Meta AI News - Page 2 of 13
241 articles • Page 2 of 13
OpenAI and Foxconn to co-design multiple generations of AI data-centre racks
Forget the algorithms. This fight is about sheet metal, screws, and the industrial hum of a million server fans.
Meta unveils open-source brain AI, adds Scrunch site audit and Suno v5.5
Meta is pushing the frontier of AI transparency with an open-source brain model, while Scrunch offers a free audit to show how AI sees your website,...
Raspberry Pi 5 Gets USD 130 AI HAT with 8GB RAM for Machine Learning Workloads
For $130, Raspberry Pi now sells an AI co-processor. This new HAT+ 2 board packs 8GB of RAM specifically for running local generative AI models like...
Meta Eyes Google TPUs, Up to One Million Units, as NVIDIA Alternative
Mark Zuckerberg is about to spend a breathtaking sum of money. Meta's capital expenditure is projected to blast past $100 billion in 2026.
Tech CEOs say AI could let them operate everywhere, Dorsey touts new layer
Tech executives Jack Dorsey and Mark Zuckerberg are separately pushing an idea that once seemed impossible: using artificial intelligence to be in...
Meta's Muse Spark 1.1 coding score hits 71.3, edges past GLM-5.2
Meta pushed out an update to its Muse Spark model on Thursday, and the numbers put it ahead of Zhipu's GLM-5.2 on coding tasks for the first time.
Bolmo architecture enables efficient byte-level LM training, easing AI
The byte is the most universal unit of data. Every character, every emoji, every garbled string from a noisy keyboard, it all collapses down to...
Meta adds Spotify AI music, Kannada/Telugu, and noise filtering to AI Glasses
Meta's AI glasses are finally getting interesting. The latest update isn't about more of the same.
SSD removes sync bottleneck in speculative decoding on MI300X
Speculative decoding could speed up large language models, but a synchronization bottleneck limited its gains.
Cerebras Leads Top 5 Fast LLM APIs with Low Latency, High Token Rate
Speed isn’t just a feature, it’s the architecture. Cerebras has engineered its inference stack around a single, uncompromising priority: raw token...
Databricks Instructed Retriever outperforms traditional RAG by 70%
That seventy percent figure is a trap. It’s meant to dazzle. The real takeaway from Databricks’ new retrieval model is simpler: most companies have...
Meta to pilot premium subscriptions for Instagram, Facebook, WhatsApp
Meta is finally making the turn. The company has spent the last few years throwing billions at AI.
Contextual AI unveils Agent Composer, hits top FACTS benchmark score
The new standard for AI trustworthiness just got a hard reset. Contextual AI didn’t just beat Google’s FACTS benchmark, they topped it, proving that...
Docker Trick: Deterministic OS Packages in One Layer to Prevent ML Failures
Your machine learning pipeline runs perfectly on your laptop. On the cluster, it crashes with a cryptic `libgomp` error.
ChatGPT Images 2.0 and Nano Banana 2 Produce Professional Results
Two AI image generators, ChatGPT Images 2.0 and Nano Banana 2, are now producing marketing banners that don't look amateurish.
MetaClaw trains AI agents via Google Calendar, turning failures into rules
Your AI agent just failed. That mistake, a mangled time format, a deleted file without a backup, a naming convention ignored, won’t stay buried in a...
OpenClaw and NVIDIA NemoClaw Enable Secure Local AI Agent via Ollama
Most local AI agents are either hopelessly exposed or just don't work. They promise autonomy but deliver a brittle script that needs constant...
Spatial priming beats semantic prompting in chart data extraction study
What works better when asking a multimodal model to read a chart: high-level reasoning guidance, or a simple grid drawn over the image?
China Orders ByteDance, Alibaba to Shut Down Custom AI Chatbots
ByteDance's Doubao chatbot has more than 300 million monthly users, making it the most popular AI assistant in China.
ChatGPT's 'Nerdy' tweak rewards goblin metaphors in answers, study finds
ChatGPT briefly, bizarrely, fell for goblins. The whole fling started with a single broken piece of code.