It's not a bad idea to do something like that (highlight the fact that there exist machine learning techniques besides deep learning) but the website is extemely heavy. It took a while for it to load in my browser and I still don't know exactly what it's doing because it's so impossibly slow. A faster implementation, or a small introduction that would describe the function of the application before actually running it would make this a lot better.
Strange. I'm on a low end android and it was impressively smooth and loaded instantaneusly. Just came off LinkedIn for a contrast, where every touch of a button gives me time to refill my cup with coffee.
This was my first thought as well. The good thing at least is that it doesn't flash and load images. That is the worst. However, I don't mean to shamelessly self promote (outright lie) but my web app https://gif.com.ai has been getting feedback like it looks old school and low res. Well that's the point. It's intention is to be GIF (old school 90's) AI. And I appreciate this artists intentions and design sensibility and the zen garden programming effort as this is one of the first demonstrations of UX and AI in a somewhat seamless fashion. It does seek refinement. But once more 9 / 10.
Thanks for the replies. The page still takes a few seconds to open in my browser (FF 72.0.2 (64-bit)). I think it's maybe better today? Anyway I could play around with it today.
Reminds me of the Mocha visualization plug for Winamp from years back [1]. You can code a visualization 'bene' in the built-in language, and open up a window of 'evolved' benes which represented different tweaks to your algorithm.
The Page is completely nonfunctional for me on a low-end Windows laptop on either Firefox or Chrome.
The viewport is empty and black. Devtools shows that the page is spending all it's time adding countless SVGs, but placing them in a div with overflow: hidden so they're all invisible.
Oddly the page appears to work without issue if I put the browser in mobile emulation mode.
It works for me, both in Chrome and Firefox on a laptop that is neither high end nor low end. Maybe, try again as I see you posted this about an hour ago??
Firefox had blocked the HTML5 canvas data request for me, which broke the page as you outlined. Once I allowed the canvas data request, and refreshed, the page eventually loaded correctly.
It's a mutation simulation with colorful shapes. You will get a autogenerated sequence of 8 shapes (on my screen at least, you might get fewer for a smaller screen) then you can choose to start mutating from either shape. The new mutated shapes have the same option to mutate from so you can mutate until you're happy with the said shape. Once satisfied you can add it you your garden (your repository of shapes you chose to keep). You can fork from any of the shapes and mutate them further. There's a refresh button that autogenerates new shapes, took me a while to spot it.
I personally find this very interesting both conceptually and visually. I've been doing myself a lot of autogenerated graphics in Racket from lists of features but never went as far as to show it to HN.
This is very similar to Picbreeder [1], which allows you to evolve art in a very similar manner, with evolutionary algorithms. It runs on the same evolutionary algorithm behind MarI/O [2], called NEAT [3].
18 comments
[ 3.1 ms ] story [ 48.0 ms ] thread[1] http://www.technosis.com/mocha/
The viewport is empty and black. Devtools shows that the page is spending all it's time adding countless SVGs, but placing them in a div with overflow: hidden so they're all invisible.
Oddly the page appears to work without issue if I put the browser in mobile emulation mode.
On my case (Android and firefox) it's fast and seems to work well.
I personally find this very interesting both conceptually and visually. I've been doing myself a lot of autogenerated graphics in Racket from lists of features but never went as far as to show it to HN.
[1] http://www.picbreeder.org/
[2] https://www.youtube.com/watch?v=qv6UVOQ0F44
[3] https://en.wikipedia.org/wiki/Neuroevolution_of_augmenting_t...