Google's Gemini AI now lets Google One AI Premium users revisit past conversations, offering context-aware responses without restating history. It can summarize previous chats and build on projects, building on earlier features like remembering user preferences. Privacy tools let users manage chat history. This update is rolling out in English, with plans to expand to more languages and Google Workspace clients. ChatGPT has similar recall capabilities.

Researchers developed a bio-inspired robotic control scheme mimicking human neuromechanics to enhance robot adaptability in unstructured environments. The system, featuring a muscle model and cerebellar network, regulates stiffness and improves motion accuracy, enabling collaborative robots to learn and adapt effectively like humans. Tested successfully, it reduces training time and enhances performance across tasks.

Wenxin's large model has been upgraded while managing costs. Starting April 1st, all users can access Wenxin Yiyang for free on PC and APP platforms. The new Deep Search function, offering advanced information retrieval and multi-modal interactions, will also be fully available for free, enhancing user experience and accessibility. Users can already test Deep Search via the official website, with the APP launch coming soon.

AI aggregator OmniGPT suffered a data breach exposing 34 million chat records, 30,000 user emails and phone numbers. The hacker, Gloomer, leaked messages, uploaded files, API keys, and credentials. This highlights security risks in AI platforms and the need for robust protection measures. Experts stress the importance of adhering to industry best practices to safeguard sensitive user information.

OpenAI updates its Model Spec focusing on control levels, public licensing, expanded topic discussion, AI rule compliance, customization for developers, unbiased perspective presentation, and enhanced transparency in governance. These changes aim to promote intellectual freedom, ethical use, and innovation in AI development while ensuring safety and accountability.

Nvidia's AI-driven cybersecurity platform, leveraging BlueField-3 DPU and the Morpheus framework, is being adopted by firms like Armis and CrowdStrike to enhance real-time threat detection in OT environments. The solution provides non-intrusive security measures, deep network visibility, and AI-powered anomaly detection to safeguard critical infrastructure against sophisticated cyber threats. This scalable approach addresses the convergence of IT and OT systems, ensuring robust protection withou

Google is testing a machine learning model in the US to estimate users' ages and identify those under 18, adjusting settings and offering age verification options. For minors, safety features like SafeSearch and YouTube restrictions will be enforced. This initiative aligns with increasing focus on online child safety, alongside regulatory efforts such as KOSA and COPPA 2.0. Meta is also using AI for similar age detection. Additionally, Google will introduce parental controls for device usage dur

Sam Altman revealed OpenAI's plans for GPT-4.5 (Orion) and GPT-5 on X. GPT-4.5 will be the last non-chained reasoning model, aiming to simplify the product line. GPT-5 will unify O-series and GPT-series models, offering advanced intelligence levels for subscribers. Release dates are expected within weeks to months.

Google's Gemini AI assistant now offers the Deep Research feature on Android and iOS for paid Gemini Advanced subscribers. This tool generates detailed research reports using advanced reasoning and long-text understanding, initially launched as a web-only option in December. Users can pose complex questions, modify research plans, and receive notifications upon completion. Reports include tables, sources, and export options to Google Docs. Powered by Gemini 1.5 Pro, the feature will eventually t

Groq secures $1.5 billion from Saudi Arabia to expand AI chip operations, including an AI data center in Dammam optimized for inference tasks using its energy-efficient LPU chips. The LPUs are specifically designed for large language models, simplifying data processing and reducing reliance on external networking components. Groq offers GroqRack systems and GroqCloud, delivering 12 petaflops of performance for AI workloads.

Adobe launches a new generative AI video model in public beta, integrated into the Firefly web app. It allows creators to generate videos from text or images with features like camera angle control and B-roll creation, supporting 1080p and future 4K resolution. Powered by Adobe's proprietary AI, it ensures IP safety and offers two subscription plans: Standard and Pro, providing credits for video/audio generation. Additional tools include 3D sketch-to-image and multilingual audio/video translatio

Baidu plans to launch its next-gen AI model, Wenxin 5.0, later this year, focusing on multimodal capabilities and cost reduction. The foundation model will support text, video, image, and audio processing while enhancing productivity through lower inference costs. This move aims to compete with US firms like OpenAI amid China's growing AI innovation. Current details remain limited, though Baidu integrates AI into various products, including cloud and content platforms.

Zhipu and Samsung collaborate to integrate Agentic GLM technology into the Galaxy S25 series. This AI enhances real-time voice/video calls, system functions, and content creation with multi-modal processing, low-latency inference, and seamless translation. The partnership marks Samsung's first global AI-powered hardware launch, offering immersive user experiences. Future advancements in intelligent agents could transform human-device interactions.

Google's annual developer conference, Google I/O 2025, will be held on May 20-21 at Shoreline Amphitheatre. CEO Sundar Pichai will deliver the keynote, with sessions focusing on AI advancements following last year's emphasis on the Gemini platform. The event comes amid heightened competition from OpenAI, DeepSeek, and Microsoft's Build Conference. NVIDIA GTC and Apple's WWDC are also scheduled earlier and later in the year.

Microsoft expands its renewable energy portfolio by 389 MW through solar projects in Illinois and Texas, supporting AI initiatives and local operations. This adds 2% to its 20 GW contracted capacity. Microsoft prefers renewables for cost efficiency and rapid deployment, aiming for carbon negativity by 2030. It invests in carbon removal tech and purchased 7 million tons of carbon credits to offset 2023 emissions. Hybrid systems with storage ensure reliable power supply.

Epoch AI's study reveals ChatGPT's energy consumption is lower than previously thought, with an average of 0.3 watt-hours per query using GPT-4o. This challenges earlier estimates that were based on outdated assumptions. While recent AI efficiency improvements exist, growing AI usage and the shift toward inference models increase overall power demands. Analysts suggest minimizing computational loads to address energy concerns.

During the AI Action Summit, OpenAI CEO Sam Altman revealed Elon Musk's $97.4 billion proposal to acquire OpenAI's non-profit arm was seen as an attempt to hinder OpenAI's progress. Altman called Musk a competitor, noting his funding for xAI. He also joked about Musk's unstable life. Altman previously rejected the offer publicly, and the OpenAI board is likely to formally decline it, though no official decision has been made. Board member Larry Summers confirmed he received no official notificat

Apple is partnering with Alibaba to bring its AI platform, Apple Intelligence, to China after previous talks with DeepSeek and ByteDance fell through. Initially, Baidu was chosen but adaptation issues arose. The absence of Apple Intelligence in China contributed to an 11% drop in iPhone sales, allowing Huawei to gain ground. Apple CEO Tim Cook emphasizes the importance of this market. Meanwhile, Apple plans to launch the fourth-generation iPhone SE and has collaborated with OpenAI for US markets

Pinkfish, an AI-powered workflow automation startup, aims to simplify digital transformation for businesses. Co-founded by Charanya Kannan and Ben Rigby, the platform uses natural language prompts to create AI-driven workflows, integrating with over 200 apps. It addresses challenges posed by legacy systems and prioritizes deterministic execution. With $7.6 million in pre-seed funding led by Norwest Venture Partners, Pinkfish focuses on sectors like retail and services, offering a balance of full

DeepSeek's valuation is speculated to range from $1 billion to $150 billion, with Liang Wenfeng holding an 84% stake. As a private company, it remains tight-lipped about its financials, sparking industry discussions. Comparisons to OpenAI and Anthropic highlight its potential impact on the global AI landscape, despite uncertainties. Liang, a former quantitative trader, founded DeepSeek in 2023, and the company has already influenced AI investment trends while facing security concerns from govern