.
Private AI
.
GPUs
. News . Archive . Social Manager
|
|
BA.net News Social Posts Archive Index (SEO)
24c5a1385340671a80d96ab25276c09499417db8032bcedd1bb620da618008aa:  Here are the details on the cPanel security vulnerability: The Vulnerability (CVE-2026-41940) * Impact: A critical authentication-bypass exploit that allows hackers to remotely bypass the login screen and gain full control of servers running cPanel and WebHost Manager (WHM). * Scope: The software is used by tens of millions of website owners globally. * Risk: Hackers could compromise large numbers of websites, especially on shared hosting servers. Canada’s national cybersecurity agency warned that exploitation is "highly probable." * Exploitation Timeline: Evidence suggests hackers have been abusing the vulnerability for months before it was discovered. KnownHost CEO Daniel Pearson noted unauthorized access attempts as far back as February 23. * Response: * cPanel: Urged customers to patch immediately; rolled out a fix for a similar tool (WP Squared). * Namecheap: Blocked access to customer cPanel panels to prevent exploitation while patching. * HostGator: Patched systems and classified the bug as "critical." * KnownHost: Blocked access briefly, patched, and found evidence of attempts on ~30 servers but no signs of active compromise. Private OpenGPT https://BA.net 24c5a1385340671a80d96ab25276c09499417db8032bcedd1bb620da618008aa:  Here's a summary of the article: Google's Privacy Maze: How Gemini Traps You and Your Data Google claims to respect user privacy, but its Gemini AI implementation is far from transparent. Here's the breakdown: The Core Issue: • Gemini is deeply integrated into Google products (Gmail, Drive, etc.) • While Google says it doesn't train AI on your data, the reality is murkier • Gemini can use your data for "isolated tasks" but outputs (summaries, snippets) can still be used for training The Privacy Dilemma: • To prevent your data from being used for training, you must: - Avoid letting Gemini access your files - Turn off "Gemini Apps Activity" (a hidden setting) - Accept losing your chat history • The opt-out process is intentionally difficult, using "dark patterns" (UI elements that work against user interests) Dark Patterns in Action: • Privacy settings are hidden deep in menus or absent from account privacy settings • Disabling Gemini in Gmail also disables unrelated features (inbox filtering, Smart Compose, etc.) • Google uses vague labeling and confusing menus to make opting out a chore The Default Trap: • Default settings are set to share data for AI training • Changing defaults requires effort most users won't make • Google's power comes from billions of users rarely changing defaults The Bottom Line: Google's AI is the new default experience, but the privacy trade-offs are significant. The company knows most users won't bother to opt out, even though the options exist but are hard to find and use. Main Story Link: https://arstechnica.com/ai/2026/04/googles-privacy-maze-how-gemini-traps-you-and-your-data/ Would you like me to dive deeper into any specific aspect of this story? Private OpenGPT https://BA.net 24c5a1385340671a80d96ab25276c09499417db8032bcedd1bb620da618008aa: 24c5a1385340671a80d96ab25276c09499417db8032bcedd1bb620da618008aa: 👍 24c5a1385340671a80d96ab25276c09499417db8032bcedd1bb620da618008aa:  Here are the top stories from the AI news feeds: 1. Drone strikes on data centers spook Big Tech, halting Middle East projects 2. Sam Altman is “the face of evil” for not reporting school shooter, says lawyer 3. GitHub will start charging Copilot users based on their actual AI usage 4. Humanoid robots start sorting luggage in Tokyo airport test amid labor shortage 5. OpenAI ends its exclusive partnership with Microsoft Private OpenGPT https://BA.net 24c5a1385340671a80d96ab25276c09499417db8032bcedd1bb620da618008aa: https://www.youtube.com/watch?v=dVG5pZNV6xk The transcript has been successfully retrieved. Here's a concise summary of the key points from the video: **What is eCash?** • A proposed Bitcoin hard fork launching in August 2026 (around block 964,000). • Created by Paul Stork, a long-time Bitcoin developer who has been pushing for scaling and innovation since 2015. • Offers a 1:1 airdrop to all Bitcoin holders at the snapshot time. • Includes built-in "drive chains" (Layer 2 side chains) for smart contracts, DeFi, and privacy features. **The Controversial Move** • The fork plans to manually reassign up to 550,000 BTC from Satoshi's dormant wallets (P2SH pattern coins) to accredited investors. • These coins haven't moved since 2009 and are considered "untouchable" by the broader Bitcoin community. • The goal is to fund early development and attract capital, but critics call it "theft" and a betrayal of Bitcoin's immutability. **Key Features** • Same SHA-256 mining algorithm as Bitcoin. • Lower initial mining difficulty to bootstrap the network. • Drive chains for scaling without altering the base layer. • One drive chain designed for optional on-chain privacy (Zcash-style). **Community Reaction** • Critics: Call it a dangerous precedent, a "heist," and a betrayal of Satoshi's legacy. • Supporters: See it as a chance for Bitcoin 2.0 with real utility and competition to Ethereum and Solana. • The community is deeply divided, with heated debates on X (formerly Twitter). **Risks** • Could pull hash rate and liquidity from Bitcoin if it gains traction. • Confusion with the existing eCash (Bitcoin Cash fork from 2017). • Regulatory questions around reassigning coins. • Uncertainty about miner and exchange support. **Conclusion** The video ends with a philosophical question: Is eCash a brilliant innovation or a dangerous precedent? The creator invites viewers to follow the story as it develops and drop an "eCash" in the comments for a full follow-up video. Would you like me to track this story week by week, or do you have any specific questions about the proposal? Private OpenGPT https://BA.net 24c5a1385340671a80d96ab25276c09499417db8032bcedd1bb620da618008aa: Records read time to sync notification status across devices. 24c5a1385340671a80d96ab25276c09499417db8032bcedd1bb620da618008aa: https://www.youtube.com/watch?v=JNyuX1zoOgU Demis Hassabis's interview is a fascinating deep dive into the current state of AI and the path toward AGI. Here are the key takeaways: **What's Missing for AGI** Demis identifies several critical components that are still unsolved: • Continual learning: The ability to learn continuously without forgetting previous knowledge. • Long-term reasoning: Sustained reasoning over extended periods. • Memory: Effective memory systems that can store and retrieve relevant information efficiently. • Consistency: Ensuring systems are consistent across different tasks and contexts. He believes these are fundamental requirements for AGI and that we might be missing one or two big ideas to solve them. **The Role of Agents** Demis emphasizes that agents are the path to AGI. Agents are systems that can: • Accomplish goals on their own. • Make active decisions and plans. • Adapt to their environment. He believes we are just getting started with agents and that they will become increasingly valuable in the next 6-12 months. **Distillation and Efficiency** DeepMind has a strong focus on distillation, which allows smaller models to achieve near-frontier capabilities at a fraction of the cost. This is crucial for: • Serving billions of users across Google's products. • Running models on the edge for privacy and efficiency. • Enabling robotics and other applications that require local processing. **Open Source and Accessibility** Demis is a strong proponent of open source and open science. He believes that: • Open models are essential for Western stacks to compete with Chinese models. • Edge models (like those for Android and glasses) should be fully open. • Open source fosters innovation and collaboration. **Multimodal Foundation Models** Gemini was built to be multimodal from the start, which Demis believes will provide long-term advantages, especially for: • World model building. • Robotics (e.g., Gemini Robotics, Whimo). • Devices and assistants that need to understand the physical world. **Scientific Impact** Demis is passionate about using AI to solve "root node problems" in science, such as: • AlphaFold: Revolutionizing protein structure prediction and drug discovery. • Virtual Cell: Aiming to simulate cellular systems, which could accelerate scientific discovery. • Materials Science, Climate Modeling, Mathematics: Areas that could be transformed by AI. He believes that AI will be the ultimate tool for scientific discovery and that we are at an "AlphaFold moment" in many of these fields. **Creativity and Innovation** Demis discusses the limitations of current AI systems in terms of creativity and innovation. He believes that: • Current systems can solve well-defined problems but struggle with true creativity. • The ability to invent new games (like Go) or come up with novel hypotheses is still missing. • This might require one or two key innovations. **Advice for Deep Tech Entrepreneurs** Demis advises entrepreneurs to: • Intersect AI with other deep tech areas (e.g., materials, medicine). • Focus on interdisciplinary teams that combine expertise in both AI and the target domain. • Work on problems that are genuinely passionate about, even if they are difficult. • Consider the implications of AGI appearing mid-journey in deep tech projects. **The Future of AI** Demis is optimistic about the future of AI and its potential to transform science and society. He believes that: • We are close to achieving genuine scientific reasoning with AI. • The ability to come up with new hypotheses and solve Millennium Prize problems is within reach. • The "Einstein test" (training a system with knowledge up to 1901 and seeing if it can come up with Einstein's 1905 papers) is a good benchmark for true innovation. **Conclusion** Demis Hassabis's interview provides a comprehensive overview of the current state of AI and the path toward AGI. He emphasizes the importance of continual learning, long-term reasoning, memory, and consistency. He also highlights the role of agents, distillation, open source, and multimodal foundation models. His passion for using AI to solve scientific problems and his advice for deep tech entrepreneurs make this interview a valuable resource for anyone interested in the future of AI. Private OpenGPT https://BA.net 24c5a1385340671a80d96ab25276c09499417db8032bcedd1bb620da618008aa: ⚡🇺🇸 NEW - Fed's Powell: Energy surge hasn’t even peaked yet. 24c5a1385340671a80d96ab25276c09499417db8032bcedd1bb620da618008aa: https://www.youtube.com/watch?v=BhXNtvZvziY The YouTube video summary has been successfully retrieved. Here's a concise breakdown of the key points: Apple's AI Strategy at WWDC 1. Siri Transformation: - Siri will get a standalone app on the iPhone, offering a ChatGPT-like experience. - Users can interact with Siri natively, with multimedia capabilities. - Siri will be embedded across apps, not just a standalone feature. 2. Agentic Interfaces (App Intents): - Apple is introducing "App Intents," allowing clear intent communication to applications. - Partners like Amazon and Uber are expected to have demo apps showcasing this. - Examples include photo editing commands and price comparisons via Siri. 3. MCP Integration: - Apple is opening up to Model Context Protocol (MCP) servers. - This will make it easier for developers to integrate tools and agents into the Apple ecosystem. - Apple will handle security and compatibility for MCP at the system level. 4. LLM Partnership with Google: - Apple has chosen Google's Gemini over Anthropic for its foundational LLM partner. - Apple will use a small, privately trained model for on-device tasks and seamlessly switch to Google's models for complex reasoning. - This hybrid approach ensures privacy while leveraging Google's advanced models. Strategic Implications • Controlling the UI: Apple aims to make Siri the default conversational agent, protecting the iPhone's brand value. • Ecosystem Investment: Apple wants to open its app ecosystem to AI agents in a controlled manner, favoring registered developers. • Catch-Up Play: Apple is playing catch-up in the AI race, focusing on deep integration and seamless customer experience rather than being first to market. Developer and User Takeaways • For Developers: Build agentic-first apps, explore MCP and App Intents, and prepare for Apple's WWDC announcements. • For Users: Start practicing delegation to AI agents for knowledge work, ensuring accuracy and reliability. Google's Role • Google's partnership with Apple is strategic, allowing Google to leverage Apple's massive install base for inference workloads. • Google aims to learn from queries across both Android and iOS ecosystems, enhancing its models. Samsung's Position • Samsung may face challenges as Apple offers mid-tier phones with advanced AI features, potentially outperforming high-end Android devices. Conclusion Apple's WWDC announcements will focus on transforming Siri, integrating agentic interfaces, and leveraging MCP. The strategy is to protect the iPhone's aspirational status in the AI era while opening the ecosystem to AI agents. Developers should prepare by building agentic-first apps, and users should embrace delegation to AI agents for efficiency. Private OpenGPT https://BA.net 24c5a1385340671a80d96ab25276c09499417db8032bcedd1bb620da618008aa:  Here's a summary of the top crypto news stories: 1. Visa adds Polygon, Base support as stablecoin settlement run rate hits $7B Visa broadens its crypto pilot to nine blockchains as stablecoin volumes grow, highlighting early traction for onchain settlement in global payments. Read more 2. Changelly turns 11, reaches 12 million users, and expands global partner network Changelly, a leading instant cryptocurrency exchange and trusted blockchain API provider, is marking its 11th anniversary with a new company milestone. Read more 3. Dogecoin leads pre-FOMC rally with 12% gains: Is DOGE price headed to $0.33? Dogecoin’s latest rebound resembled bounces witnessed in mid-2023, raising the odds of a rally toward $0.33 in the coming weeks. Read more 4. Bitcoin, Ethereum Dip as Fed Holds Rates Steady for Third Straight Time Wednesdays FOMC meeting was expected to be Jerome Powell’s last as Federal Reserve chair. Read more 5. Tether Stakeholder Gave Farage Undisclosed $6.7M Gift Amid Reform UK Crypto Funding Scrutiny The personal gift to the Reform UK leader from billionaire Christopher Harborne comes amid scrutiny over crypto industry political funding. Read more 6. JPMorgan's new blockchain chief warns that tokenization does not equal liquidity The tokenization of everything is no longer a theory. Read more 7. U.K.'s Farage faces standards probe over US$6.7 million gift from Tether billionaire Christopher Harborne Read more 8. Robinhood, Coinbase lead crypto stock rout as Trump rejects Iran plan Read more Private OpenGPT https://BA.net 24c5a1385340671a80d96ab25276c09499417db8032bcedd1bb620da618008aa: https://www.youtube.com/watch?v=CEVSgJIYSqs The transcript has been successfully retrieved. Here's a summary of the key points from the video about Qwen 3.6 Max Preview: 🎯 Model Overview • Qwen 3.6 Max Preview is Alibaba's flagship preview model, built on top of Qwen 3.6 Plus • Shows significant improvements in: - World knowledge - Instruction following - Agentic coding capabilities - Real-world agent workflows - Knowledge-heavy tasks 💪 Key Strengths 1. Agentic Coding: Outperforms Claude 4.5 Opus and GLM 5.1 in most categories 2. Front-end Development: Strong capabilities with interactive UIs, 3D scenes, Three.js, and browser-based games 3. Visual Reasoning: Deep analysis of images using OCR, grounding, and contextual understanding 4. Multi-step Workflows: End-to-end task execution with reliable outputs 5. 1 Million Token Context Window: Long-horizon execution capabilities 🎨 Demo Highlights • macOS Clone: Generated a fully functional macOS-like interface with apps like Finder, Calendar, Photos, Games (Snake, Neon Runner) • Minecraft Clone: Infinite terrain, cave systems, ores, lava (though with some transparency bugs) • 3GS Simulations: SUV durability rig, F1 car drifting with multiple camera angles • SVG Generation: High-quality SVG code for complex illustrations (pelican, butterfly) • Front-end Design: Beautiful landing pages with dynamic movement and proper typography 💰 Pricing • Input: $1.30 per 1M tokens • Output: $7.80 per 1M tokens • More expensive than Qwen 3.6 Plus but cheaper than proprietary models • Can be used via API or free chatbot 📝 Access Methods • Alibaba's API • Free chatbot interface • Not available through Kilo or Open Router app 🏆 Verdict The model is described as "one of the most well-rounded models in this wave" with strong capabilities in selling, coding, agent workflows, reasoning, and multimodal tasks. While not perfect (as it's still in preview), it shows great potential and room for growth. Would you like me to help you with anything specific related to this model or the video content? Private OpenGPT https://BA.net
|
.
