Cerebras overtakes peers as #1 AI Semiconductor Startup In the freshly-updated 2023 State of AI Report Compute Index from Air Street Capital and Zeta Alpha, Cerebras is highlighted for leading all AI semiconductor startups in research publication counts, open source
AI HARDWARE
-
Neural Operators Mixed Precision Training Achieves 50% Throughput Gain
By
–
Super excited to share our paper on efficient training of Neural Operators with mixed precision, recently accepted at #ICLR2024! We show more than 50% gain in throughput with almost no loss in accuracy. Neural operators are #AI methods for solving #PDE . Unlike traditional
-
Sharing spatial content format for Vision Pro devices
By
–
I wish there was a easy way for you to share this in the spatial format so we could all watch it on our Vision Pro 🙂
-
Resolution Limitations for Professional Work Applications
By
–
I don’t find the resolution sufficient for doing work. Good enough for movies though
-
Vision Pro 1x offers better value than premium business class travel
By
–
Flying the Pacific in business class is $4k minimum. Buying a Vision Pro 1x and flying economy multiple times is a massive life improvement that is far cheaper
-
Nvidia Controls 80% High-End AI Chip Market Worth $1.73 Trillion
By
–
“The Santa Clara, California-based company controls about 80% of high-end AI chip market, a position that has sent its stock market value up 40% so far this year to $1.73 trillion after it more than tripled in 2023.” https://
reuters.com/technology/nvi
dia-chases-30-billion-custom-chip-market-with-new-unit-sources-2024-02-09/
… -
Apple Poised to Dominate Metaverse Market with Vision Pro
By
–
Apple is poised to bring #Metaverse mainstream and dominate the market https://
cointelegraph.com/news/apple-poi
sed-metaverse-mainstream-dominate-vision-pro-augmented-reality-market
… via @cointelegraph -
Strong AI Creation Possible with Budget Laptop Computer
By
–
Someone is going to create strong AI using a $700 laptop computer.
-
GPU Throughput Increase with Multiple Processes Using MPS
By
–
With workloads that don’t fully load the GPU you can often get a substantial throughput increase by running multiple processes on each GPU, but you need to manually enable MPS to get concurrency: https://
docs.nvidia.com/deploy/mps/ind
ex.html
…