Do you think there is a silent war going on between walled gardens where each data scraping team is trying to automate scraping of other platform’s copyrighted data, disregarding any TOS?
@bitinn at this point - they have enough data, and I think - they planning or already making - removing information from internet to make "quality precise information" available only to them.
Flooding internet with AI-generated average-precision data is part of it.
And only half of that is related to Rust! The rest are very good points about hype cycles, tech zealotry, search for silver bullets, being constantly sidetracked by shiny interesting tech instead of solving a problem, etc. etc. Very long, but also very good read.
@Taffer >I want to use Wayland. I have no idea how to debug the hang
I assume you have amd-gpu.
Do you have - xf86-video-amdgpu - installed?
I installed clean opensuse week ago - and this package was not installed for some reason by default, so I had only software-rendering in x11.
Emulation is often in proximity with piracy talk, but most companies will use that to distract from poor service they provide in terms of access to legacy content. Providing a good legal service to access that catalogue would help, we just have to look what Steam did for PC
1/3
@djlink >Xbox did a good job bringing X1/360 titles to Xbox One
idk if you know - PS3 had builtin PS2 and that PS2 were used for PS2 compatibility, latter they released "PS3 slim" that lose ps2 support because they cut needed hardware part.
Same with xbox - they had some parts of xbox360 in xboxone for emulation to work.
in context of wii emulation - it had completely different hardware than modern switch.
I think this is hilarious and just goes to show that even they think it's wrong, but as long as they can profit on others work, and not the other way around it's all good.
@djlink download internet today - and sell it in 1 year, because in 1 year internet will be flooded with AI-gen content, and AI-gen content can not be used for training.
I've put Not By AI badges on my website (see https://notbyai.fyi/); is there something similar for sites that don't have any tracking/analytics/etc. garbage on them?
@Taffer OpenAI - is only competitor to Google in web search, and ChatGPT alone did alot of "damage" to google-search by allowing people to easy generate "statistically correct distribution of tokens" as web-pages.
Google dont have way to filter AI-generated content.
Google also need "human made content" to train their AI.
Using badge like this - you just "hand sort" data for Google, allowing them easy find "not AI website".
I hope this "not by AI" will be used by every AI-website.
@bitinn Adobe license allow Adobe to use all images created with photoshop in Ai-training. Because photoshop license say "Adobe own all products created with its software".
Same with Microsoft and office - Microsoft own all documents created with office-software.
NEW: Tumblr and Wordpress are preparing to sell user data to Midjourney and OpenAI for AI training, according to internal documentation we've reviewed at 404 Media
Though as Sebastian Aaltonen showed in his talk on mobile rendering, if you want a good coverage of mobile devices right now, then Forward rendering is still the way to go.
GPU driven rendering has multiple bottlenecks on mobile that makes it really more of a tech demo than real possibility unless you are only targeting the most recent top tier hardware. This is not currently viable business wise.
@wawe@memes in releasing indie game you have much better chances of winning than in lottery.
Release multiple games, make simplest concepts to randomly get one of game to trends
I can't help but feel like it should be running better than my CPU in cycles renders (seems to get the same time as my CPU of 48 seconds on medium-detail render) but it isn't. I'm borrowing my brother's RX 6650 XT just to test blender performance with the card and it's seemingly.... mid? #b3d#blender3d#3d
@eoinoneill I thought getting AMD RDNA3 as "playground" to do some commits to mesa/linux kernel.
I have amd Ryzen with integrated GPU, and it was insanely extremely bugged i Linux first year when after I got it, could not even work for 10min before PC freeze, but with my little help and others - since kernel 6.2 this integrated GPU started working perfectly fine, I used it as main GPU since then.
Maybe those bugs also can be fixed by driver updates and since they opensource - you can do it.
Think I know why I wouldn't be a good fit for the Vulkanised talks: they have a talk on how to easily implement barriers, but my talk would be on how I never use them at all.
"No transitions, no barriers, no cache flushes, no render passes, no correctness, and no problems" anyone want that talk? haha
@sascha@nicebyte@NOTimothyLottes this why you can make working single triangle for Vuulkan in 100 lines of code in C but it will work only on Nvidia, when AMD require 3000 lines of code to have working single triangle there.
The inventor of blue and white leds got 170$ per patent and got his salary increased to 60k. This was in the 90s and 2000s.
He succeeded by ignoring orders and doing what he thought was right.
He left to a different company and his employer sued him. He counter sued and ended up just getting enough to pay his legal fees.
He seems happier and better appreciated now, which is nice to see. https://youtu.be/AF8d72mA41M?si=R5D7uwdMH4F5zNVX
IMHO this makes a lot of sense, the native WebGPU implementation libraries turned out to contain quite a bit of fat.
TBH looking at Dawn I was rolling around that idea in my head too (WebGPU without the shader compiler isn't all that much code, and I would have moved that offline like in sokol-gfx).
@floooh you can "compile" shaders only for AMD GPU in opensource Linux drivers - means you can make "shader compiler" from language to native GPU machine code - only in AMD GPU in Linux.
Everything else - is not even "java" especially webGPU with 3 layers of translation - WGSL->spirv->gpu driver IR->binary commands
Adding "one more translation" on top of it - is not "compiling" it translation, and translation it is overhead and explode of complexity that lead to more bugs.
Hey Tech Art folks, I have seen many variant of mesh-based outline shader but I haven’t seen one that maintain consistent outline width under different fov and camera distance.
I think we usually guess the width factor based on fov and distance as the correct calculation is too costly. Is there a mathematical sound yet fast solution?
Alternatively I would do the calculation in screen space (fov and distance no longer matters), I wonder if there are equivalent ways that I missed?
And game has 27000 reviews - and if you go to steadb and filter f2p+multiplayer+27k+ reviews:
There only 93 F2P multiplayer games on Steam with more than 27k reviews - and this is 3% of all F2P multiplayer games on Steam.
You can assume that only 3% of f2p multiplayer games being profitable on Steam.