AI Internet of Vehicles

The AI World Today

• French Mistral Unveils Large 2

• Nvidia's New China AI Server

• Chinese Kling AI Goes Global

• Alibaba Open Sources Speech Models

• India’s AI Supercomputer AIRAWAT

• Meet AI-driven Internet of Vehicles

  +

• Heads Up

• Save the Date

• AI Solutions

Mistral's Large 2 Outpaces Competitors in AI

French AI startup, Mistral has unveiled its latest AI model, Large 2, positioning it as a formidable competitor to the recent advancements from OpenAI and Meta. Released on July 24, Large 2 sets a new benchmark in code generation, mathematics, and reasoning while being highly cost-efficient. Mistral's Large 2 boasts 123 billion parameters, significantly outpacing Meta's Llama 3.1 405B in key areas with fewer parameters. With a 128,000 token window, the model can process vast amounts of data in one go, equating to around 300 pages of text. Large 2 supports multiple languages, including English, French, German, Spanish, Italian, Portuguese, Arabic, Hindi, Russian, Chinese, Japanese, and Korean, along with 80 coding languages. Available on platforms such as Google Vertex AI, Amazon Bedrock, Azure AI Studio, and IBM watsonx.ai, Mistral Large 2 can also be accessed on Mistral’s le Plateforme and tested for free on their ChatGPT competitor, le Chat.

Nvidia Designs Compliance-Friendly Server for China

Nvidia has designed a new server for China, dubbed GB20, featuring the B20 chip, in collaboration with Inspur, a key distributor in China. This move ensures compliance with U.S. export restrictions on advanced chips. The B20, part of Nvidia's Blackwell series, performs AI calculations slower than other Blackwell chips, but its performance is bolstered by using multiple chips in the GB20 server. This setup maintains competitiveness against Chinese rivals like Huawei while adhering to U.S. computing power caps. The B20 chip and GB20 server, slated for release next year, aim to balance performance and regulatory compliance. The Blackwell series was initially released in March, with the B20 chip launching in 2024.

Kling AI Video Generator Now Accessable Globally

Kuaishou, a Chinese tech giant, globally launched Kling AI version 1.0 on July 24. This AI video generator, now accessible worldwide at KlingAI.com, offers users 66 free daily credits for video creation upon email registration. The platform supports text-to-video and image-plus-text-to-video generation, challenging OpenAI's Sora. Kling AI can produce videos up to two minutes long in 1080p resolution at 30fps. Users can choose from 960x960, 720p landscape, and 720p portrait modes, with options for various camera movements. Multilingual support includes English and Mandarin. Future updates will enhance features, including a high-quality mode and extended generation times. Content restrictions apply, preventing violent, NSFW, or famous person depictions.

Alibaba's Tongyi Team Launches Open-Source Models

Alibaba AI Speech's Tongyi Speech Team has open-sourced two foundational speech models: SenseVoice and CosyVoice. These multilingual models offer advanced capabilities in speech recognition and generation. SenseVoice excels in multilingual audio understanding, outperforming Whisper by 50% in Chinese and Cantonese speech recognition, with an inference speed 15 times faster. It also supports state-of-the-art emotion recognition and audio event detection. CosyVoice, trained on over 170,000 hours of multilingual audio data, is proficient in multilingual speech generation, timbre, and emotion control. It enables zero-shot speech generation, cross-lingual voice synthesis, and instruction execution.

AIRAWAT Supercomputer Elevates India’s AI Capabilities

The AI supercomputer AIRAWAT, developed by C-DAC and installed in Pune, India, has made significant strides in the AI landscape by ranking 75th in the Top 500 Global Supercomputing List. This achievement positions India among the leading AI supercomputing nations globally. AIRAWAT is part of India's National AI Programme and features a peak compute capacity of 200 AI Petaflops, integrated with PARAM Siddhi for a total of 410 AI Petaflops. The supercomputer supports AI research, analytics, and knowledge dissemination, empowering academia, research labs, and industries to develop indigenous AI solutions. This infrastructure aims to tackle complex real-life problems and boost India’s AI capabilities while supporting the government's vision of "AI for All"

China Unicom Launches AI-Powered Vehicle Solutions

China Unicom has launched three AI-driven Internet of Vehicles products: ZhiYu, ZhiLue, and ZhiTu. These were unveiled at a dedicated forum during the company’s partner conference in Shanghai.ZhiYu focuses on optimizing connectivity resources, enhancing the efficiency of vehicle communication networks. ZhiLue uses multimodal interfaces to improve operational efficiency through industry data analysis. ZhiTu provides integrated perception for intelligent mobility, enhancing the autonomous driving experience. Additionally, the China Unicom Vehicle Life Index supports automakers across various stages from research to service by analyzing 36 user life cycle metrics. Strategic partnerships, including an agreement with Huawei, were announced to bolster AI-powered vehicle solutions, aiming to transform the automotive and transportation sectors through advanced AI and connectivity technologies.

Heads Up

AI Safety: OpenAI announced a new way to teach AI models to align with safety policies called Rules Based Rewards 

AI Search: Microsoft Introduces Bing generative search

AI Video: Stability AI steps into a new gen AI dimension with Stable Video 4D

AI Funding: US-based AI-powered customer intelligence platform secures $13M funding

AI Robotics: A US company is making AI robots to do America's toughest jobs

AI Jobs: A new survey shows Canadian graduates optimistic about finance careers amid rising AI and tech challenges

AI Grants: NIST Announces Funding Opportunity for AI-Focused Manufacturing USA Institute

AI Media: India’s HT Media uses AI to boost traffic and editorial efficiency

AI Funding: Chinese AI Startup Baichuan Bags USD690.1 Million in Series A Fundraiser

AI Invest: Saudi’s Wa’ed Ventures invests $15 M in South Korean chipmaker

AI Medical: Qatar adopts AI-driven solution for breast cancer screening programme

Save the Date

September 10-11: Qatar Hosts Wold AI Summit

AI Solutions

RoboflowSports Enhances Real-Time Sports Analysis with AI

RoboflowSports has launched advanced motion tracking technology to revolutionize sports analysis. By utilizing precise object detection and image segmentation, this technology enables real-time analysis of competitions, offering coaches and analysts deeper insights into game situations and athlete performance. Key functions of RoboflowSports include football player and ball detection, which allows for accurate tracking of player movements and ball positions. Image segmentation further refines the analysis by distinguishing different objects within an image, such as athletes and footballs. Additionally, key point detection identifies critical points on athletes' bodies and the football field, aiding in posture estimation, motion analysis, and tactical planning. The technology also features visualization tools that generate annotated images and videos, enhancing the comprehensiveness and clarity of the analysis.