AI News, Research and Tool Reviews
-
H200 memory bandwidth inefficiencies in distributed training
Improper NUMA settings and hardware selection can waste the 4.8 TB/s memory bandwidth of the NVIDIA H200. Communication errors with the Power Management Unit cause job failures 97%…
-
Gemini Enterprise updates and Salesforce integration expansion
Gemini Enterprise introduces agentic workflows and in-window editing to automate complex business processes. New integrations with Salesforce Agentforce and Google Cloud enable…
-
Sakana AI optimizes model performance through evolutionary merging
Sakana AI utilizes evolutionary techniques to combine open-source models, creating specialized systems like the EvoLLM-JP-v1-7B which outperforms larger 70B parameter models. The…
-
The capital and intelligence race in physical AI
Physical Intelligence negotiates an $11 billion valuation as Figure AI deploys its Figure 02 robot to handle 90,000 components for BMW at…
-
The cost of scaling AI interconnects
Marvell’s $3.25 billion acquisition of Celestial AI targets the 16 Tbps Photonic Fabric platform to scale XPU connections. This optical interconnect offers…
-
Liquid AI LFM-3B edge deployment on Apple Silicon
Liquid AI utilizes MIT-inspired liquid neural networks to deliver efficient edge computing performance. The Cactus engine achieves 27 tokens per second decode…
-
Character.AI faces lawsuits and infrastructure shifts
Character.AI faces intense legal scrutiny following the death of 14-year-old Sewell Setzer III and regulatory action from the Pennsylvania Attorney General. The…
-
Deploying Mixtral 8x22B on Fireworks AI
Fireworks AI provides serverless GPU inference for the Mixtral 8x22B model, utilizing a sparse Mixture-of-Experts architecture with 141 billion parameters. Developers can…
-
The evolution of OpenAI from research lab to $852 billion entity
OpenAI transitioned from a 2015 nonprofit research lab to an $852 billion entity through a strategic compute-for-equity model. The 2019 Microsoft partnership…