ansuz @ansuz@social.cryptography.dog

in the end, I can't really get my hands on a GPU thanks to cryptocurrencies and supply chain issues, so platforms like these are the best option available.

I want to learn how these things work because I see critics of "AI" getting ignored by ML insiders who say they just don't understand it. Those who don't participate are shut out of the discourse.

All I can say is that this is a very classic Google Inc. move, spread the word about who actually owns what, and laugh at the absurdity of it all

I find it difficult to interpret this as anything other than corporate strategy:

1. running what is effectively a corporate sock-puppet makes it look like there's more competition in the space than is actually true

2. people unknowingly end up on a google site even when they consciously try to avoid it

3. google colab looks better by comparison because kaggle's UI is slightly worse

4. blogs listing ML cloud resources list google twice as much as their competition

I'd seen several google-related options on the site.. it integrates with gdrive and lets you sign in with your account credentials, but lots of independent services integrate with google in a similar way. Professional ML researchers that I know were completely unaware of this despite being familiar with the platform, and again, I'd heard it explicitly cited in a few places as a google alternative.

I don't have a spare computer with a GPU, but I was aware of google colab as a cloud computing option, and had heard kaggle.com mentioned as an alternative for people that wanted to get into machine learning while still avoiding google services.

After playing around with kaggle for a bit I got curious and decided to read a bit about the platform: their history, business model, and owners...

I was surprised to learn that Google bought them in 2017 😅

my partner is a neuroscientist with an extensive computational background. she's been showing me a bit about how some modenr machine learning techniques work, both because we've been thinking of working on some hobby projects together. I've been curious for a while whether these things live up to the hype and I wanted to get a sense of how much electricity, time, and computation goes into the sorts of projects that get so much attention in the media.

🎵
When the phasing of lines
frustrates machines and minds
that's a moiré
🎵

article via @grimmware

https://www.getrevue.co/profile/shift-happens/issues/moire-no-more-688319

1. submitting a feature request
2. reporting a reliably reproducible bug with full details
3. submittting a well-commented PR that fixes a bug
4. making a donation with no strings attached

It was merged!

https://ansuz.science/sprites/ animates a png sprite map from a final fantasy game (using the canvas API again).

If I keep having fun with little experiements like this I might end up integrating them with some other procedural generation ideas to make some tools for tabletop RPGs or something like that.

Of course, I'll probably want to drop the final fantasy sprites and use something creative commons like these: https://itch.io/game-assets/assets-cc0

https://ansuz.science/attractor/ draws a sierpinski triangle to an HTML5 canvas based off some random parameters.