STMicroelectronics to launch data centre photonics chip developed with Amazon | Reuters https://www.reuters.com/technology/artificial-intelligence/stmicroelectronics-launch-data-centre-photonics-chip-developed-with-amazon-2025-02-20/
“Openrouter is now supported in ai-gradio in a few lines of code you can use deepseek-r1, claude, gemini and more with coder mode pip install –upgrade “ai-gradio[openrouter]” import gradio as gr import ai_gradio gr.load( name=’openrouter:anthropic/claude-3.5-sonnet”, https://x.com/_akhaliq/status/1890543241017405695
“To wit: DeepSeek, Tencent, Qihoo 360 (cybersecurity), Xiaomi, Will Semiconductor, BYD, Huawei, New Hope (agriculture), Unitree, CHINT (electrics), Alibaba (Jack Ma!), CATL. This is the lineup of crucial technology leaders. Our bro Wenfeng is the sole representative of AI wing. https://x.com/teortaxesTex/status/1891446214236807451
Together AI Announces $305M Series B to Scale AI Acceleration Cloud for Open Source and Enterprise AI https://www.together.ai/blog/together-ai-announcing-305m-series-b
“Based on the early stats, looks like Grok 3 base is going to be a very solid frontier model (leads Chatbot Arena), suggesting pre-training scaling law continues with linear improvements to 10x compute No Reasoner, yet (one is coming?) so GPQA scores are still below o3-mini (77%) https://x.com/emollick/status/1891707120879345788
“We’re excited to receive our first #NVIDIADGX B200 system which we’ll use for vLLM research and development! Thank you @nvidia! https://x.com/vllm_project/status/1893001644037566610
[2502.11089] Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention https://arxiv.org/abs/2502.11089
“🚀 Introducing NSA: A Hardware-Aligned and Natively Trainable Sparse Attention mechanism for ultra-fast long-context training & inference! Core components of NSA: • Dynamic hierarchical sparse strategy • Coarse-grained token compression • Fine-grained token selection 💡 With https://x.com/deepseek_ai/status/1891745487071609327
“Can frontier models cost-effectively accelerate ML workloads via optimizing GPU kernels? Our take at METR: yes, and they’re improving pretty steeply – but it’s easy to miss these capabilities without good elicitation and “fair” compute spend. https://x.com/METR_Evals/status/1890531685495685382
“🏎️ Test Drive NVIDIA Blackwell GPUs with Together GPU Clusters featuring Together Kernel Collection! 🚀 Eight AI teams get FREE access to NVIDIA HGX B200 nodes on Together GPU Clusters! 🤖 Collaborate with NVIDIA & Together AI experts to optimize your models and accelerate https://x.com/togethercompute/status/1892256276576686131
“Llamba: Scaling Distilled Recurrent Models for Efficient Language Processing “We introduce Llamba, a family of efficient recurrent language models distilled from Llama-3.x into the Mamba architecture. The series includes Llamba-1B, Llamba-3B, and Llamba-8B, which achieve higher https://x.com/iScienceLuvr/status/1892875837772615839
“The significance of Grok 3, outside of X drama, is that it is the first full model release that we definitely know is at least an order of magnitude larger than GPT-4 class models in training compute, so it will help us understand whether 1st scaling law (pre-training) holds up.” / X https://x.com/emollick/status/1890982179355639881
“LETS GOOOO – Hugging Face just released Ultra Scale Playbook for Training LLMs on GPU Clusters! 🤯 A free, open-source, book to learn everything about 5D parallelism, ZeRO, fast CUDA kernels, how and why overlap compute & communication – all scaling bottlenecks and tools https://x.com/reach_vb/status/1892276287039033473
“I think Grok 3 came in right at expectations, so I don’t think there is much to update in terms of consensus projections on AI: still accelerating development, speed is a moat, compute still matters, no obvious secret sauce to making a frontier model if you have talent & chips.” / X https://x.com/emollick/status/1891749764212900242
“Grok 3 also excels at creative coding like generating creative and novel games. Elon emphasized Grok 3’s creative emergent capabilities. You can also use the Big Brain mode to use more compute and reasoning with Grok 3. https://x.com/omarsar0/status/1891709371802910967
“the grok 3 release made me sad. something fatalistic about falling back to bruteforce scaling — 100x more compute than R1 for a model that’s at most 10% better all that time, money, and electricity spent on a system that will be obsolete before my semester ends AI needs new” / X https://x.com/jxmnop/status/1892725541796446350
AI Platform for Teaching American Sign Language | NVIDIA Blog https://blogs.nvidia.com/blog/ai-sign-language/
“Github 👨🔧: the LLM vulnerability scanner from @nvidia → LLM vulnerability scanning and security assessment → Comprehensive red-teaming framework for generative AI → Extensible plugin architecture for probes and detectors It supports multiple vulnerability checks: 🔹 https://x.com/rohanpaul_ai/status/1890551385680163284
“A team at @deepseek_ai plans to open-source 5 repositories next week, one per day. Focused on infrastructure and building blocks of their online services. https://x.com/_philschmid/status/1892857906669715779
“Join us, @gokoyeb and @tenstorrent for an exclusive night of eye-opening talks on Wednesday, March 5! Learn about: 🧠 Cutting-edge innovations in AI infrastructure 🔧 Practical applications for training, fine-tuning, inference, and RAG 🚀 Superior performance at lower cost https://x.com/llama_index/status/1893012260785987667
“What is Mixture-of-Mamba (MoM)? MoM expands Mixture-of-Experts (MoE) concept on State Space Models SSMs). This development brings a new architecture that can handle all modalities by applying modality-aware sparsity inside the core of the Mamba block. ▪️ What is this https://x.com/TheTuringPost/status/1892695756290834941
“Total GPUs: 200K The capacity was doubled in 92 days! All of this compute was used to improve Grok — which has lead to Grok 3. https://x.com/omarsar0/status/1891705593125105936
“Scaling Test-Time Compute Without Verification or RL is Suboptimal “In this paper, we prove that finetuning LLMs with verifier-based (VB) methods based on RL or search is far superior to verifier-free (VF) approaches based on distilling or cloning search traces, given a fixed https://x.com/iScienceLuvr/status/1891839822257586310
“NVIDIA + Arc Institute’s new model Evo 2 just demonstrated that deep learning can directly model biological function It stands as a breakthrough in computational biology, 🧵 1/n Evo 2 just redefined genomic modeling by processing over 9 trillion nucleotides to seamlessly https://x.com/rohanpaul_ai/status/1892383673887985738
Microsoft’s Majorana 1 chip carves new path for quantum computing – Source https://news.microsoft.com/source/features/ai/microsofts-majorana-1-chip-carves-new-path-for-quantum-computing/
“grok-3 is 8e26 FLOPs of training compute” / X https://x.com/ethanCaballero/status/1891712442893312151
Together AI Achieves 90% Faster BF16 Training with NVIDIA Blackwell Platform and Together Kernel Collection https://www.together.ai/blog/nvidia-hgx-b200-with-together-kernel-collection
“> Efficient Triton implementations for Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention. Not sure how this addresses @main_horse’s doubts about implementation details, but great to have this variant” / X https://x.com/teortaxesTex/status/1893019673043566670
“Update: Combining evolutionary optimization with LLMs is powerful but can also find ways to trick the verification sandbox. We are fortunate to have readers, like @main_horse test our CUDA kernels, to identify that the system had found a way to “cheat”. For example, the system” / X https://x.com/SakanaAILabs/status/1892992938013270019
“chat, google just dropped apache 2.0 licensed, multilingual vision encoder, SigLIP 2! 🔥 drop-in replacement to SigLIP 1, works w/ transformers! https://x.com/reach_vb/status/1892870777197764703
“The AI-native (edge and LLM) proxy for agents. Move faster by letting Arch handle all the pesky heavy lifting in securing, processing, routing, and tracing prompts. Built by the contributors of Envoy. Key features include: 🛡️ Guardrails at the edge: reject jailbreak attempts https://x.com/_akhaliq/status/1891289618978316749
HadaCore: Tensor Core Accelerated Hadamard Transform Kernel | PyTorch https://pytorch.org/blog/hadacore/
Lambda Raises $480M to Expand AI Cloud Platform https://lambdalabs.com/blog/lambda-raises-480m-to-expand-ai-cloud-platform
“on-demand H100 for $0.99/hr, 4090 for $0.20/hr at Hyperbolic likely the cheapest GPUs around tell me what you’re building, and I’ll spot you free credits for an 8xH100 node for at least a few hours to start. https://x.com/Yuchenj_UW/status/1892990427139318007
“We’re building the fastest & most efficient AI download & upload platform to accelerate AI development. Great progress from the Xethub team @huggingface! https://x.com/ClementDelangue/status/1890416797007900738
“SemiAnalysis is hosting Blackwell & low level GPU Hackathon 🚀 Hacking, prizes, and insights from top industry leaders like @cHHillee, @tri_dao, @marksaroufim, Phil Tillet from TogetherAI, GPUMode, OpenAI, Coreweave, Lambda, etc Limited spots, apply here! https://x.com/dylan522p/status/1893026079931277636
