HackerTrans
TopNewTrendsCommentsPastAskShowJobs

MasterScrat

2,633 karmajoined il y a 16 ans
General Intuition

florian dot laurent at gmail

Submissions

Show HN: MIRA – Multiplayer World Model Trained on Rocket League

mira-wm.com
12 points·by MasterScrat·il y a 4 jours·1 comments

Meta's AI Storage Blueprint at Scale

engineering.fb.com
1 points·by MasterScrat·il y a 6 jours·0 comments

Machines do not press play: From temporal compression to model-ready tensors

spiraldb.com
1 points·by MasterScrat·il y a 24 jours·0 comments

The Origins of GPU Computing

cacm.acm.org
1 points·by MasterScrat·il y a 3 mois·0 comments

Poolside hunts data centre partners after CoreWeave deal falls through

ft.com
1 points·by MasterScrat·il y a 3 mois·1 comments

Keeping 20k GPUs Healthy

modal.com
2 points·by MasterScrat·il y a 6 mois·0 comments

ClusterMAX 2.0: The Industry Standard GPU Cloud Rating System

newsletter.semianalysis.com
2 points·by MasterScrat·il y a 8 mois·0 comments

General Intuition $134M seed to teach agents spatial reasoning using game clips

techcrunch.com
2 points·by MasterScrat·il y a 9 mois·0 comments

Project Horizon: Why we're building a 2 gigawatt AI campus in Texas

poolside.ai
3 points·by MasterScrat·il y a 9 mois·1 comments

comments

MasterScrat
·il y a 11 heures·discuss
Playing backward is actually a great idea for testing generalization
MasterScrat
·il y a 3 jours·discuss
Would be a great idea to see how much we could manipulate the latent space and whether it has some internal structure w.r.t the physical state. I guess the only unknown is how the world model would show robustness to latent states that are transformed through this network
MasterScrat
·il y a 3 jours·discuss
Hey all, happy to see this here! This was a colab between General Intuition (that I’m part of), Kyutai and Epic Games.

You can read plenty of details in the blog post and tech report but the TLDR is that we trained a multiplayer world model on 10k hours of Rocket League data. We optimized it to be playable at 20fps on a single GPU.

So what you see in the demo is fully generated: there’s no graphics or physics engine. Instead it’s a 5b neural network that takes actions in and gives pixels out.
MasterScrat
·il y a 4 jours·discuss
We're happy to release MIRA, a collaboration between General Intuition, Kyutai, and Epic Games.

Mira was trained on 10k hours of Rocket League data. The model has 5B parameters and runs 4-player games at 20 fps on a single B200 GPU.

We've released a playable online demo, an in-depth technical report as well as a 1k hour dataset of 4-players gameplay:

Technical report: https://mira-wm.com/paper Repo: https://github.com/mira-wm/mira
MasterScrat
·il y a 4 jours·discuss
The checkpoint also weights half less than the game install! ;-)
MasterScrat
·il y a 14 jours·discuss
This sounds like a strong statement with little backing. The author does infra at DeepSeek if his LinkedIn is to be trusted, and is the author of Foyer.
MasterScrat
·le mois dernier·discuss
This is exactly what i would have needed 6 months ago
MasterScrat
·il y a 2 mois·discuss
I have mixed feelings (as in, I'm unsure how to feel) about projects where the code, the README and the HN/Reddit posts are mostly AI-generated.

I feel the frustration of reading "slop", but on the other hand the projects that surface do usually bring something useful to the table.

Should we simply judge the submission based on its technical merit? Why do I feel annoyed that an otherwise cool project uses typical LLM prose? For how long will we be able to recognize LLM-generated text, and what happens when we can't?
MasterScrat
·il y a 3 mois·discuss
and so does OpenAI
MasterScrat
·il y a 3 mois·discuss
Clearly they need more datacenters!
MasterScrat
·il y a 3 mois·discuss
Without paywall: https://archive.is/u7t8U
MasterScrat
·il y a 4 mois·discuss
> As we focus and compute demand grows, the Sora research team continues to focus on world simulation research to advance robotics that will help people solve real-world, physical tasks.

— https://www.businessinsider.com/openai-discontinues-sora-vid...

So yeah, focusing on world models
MasterScrat
·il y a 4 mois·discuss
This has a simple pragmatic solution though:

https://duckdb.org/2024/12/06/duckdb-tpch-sf100-on-mobile#a-...
MasterScrat
·il y a 4 mois·discuss
Are you saying Carmack isn't a "talented and successful developer"?
MasterScrat
·il y a 4 mois·discuss
Why would a game development pedigree correlate with rejecting AI? As Carmack said:

> AI tools will allow the best to reach even greater heights, while enabling smaller teams to accomplish more, and bring in some completely new creator demographics.
MasterScrat
·il y a 6 mois·discuss
> Have you ever noticed that the food graphics in Super Smash Bros. and Kirby Air Riders is flat “billboarded” stock images of food?

I have not... A video showing what it's talking about would have been a good addition to this article.
MasterScrat
·il y a 6 mois·discuss
Is there a reliable service that plots hourly price per GPU per cloud through time?
MasterScrat
·il y a 6 mois·discuss
Just so you know, this is already very much a thing on TikTok: AI-generated movie summaries with narrator voice explaining the plot while showing only major beats, reducing movie from 2h to shorts totaling 10min.

It’s honestly not the worse AI content out there! Lots of movies I wouldn’t consider watching but that I’m curious enough to see summarized (eg a movie where only the first title was good but two more were still published)
MasterScrat
·il y a 7 mois·discuss
Exactly - best case scenario is Reddit/HN front page with a cool project you enjoyed working on, have some nice conversations there, reach a few 100s stars which look good on your CV, and that’s it.

If you expect more long term support you better be paying me for my time.
MasterScrat
·il y a 7 mois·discuss
> Without world models. you cannot achieve reliability. And without reliability, profits are limited.

Surprising to simultaneously announce the end of the road yet point to the road ahead