OpenAI offers 50% off GPT usage during off-peak hours to boost AI resource utilization. The new Batch API allows batch query processing, enhancing efficiency and cost-effectiveness. By utilizing this feature, users can enjoy discounted pricing and improved task processing capabilities.

Google researchers introduce Infini-attention to enhance large language models, enabling them to process unlimited-length text with high efficiency and performance. This technique combines compressed memory and local causal attention to handle long-range and short-range contextual dependencies effectively. Experiments show improved model performance and reduced memory requirements, making it ideal for various practical applications and customizations in the AI field.

The Limitless Pendant, part of the Limitless system by Dan Siroker's company, aids in meeting preparation, note-taking, and data retrieval through AI. Integration with various applications is a key focus for the future. The upcoming August release features a $99 device with 100 hours battery life and advanced voice recording technology. Consent mode ensures privacy by not recording without permission.

Medium will no longer pay for AI-generated content starting May 1, 2024. Partners must focus on human storytelling. Using AI for articles will result in removal from the program. Disclosure is required if AI content is published for free. Medium celebrates one million members.

Zoom is using AI to optimize the video calling experience in its first major update in over 10 years. New features include portrait lighting, audio isolation, dynamic reactions, and enhanced collaboration tools like a "Meetings" tab. Upgrade to version 6.0.0 for access to Zoom Workplace.

OpenAI released GPT-4 Turbo model for devs and ChatGPT subscribers. It dominates Chatbot Arena with superior performance in coding, math, and English. To compare, visit Chatbot Arena website. For direct use, subscribe to ChatGPT Plus for $20/month.

At the 2024 Baidu Create AI Developer Conference, Baidu launched Wan Yuan, an intelligent computing operating system. Wan Yuan optimizes computing resources, offers diverse language models, and includes tools for efficient model management. It aims to simplify AI application development, leading the AI-native era.

Xiaodu Technology launched the world's first AI-native operating system, DuerOS X, at Create 2024 Baidu AI Developer Conference. It brings more natural and intelligent interaction, integrating Baidu's "Wenxin Yiyu" and supporting wake-up-free voice interaction with high accuracy. It offers 5200+ intelligent skills and services, enhancing user experience and opening new possibilities for developers in AI technology.

360 announced the open source of its 7B model with 7 billion parameters on GitHub, marking a breakthrough in AI. The model covers multiple languages, excels in evaluations, and provides tools for developers. Founder Zhou emphasized the importance of open source for driving tech development, aiming for practical application effects over text length.

Alibaba's Alimama introduces AI-powered "Super Short Video" feature, revolutionizing e-commerce content creation. Businesses can easily generate captivating short videos with one-click, addressing production challenges and enhancing marketing effectiveness. AI technology shapes the future of personalized content creation in e-commerce. Alimama commits to continuous optimization and research, providing comprehensive solutions for businesses.

x.AI's Grok-1.5V unveils powerful multimodal model with enhanced capabilities in understanding and interacting with the physical world. It excels in tasks like code writing, meme explanation, and home maintenance suggestions. Exciting developments signal progress towards beneficial AGI.

ChatGPT receives major update, introducing GPT-4 Turbo model with enhanced writing, reasoning, math, coding skills and updated knowledge base trained until 2023. OpenAI faces challenges, including copyright allegations for video model Sora.

Adobe used competitors' AI images to train Firefly, contrary to its claims of using only copyrighted or public domain content. AI images, including Midjourney's, made up 5% of Firefly's training data. This raises ethical and legal concerns about AI-generated content integration in training.

Google Research introduced Patchscopes, a framework to explain large language models' internal mechanisms. It offers natural language explanations for hidden representations, addressing limitations of previous methods. Patchscopes aims to detect and correct model illusions, explore multimodal expressions, and study predictions in complex contexts.

In its latest update, Ideogram enhances its AI image generation platform with new features like description-based prompts, negative prompts, and speed modes. The upgrades improve image coherence and realism, surpassing competitors like Midjourney and DALL-E.

Large language models (LLMs) like OpenAI's ChatGPT, Claude AI, and Gemini show limited decision-making abilities. Research explores decision-making challenges and proposes solutions like AutoGPT and "Tree of Thoughts" for context-aware decision-making enhancements. These methods aim to improve LLMs' capabilities to collaborate with humans in strategic decision-making across various fields.

Microsoft researchers propose a new language model, RHO-1, utilizing Selective Language Modeling (SLM) to improve few-shot accuracy on mathematical tasks. SLM focuses on training tokens aligned with desired distribution, enhancing model performance and data quality. Future work includes extending SLM to other fields and exploring token reweighting for improved training.

iOS 18 to bring major AI tools without cloud reliance, boosting Siri and Spotlight. iPhone 15 series and older models will benefit, with future dependencies on powerful neural engines in iPhone 16 series. Apple aims for device-based AI processing, with updates planned at WWDC 2024. Potential for AI chatbot competition in future releases.

Microsoft announces full support for Windows on Arm and teases major changes at Build conference. Windows users and developers are set for a tech feast with improved Arm processor compatibility and a new AI Explorer feature. New Surface devices with Qualcomm Snapdragon X Elite processor will also be unveiled, highlighting Microsoft's hardware-software integration. Leading the industry with innovation, Microsoft continues to fuel technological progress.

Meta AI's research team has released OpenEQA, an open-source benchmark dataset for evaluating AI systems' understanding of real-world questions. The dataset covers 7 categories and tests object recognition, reasoning, and common-sense knowledge. Researchers aim to create intelligent agents for augmented reality and mobile robots. The dataset includes 1636 questions with human-generated answers for evaluation.