Samsung launched Galaxy S25 series at Unpacked 2025. Google updated Android, enhancing AI assistant Gemini with deeper app integration and new features like speech practice and multimodal interactions. Gemini now supports commands across apps and will soon include Astra's screen sharing and live streaming on Galaxy S25 devices.

Samsung's SmartThings platform will introduce AI features like ambient sensing and generative AI to enhance daily living. Using mmWave and sound sensors, it can detect activities and adjust home settings accordingly. Generative AI enables personalized configurations and 3D mapping for easier device control. All data remains local for privacy. Updates roll out between 2025 and 2026.

Samsung's upcoming Galaxy S25 series will support the C2PA's Content Credentials standard for identifying AI generated content aiding in the fight against misinformation starting with images and expanding to other media Samsung joins Adobe and other tech giants in this initiative aiming to provide transparency in digital content creation

Samsung launched the Galaxy S25 series featuring three models with advanced AI capabilities, custom Snapdragon 8 Elite platform, and One UI 7 OS. Displays range from 6.2-inch to 6.9-inch QHD+ with 120Hz refresh rate. Camera setups vary, with Ultra model offering a quad-camera array including a 200MP lens. Storage options range from 128GB to 1TB, and prices start at $800, $1000, and $1300 for S25, S25 Plus, and S25 Ultra respectively.

Researchers from App-In Club developed an affordable robotic handwriting system using Raspberry Pi Pico, 3D-printed components, and TensorFlow.js. It replicates human handwriting with ±0.3 mm accuracy at 200 mm/min, costing about $56, significantly cheaper than existing systems.

Deepseek R1, an open-source language model, excels in coding, reasoning, and math, outperforming proprietary models like GPT-3.5 with a 97% success rate in coding tasks. It offers advanced features such as functional code generation and debugging support, along with customizable architecture for various hardware platforms. Future developments include more efficient versions to enhance accessibility and performance. Its transparency and flexibility make it ideal for developers and researchers see

Netease Youdao launches Zi Yue-O1, a 14B parameter reasoning model offering step-by-step explanations, optimized for Chinese users and deployable on consumer-grade GPUs. It aims to provide an accessible tool for AI education.

ByteDance releases DouBao 1.5 Pro, excelling in knowledge, coding, reasoning, and Chinese processing, surpassing GPT-4o and Claude 3.5 Sonnet. Features include innovative pre-training, MoE architecture, and enhanced visual and speech models. Available via API on Volcano Engine, ensuring independence and originality without price increases.

SoftBank, OpenAI, Oracle, and MGX announced a $500 billion AI infrastructure investment for OpenAI in the US over four years. Initial partners include Arm, Microsoft, Nvidia, Oracle, and OpenAI in Texas. Microsoft extends partnership with OpenAI until 2030, including IP access and revenue sharing. Project gains attention, previously speculated since last year. Note: Verify Jack Ma's involvement.

Mistral AI CEO Arthur Mensch reveals plans for an IPO, denies sale intentions, and announces expansion into Singapore to focus on Asia-Pacific while growing in Europe and US. Founded in 2023, Mistral has raised $1.14B and is valued at $6B, competing with OpenAI through products like Le Chat.

Perplexity launches Sonar API for integrating generative AI search into apps. Sonar connects to reliable web sources in real-time, offering standard and Pro versions. Pro provides more detailed responses and twice the references. Sonar's pricing starts at $5 per 1000 queries with varying costs for input and output tokens. In SimpleQA benchmark tests, Sonar Pro outperformed top models from Google, OpenAI, and Anthropic. The company has an ARR of $5-10M and recently raised $73.6M at a $520M valuat

Microsoft and OpenAI revised their partnership, allowing OpenAI to use other suppliers' computing resources. Microsoft retains priority purchasing rights and exclusive access to OpenAI's tech for products like Copilot until 2030. OpenAI will receive expanded computing capacity mainly for model development. Despite tensions and shared revenues, with Microsoft getting 20% of OpenAI's income, both companies plan to continue collaboration. OpenAI is considering removing a clause that could restrict

Anthropic CEO Dario Amodei announced new two-way voice mode and memory function for chatbot Claude to enhance user experience and personalization. Facing growing demand and computational limits, Anthropic seeks solutions and plans advanced AI models. Despite raising $13.7B, facing losses, negotiating to raise $2B more at $60B valuation to compete with OpenAI.

Grok AI, developed by xAI under Elon Musk, stands out with its bold personality and real-time awareness via X platform integration. It excels in various tasks but is limited to X's ecosystem, raising privacy concerns. Despite superior performance, its unconventional tone makes it less ideal for all business contexts.

Google plans to release Gemini 2.0 Flash Thinking Exp-0123 on January 23, 2025, an upgrade over Exp-1219 with enhanced reasoning capabilities for quicker decision-making and dynamic adaptability. Part of Google's AI Studio platform, it integrates large language models with multimodal features like image and video understanding, aiming to optimize AI tools for broader applications. Initial internal tests suggest imminent public availability.

OpenAI is developing a memory-integrated search feature for ChatGPT, enhancing personalized responses based on user interactions. This functionality, akin to web cookies, allows precise, context-aware replies and raises discussions on privacy and targeted advertising. The new feature aims to improve user interaction by leveraging stored data for tailored content. Key points: Memory-search integration, personalized responses, privacy concerns, enhanced user experience.

At a post-inauguration rally, President Trump signed executive orders revoking 78 from the Biden administration, including AI guidelines. These actions aim to repair national institutions and the economy but lack detailed explanations. Biden's orders had focused on establishing AI safeguards, promoting protocols, and mitigating AI's adverse impacts on workers.

OpenAI is set to launch Operator, an AI tool assisting with various computer tasks. Evidence includes hidden shortcuts in macOS ChatGPT client and performance comparisons on OpenAI's site. Operator shows mixed results in benchmarks, excelling in some areas but struggling with basic tasks like cloud service registration. Safety concerns are highlighted, with Operator performing well in security tests. The AI agent market is projected to reach $47.1 billion by 2030.

DeepSeek released R1 series LLMs for reasoning tasks with MoE architecture and 671B parameters. R1 outperforms OpenAI's o1 in benchmarks, while R1-Zero achieves reasoning via reinforcement learning alone. Distilled models also show competitive performance.

In Beijing, 12,000 human runners will compete against advanced humanoid robots in a half-marathon. Robots from over 20 companies will participate, with strict rules mandating bipedal movement and height limits. The highly anticipated Tian Gong robot, capable of running at 10 km/h, aims to complete its first full 21km race. This event highlights China's growing investment in AI and robotics, with plans for further competitions and substantial R&D funding.