HARDWARE

–

18 February 2025 22h29

When the frontier labs say they need to unlock orders of magnitude more compute to advance AI…. take their word for it. @XAI quickly moving to cluster that houses a million GPUs. (5x bigger than Colossus now.) After #Grok3 wins, Elon Musk is the strongest horse in the race

→ View original post on X — @learnopencv

Apple Intelligence Dictation Feature Works on Latest macOS

By

@learnopencv

–

18 February 2025 19h43

I upgraded to the latest version of macOS and finally I'm seeing some signs of Apple Intelligence. For example, this post is dictated without any manual intervention. So far it looks good.

SpaceX Engineers Master 200k H100 GPU Orchestration in Record Time

By

–

18 February 2025 14h03

Also, @elonmusk has the best engineers in the world at @SpaceX & @Tesla
, that’s a huge advantage! They figured out how to make 200k H100s work together in perfect symphony – from the power, to the cooling to the networking. And in 122 days? A insane (impossible one might have

Compute Power Drives AI Progress: Full-Stack Integration Essential

By

–

18 February 2025 5h53

Astonishing progress in 18 months with brute force compute and brilliant engineering. Compute is king. Players who are full-stack are better positioned.

xAI Secretly Expands Cluster with 100,000 Additional GPUs

By

–

18 February 2025 5h31

WOW @xai secretly added 100,000 GPUs to its cluster!

As I mentioned, compute is king!

PS loving Elon’s international team of engineers, I think I hear German, Chinese and American. https://t.co/oFtyMkAj1q
— Nina Schick (@NinaDSchick) 18 février 2025

WOW @xai secretly added 100,000 GPUs to its cluster! As I mentioned, compute is king! PS loving Elon’s international team of engineers, I think I hear German, Chinese and American.

Accelerating Progress Through Massive Compute Infrastructure

By

–

18 February 2025 5h22

Progressing at ludicrous speed thanks to COMPUTE.

→ View original post on X — @sama

Open Source AI Model: o3-mini vs Phone-Optimized Deployment

By

@sama

–

18 February 2025 2h53

for our next open source project, would it be more useful to do an o3-mini level model that is pretty small but still needs to run on GPUs, or the best phone-sized model we can do?