed me to do an impromptu experiment and let herself be the guinea pig for tasting. 🤍🫂
One of them (juniper) is easily on the top of the list for the best tasting ice-creams I have ever tried, not joking. 🤤 by @shegoeslocofoco

27

10

a year ago