State Grid Corp of China launched the "Bright Power Mega Model," a trillion-parameter AI model for the power industry, excelling in data integration, analysis, and professional capabilities. It has been widely applied in grid planning, operation, and customer service, enhancing productivity and intelligence in the sector.

Google has introduced Gemini 2.0 Flash Thinking Experimental, an AI model focusing on reasoning, available on AI Studio. It excels in multi-modal understanding and complex problem-solving but is still experimental and requires more time for better results. Other companies like DeepSeek and Alibaba have also launched similar models.

GitHub launched a free Copilot subscription, allowing 150 million developers to use AI features in VS Code, including code generation and chat assistance, up to 2,000 times monthly.

OpenAI announced new integrations for the ChatGPT desktop app, including support for Apple Notes, Notion, Quip, and multiple IDEs. The update enhances productivity and programming workflows, with a focus on user privacy. OpenAI plans to share more about AGI and new agents in 2025.

ByteDance launched the Doupao Visual Understanding Model at the Volcano Engine Force Conference, offering cost-effective multimodal capabilities at 3 cents per thousand tokens, 85% cheaper than the industry average. The Doupao 3D Generation Model and other updated models like the General Model Pro, Music Model, and Text-to-Image Model 2.1 were also introduced, enhancing AI technology accessibility and digital content creation.

IBM launched Granite 3.1, an upgraded LLM with 128K token context, new embedding model, and hallucination detection, outperforming competitors in benchmarks. It aims to enhance enterprise AI with smaller, more efficient models.

NVIDIA's new Jetson Orin Nano Super Developer Kit offers a 70% performance boost, advanced AI capabilities, and is priced at $249. It features an NVIDIA Ampere GPU, 6-core Arm CPU, 67 INT8 TOPS, 102GB/s memory bandwidth, and supports up to four cameras.

Tech giants like Microsoft, Meta, and Google are competing in AI. Microsoft is the largest buyer of NVIDIA AI chips, purchasing 485,000 Hopper chips, 20% of NVIDIA's 2023 revenue. NVIDIA's GPUs, essential for AI, face high demand, boosting its market value to over $3 trillion. Microsoft invests heavily in AI, including a $100 billion Stargate project with OpenAI and developing its own Maia chip.

Apple and NVIDIA collaborated to integrate Apple's ReDrafter technology into NVIDIA's TensorRT-LLM, boosting token generation speed by 2.7x and improving inference efficiency for large language models.

PromptWizard, an AI framework by Microsoft India Research, optimizes prompts for black-box LLMs using a feedback-driven mechanism. It improves task performance through iterative enhancement and structured critique, outperforming existing methods in 45 tasks with reduced computational overhead and cost.

Odyssey, founded by Oliver Cameron and Jeff Hawke, develops Explorer, an AI tool converting text or images into 3D renderings. Trained with real-world data, it excels in creating realistic scenes for use in Unreal Engine, Blender, and Adobe After Effects. Despite current limitations, it's available to production companies and independent artists. Odyssey aims to support, not replace, creative professionals, with Ed Catmull joining its board.

Alibaba's CosyVoice 2 enhances speech synthesis with unified streaming and non-streaming modes, improved pronunciation accuracy, and advanced instruction control. It features low latency, high speaker consistency, and multilingual capabilities, making it ideal for real-time applications.

GitHub is releasing a free version of its Copilot AI coding tool, integrated into VS Code. It aims to expand access, especially for students and occasional users, with limitations like 2,000 code completions per month.

Microsoft purchased 485,000 Nvidia Hopper chips in 2024, more than doubling Meta's acquisition. This surge, along with developing its own AI chips and deepening partnerships, underscores Microsoft's commitment to AI and high-performance computing.

Samsung's Unpacked event on January 22nd is expected to unveil the Galaxy S25 series and provide more details on the "Project Moohan" XR headset, with anticipated upgrades in performance, camera, and design.

OpenAI launched a new service allowing users to interact with ChatGPT via phone or WhatsApp. U.S. users get 15 minutes of free monthly voice calls, while global users can text on WhatsApp with daily limits. The service aims to make AI more accessible, especially for those without high-speed internet or modern smartphones. Interactions are limited to text, and data is temporarily stored for security.

WeChat introduces "Author's Reading Voice" allowing users to listen to articles in the author's voice, enhancing interactivity. Authors use the "Public Account Assistant" app to record and apply their voices. Another feature, "AI Image Matching," recommends images based on content, improving visual appeal. Both features aim to enhance user experience.

Meta's Ray-Ban smart glasses will get a new update, integrating Shazam for song recognition and adding live AI and real-time translation features.

Intel faces pressure but sees potential in its Battlemage GPU, which has received positive reception and high demand. The company plans to release new batches of the Arc B580 Limited Edition weekly, aiming to maintain market presence despite competition from Nvidia and AMD.

OpenAI is developing a new feature called Tasks, allowing users to schedule and repeat prompts for ChatGPT. Set to launch in beta on its 12th anniversary, it includes an intuitive UI for managing tasks. Google is also working on a similar feature for Gemini, called Scheduled Prompts.