SambaNova Sets New AI Performance Record: Processing 1000 Tokens Per Second

2024-05-30

SambaNova Systems achieved a milestone with Llama 3, hitting 1000 tokens/second in generative AI processing speed. This advancement offers faster response times, better hardware utilization, and cost savings for businesses. SambaNova combines hardware like the RDU chip with software like the Samba-1 model to optimize generative AI. Their Reconfigurable Dataflow Architecture enables iterative optimization for efficient resource allocation. Speed milestone reached with 16-bit precision, emphasizin

Read more

Apple may announce collaboration with OpenAI at WWDC

2024-05-27

Apple will reveal its AI strategy at WWDC 2024 with basic functionalities like voice memos transcription and emojis. Possible collaborations with OpenAI, Google, and Anthropic. Introducing "Smart Summary" feature to manage notifications. Enhancements to Voice Memo app, Spotlight search, Siri, and iOS 18 app icon customization.

Read more

Apple iOS 18 and other operating systems AI features may be released in "beta" or "preview" form

2024-05-27

Apple's iOS 18 and other systems' AI features may be released as "beta" or "preview" to reflect cautious AI approach. Newer Apple devices support on-device AI processing for enhanced user experience. AI features cover various aspects and Siri will provide smarter services. Apple faces challenge in balancing user privacy with advanced AI capabilities. Apple's cloud will run on Apple Silicon chips, potentially raising privacy concerns. iOS 18 to include ChatGPT-powered chatbot and continue collabo

Read more

"Domestic GPU and AI Platform Join Hands to Breakthrough: Moore Threads and Wuwen Core Successfully Train 3B Large Model"

2024-05-27

Moor Thread and Wuwen Xinquan announced the successful completion of the 3B-scale model training using domestic GPU Qianka cluster. This milestone validates domestic GPU's strength in large-scale model training and marks a new chapter in deep cooperation between domestic GPU and AI models. The training achieved stability and efficiency, showcasing domestic GPU's capabilities in AI model training. Wuwen Xinquan is committed to creating an intermediate layer product for efficient deployment of lar

Read more

New ETS Conversion Model Enhances Naturalness in Speech Synthesis

2024-05-28

Recent introduction of Diff-ETS by researchers has enhanced ETS technology, enabling more natural synthetic speech for those with communication difficulties. The new model utilizes a diffusion probability model to improve the naturalness of synthesized speech. Compared to traditional ETS models, Diff-ETS produces more human-like voices, opening new communication pathways for individuals unable to speak. Future research will focus on optimizing model parameters for real-time speech generation.

Read more

NVIDIA CEO Jensen Huang's Net Worth Surges to $90 Billion Due to Soaring Demand for AI Chips

2024-05-28

Nvidia CEO Jensen Huang's net worth surged to over $90 billion as the stock price hit a record high. The company's AI processors drove revenue growth exceeding 200%, making Huang one of the world's top 20 richest individuals. Huang's stake increased from $3 billion to $90 billion, with optimism on future demand for Nvidia's GPUs. Huang, a key figure in Nvidia's AI transformation, has propelled the company's growth in various markets.

Read more

AI Intelligent Agents Bring Sci-Fi Novel Plots to Life

2024-05-28

AI agents are the latest trend in tech, being developed by Google, Microsoft, and OpenAI. These agents can perform tasks independently. Challenges include making multiple decisions autonomously. The future holds potential for AI agents interacting with physical tasks. Concerns arise about designing agents too specialized. Divided opinions exist on the impact of AI agents, with some viewing them as empowering tools. Applications include autonomous vehicles. Updates to AI models are needed for bet

Read more

China Telecom Releases Industry-Leading Large-Scale Dialect Speech Recognition Model, Supporting 30 Dialects Mixed Speech

2024-05-27

TeleAI, a research institute of China Telecom, released Starry Sky Multi-Dialect Speech Recognition Model supporting 30 dialects, enhancing language services nationwide. With technical breakthroughs, the model benefits elderly and remote users, ensuring efficient information access. China Telecom aims to expand to 333 dialects, enhancing innovation in intelligent customer service and smart home sectors. The company's growing user base, including 332 million 5G users, supports its leading positio

Read more

Apple WWDC Conference Set to Unveil, New Artificial Intelligence Strategy and Features Awaited

2024-05-27

At WWDC, Apple will unveil new AI features through Project Greymatter, focusing on practical tools for consumers. Functions will run on devices and in the cloud, including voice memo transcription, photo restoration, improved search speed, enhanced Siri, and generative AI for emojis. Apple aims to enhance user experience on iPhone and Apple Watch, but faces competition from Google and Samsung.

Read more

Elon Musk Plans to Build a xAI Supercomputer Factory with 100,000 Chips, Expected to be Operational by 2025

2024-05-27

Elon Musk's xAI plans to build its own supercomputer for next-gen chatbot Grok. It aims to combine data and technology from Musk's other ventures to compete with OpenAI, Anthropic, and Google. The project collaborates with Nvidia and may involve Oracle. Musk also promises a super app for X and a self-driving taxi by year-end. The new funding round brings xAI's valuation to $18 billion. The open-source Grok project aims to democratize AI tech and foster innovation in commercial applications.

Read more