OpenAI releases "Model Specification" document for transparency on AI behavior, aiming to shape model reactions to user input. Share feedback by May 22.

Microsoft's research report reveals AI's growing impact on knowledge workers, with AI usage doubling to cover 75% of them. The report emphasizes AI's positive role in boosting productivity. Microsoft warns about the trend of "Bring Your Own AI" and introduces Microsoft 365 Copilot. AI skills on LinkedIn profiles are increasing job opportunities. Despite AI's benefits, companies must focus on employee retention as many consider resigning.

Multi-modality describes how humans understand the world through senses. Communication between people is multi-modal, and future communication with machines will be too. Multi-modal AI uses various strategies for different data types, involving fusion of single-modal models for a comprehensive understanding of reality.

"Alibaba Cloud releases Tongyi Qianwen 2.5, surpassing GPT-4 Turbo. Achieves top performance in benchmarks, proving leading AI position. Also launches Qwen1.5-110B model, leads industry. Tongyi's multimodal and code models excel, setting industry standards."

Apple highlighted AI application and significance in devices at iPad launch. New iPad Air boasts M4 chip with specific neural engine, enhancing AI performance. iPad Pro upgraded to M4 chip with impressive CPU and GPU improvement. AI advancements integrated in apps like Pixelmator for enhanced user experience. Apple hints at empowering iPadOS app developers with AI features. Logic Pro adds AI accompanists for music production. iPad Pro uses AI for document scanning and flash adjustment. Apple foc

OpenAI introduces new image detection tool for its DALL-E AI generator with watermarking method for source attribution. Classifier can detect DALL-E 3 images with 98% accuracy, but only 5-10% on other AI models like Midjourney. Watermarks added to Voice Engine platform.

Meta launches enhanced generative AI tools for advertisers, allowing complete image transformations and text overlays. Updates include text variations, font styles, aspect ratio adjustments, and integration with Instagram and Facebook. AI capabilities aim to expedite ad creation, with global rollout by year-end.

Itamar Arel of AI startup Tenyx fine-tuned Meta's Llama-3 model to surpass GPT-4. Their method addresses catastrophic forgetting, enabling fast, selective parameter updates. Tenyx-70B is open-source, reshaping the AI landscape with potential for specialized chatbots and frequent model updates. This breakthrough shows open-source models can rival proprietary ones, paving the way for a more collaborative AI future.

OpenAI is enhancing ChatGPT to perform web searches and cite sources, challenging search engines like Google. Users can expect more dynamic and referenced information directly from the web, including visual aids. Competition in the AI field is driving OpenAI to improve search functionality in order to provide comprehensive information retrieval and task assistance tools. Current web search feature is limited to ChatGPT Plus subscribers and may have issues with outdated responses.

Freepik acquires AI image upscaling tool Magnific, enhancing design platform with powerful AI technology. Magnific's success driven by word-of-mouth. Acquisition aligns with vision of empowering users with innovative design tools. Integration to drive growth and expansion.

YouTube is launching a new AI feature called "Jump Ahead" for premium subscribers. It analyzes viewer behavior to help users skip to popular video segments with a double-click. The feature is currently available for YouTube Android users with premium service.

GitHub announced the full availability of Copilot Chat in its mobile app, enabling users to ask programming questions and generate code instantly. This move reflects GitHub's commitment to driving technological innovation and enhancing user convenience. Copilot Chat on mobile aims to help developers solve programming problems quickly and efficiently, further expanding its application scenarios. Users can easily access the AI assistant within the app, promoting a more comprehensive and intelligen

In the field of AI large models, domestic unicorn company Brain AI is brewing a technical innovation to develop a video model comparable to OpenAI's Sora, expected to be released this year. Founded in 2019, Brain AI has a strong academic background and has released several notable products in the large model field. The upcoming video model aims to compete with overseas top models.

Recently, Apple is secretly working on "Project ACDC", developing AI chips for data centers, aiming to lead in AI field with self-developed technology, enhancing independence and server performance.

Microsoft's Consumer AI team is developing a new language model, MAI-1, with 500 billion parameters, positioning it between GPT-3 and GPT-4. Led by Mustafa Suleiman, the model aims to enhance Microsoft's AI capabilities.

At Automate conference, Intrinsic, spun off from Alphabet X, announced integrating Flowstate robot platform with NVIDIA products. Collaboration focuses on grasping tech for automation. Intrinsic also working with DeepMind on pose estimation and path planning in automation.

Google plans to leverage AI for cybersecurity with Google Threat Intelligence, utilizing the Gemini AI model to summarize threat reports, making them easier to read for companies evaluating potential attacks' impact.

Hugging Face introduces LeRobot, an open-source toolkit for AI robots, aiming to democratize robotics and foster a new generation of experts. Features pre-trained models, simulation capabilities, versatile library, and hardware integration. Join the community to revolutionize AI robotics!

Internal preview Build 26120.461 (KB5037009) of Windows Dev Channel is now available, featuring general improvements, bug fixes, and significant updates to Copilot. Copilot's new features have been temporarily put on hold based on user feedback. Issues like auto-start after PC restart have been fixed.

OpenAI and developer platform Stack Overflow announced a partnership to enhance AI models and integrate more technical information into ChatGPT. Stack Overflow will use OpenAI's language model to enhance its AI application Overflow AI. The first integration will be rolled out in the first half of this year. Stack Overflow previously banned users from sharing ChatGPT responses on its platform.