tostane

joined 1 month ago
 

Honest Limitations Though

TurboQuant was designed for attention score accuracy, not perceptual image quality — PSNR and SSIM care about different things than transformer attention distributions The "zero accuracy loss" claim is specific to its use case; adapted to video you'd need to re-validate quality metrics Motion estimation (the biggest win in video codecs) isn't addressed at all by TurboQuant — you'd still need that separately Real-time decoding constraints are much stricter in video than in LLM inference

Bottom Line It probably wouldn't replace a traditional codec on its own, but the PolarQuant + QJL two-stage approach applied to neural codec latents is a genuinely novel research direction that doesn't seem to have been explored yet. If you combined it with solid motion estimation, you could have something interesting — especially for scenarios where you want very aggressive compression (3-4 bits) with provable quality bounds.

[–] tostane@thelemmy.club 10 points 2 weeks ago (1 children)

iran seem smart they seen trump say the same lies with ukrane then russia would attack ukrane hard. like who side is trump on.

[–] tostane@thelemmy.club 7 points 3 weeks ago (1 children)

lets hope the Devil stops controlling him. And all the other nations can forgive those who voted this nightmare on the world. he wanted a nobel prize lets give him a trophy for starting Armageddon

[–] tostane@thelemmy.club -4 points 1 month ago (1 children)

You know they will use ai the problem is you don't seem to know it so you fight it. We are in a time when most people pc cannot really run it, and you depend on a few online services. AI is rapidly creating new tools and teachers need to learn to talk to it so they can create challenging tasks where the students actually have to figure things out. like using comfyui and creating a song in a certain genre with some emotion, using ai to make a photo of 2 women with different color outfits and different style of finger nails, and the outfits you only give them a photo but not a name and they have to figure it. ai is not easy if you actually try to create something worth creating. students in china are learning to use it at 5 years old.

[–] tostane@thelemmy.club 8 points 1 month ago

Amazon, gemini, perplexity are also in the us goverment baby bottle nipples.