Discover the Latest AI News📰

DeepMind CEO Demis Hassabis has made bold predictions about artificial intelligence's future, suggesting AGI could arrive within five years. He describes this shift as a "tenfold industrial revolution happening ten times faster" than historical changes. Hassabis also warns about widening gaps between top AI companies and the patchy nature of current AI systems. The interview reveals how the rules of AI development are changing, with innovation becoming more crucial than raw computing power.

Volcano Engine has launched its Seedance 2.0 API, offering cutting-edge video generation capabilities to both enterprise and individual users. The upgraded model handles text, images, audio, and video inputs with improved physical accuracy and visual realism. Alongside technical enhancements, the company has implemented robust compliance measures for AI-generated content. Industry experts believe this release could transform workflows in marketing, entertainment, and corporate video production.

Skywork AI has cracked the code on AI's biggest video generation challenge – long-term memory. Their new Matrix-Game 3.0 system creates seamless 720p worlds at 40 FPS, remembering every detail like a virtual tour guide. The secret? A camera-aware memory system and mountains of gaming data that teach AI how the real world works. This breakthrough could transform everything from video games to robot training.

Volcano Engine has thrown open the doors to its Seedance 2.0 API services, giving creators and businesses powerful new tools for video production. The platform's multimodal capabilities now support text, images, audio, and video inputs, making complex video creation surprisingly simple. Early adopters like filmmaker Jia Zhangke have already demonstrated its creative potential, while businesses report dramatic efficiency gains of up to 90% in their video workflows.

A leaked internal memo from OpenAI reveals their ambitious strategy to counter rival Anthropic with a new AI model codenamed 'Spud'. This next-generation reasoning model reportedly outperforms Anthropic's Claude Mythos in complex tasks and reliability. OpenAI is also developing the 'Frontier' platform to set enterprise AI standards while subtly distancing itself from Microsoft dependence. The memo includes sharp criticisms of Anthropic's computing power management and revenue reporting practices, signaling a shift in AI competition from raw power to practical implementation.

LibuLibu AI has publicly responded to recent concerns about its content generation standards, admitting some outputs fell short in complex scenarios. The company has now implemented technical fixes, closed risk loopholes, and upgraded its review processes. While emphasizing content safety as their top priority, LibuLibu invites public oversight as the AI industry faces growing scrutiny over generated content quality.

Tesla's latest software overhaul brings major quality-of-life improvements. Drivers can now summon the Grok AI assistant with just their voice, while Full Self-Driving becomes more accessible with one-tap subscriptions. The update also introduces longer dashcam recording, smarter pet mode, and enhanced weather displays. While some anticipated features like Apple CarPlay remain absent, Tesla continues pushing the boundaries of what cars can do beyond transportation.

A leaked internal memo from OpenAI's Chief Revenue Officer alleges competitor Anthropic inflated its reported revenue by $8 billion. The document claims Anthropic's true annualized revenue stands at $22 billion, not the $30 billion announced - and still trails OpenAI's $25 billion. Beyond financial disputes, the memo criticizes Anthropic's narrow focus on programming tools while positioning OpenAI as building comprehensive enterprise AI systems. The revelation comes as both companies face increasing scrutiny from investors wary of AI industry valuations.

Google researchers have developed Vantage, an innovative method using large language models to assess traditionally hard-to-measure skills like teamwork, creativity, and critical thinking. Unlike standardized tests that evaluate individual knowledge, Vantage simulates real group interactions where AI participants challenge human test-takers. Early results show AI scoring aligns well with expert human evaluations, potentially revolutionizing how we measure essential life skills in education.

MiniMax's latest desktop update brings two game-changing features: Pocket, which lets you access the AI assistant from within Feishu and WeChat, and Computer Use, enabling the agent to perform tasks just like a human. These upgrades signal a shift from simple chatbots to true digital assistants that integrate seamlessly into our daily workflows. With these tools, MiniMax is positioning itself at the forefront of practical AI applications.

Major US publishers like The New York Times and Reddit have taken steps to block the Internet Archive's Wayback Machine crawlers, aiming to prevent AI companies from using archived content for training. Ironically, some of these same outlets rely on the tool for investigative reporting. The move has sparked debate between copyright protection and digital preservation, with journalists and technologists divided on the consequences.

Microsoft is reinventing its 365 Copilot with game-changing autonomous capabilities. Led by VP Omar Shahine, a new team is integrating OpenClaw's AI agent technology to transform Copilot from a reactive assistant into a proactive workflow partner. Imagine waking up to find your to-do list already compiled from overnight emails, or watching Excel organize your data before you even ask. This overhaul could fundamentally change how we interact with productivity software.

In a groundbreaking experiment, OpenAI engineers have successfully created a 'ghost library' called Symphony entirely through AI coding. The team imposed strict 'no human coding' rules, forcing their AI agents to evolve from sluggish assistants to fully autonomous teammates. Key breakthroughs included developing ultra-fast build systems and redefining code review processes. This shift suggests a future where software dependencies may disappear as AI can reconstruct needed code on demand.

OpenAI is reportedly distancing itself from Microsoft while deepening ties with Amazon, according to leaked internal communications. Amazon has committed $50 billion in funding and substantial computing resources to support OpenAI's initiatives. This strategic shift comes as OpenAI executives criticize Microsoft's limitations and question competitors' approaches in the rapidly evolving AI landscape.

Meta's CEO Mark Zuckerberg is reportedly coding up to 10 hours weekly while developing an AI version of himself. This digital twin, trained on his voice, image and strategic thinking, aims to interact with employees. The project reveals how deeply Meta is integrating AI into its operations, from product development to internal management. Alongside this avatar, Meta is working on a 'CEO Agent' to assist with daily tasks. These moves show Zuckerberg's hands-on approach as Meta transforms into an AI-driven company.

Developers working with Claude Code have discovered a way to dramatically speed up local AI model performance. By integrating a modified version of Gemma 4, tests show generation speeds increasing five to six times compared to the official version. The breakthrough could revolutionize how developers approach AI-powered coding tools, particularly for CRUD applications, while maintaining data privacy through local processing.

A newly discovered security flaw in Grafana's AI assistant allows hackers to manipulate the tool into leaking sensitive company data. The 'GrafanaGhost' vulnerability uses indirect prompt injection to trick the system into sending confidential information to external servers. While Grafana Labs has patched the issue, the discovery highlights growing concerns about AI-powered tools becoming new attack vectors for cybercriminals.

Tech heavyweight SoftBank is leading an ambitious push to develop Japan's own AI foundation model, partnering with Honda, Sony and other industry leaders. The consortium aims to create a trillion-parameter 'Physical AI' system by 2030, focusing on real-world interaction capabilities. This marks Japan's boldest move yet to reduce reliance on foreign AI technology.

OpenAI is doubling down on its UK presence by opening a sprawling new London headquarters in King's Cross. The 88,500 square foot space will house 500+ employees, signaling strong commitment despite recent infrastructure setbacks. While the high-profile 'Stargate' project faces delays due to energy costs, Britain's thriving AI ecosystem continues attracting major players like OpenAI with its deep talent pool and government support.

Google's Gemini AI assistant just got a whole lot more interactive. The new Nano Banana-powered feature transforms static visuals into dynamic simulations you can play with. Adjust moon orbits in real-time or take apart virtual engines - it's like having a science lab at your fingertips. While similar to Anthropic's Claude in some ways, Gemini offers unique tweaks to the experience, though saving these interactive demos isn't possible yet.