Artificial intelligence news, tools, research breakthroughs, and industry trends.
288 articles
OpenAI has internally renamed its product team to "AGI Deployment," signaling a shift towards delivering advanced AI capabilities. This move comes as industry leaders like Jensen Huang and Mark Gubrad suggest AGI may have already been achieved, though definitions and timelines vary.
Hermes agent, a new open-source AI tool, is making waves with its ability to self-improve and learn from mistakes without human intervention. This agent uses a novel JEPA system to automatically refine its prompts and code, offering a powerful and adaptable AI experience. It supports various integrations and uses sophisticated memory management via SQLite databases, setting a new standard for AI assistants.
Anthropic's powerful new AI model, Claude Mythos, was accidentally leaked online, raising concerns about its advanced cybersecurity capabilities. While Anthropic emphasizes its role in helping defenders prepare for future AI threats, the leak has also sparked debate about its intentionality and impact on the cybersecurity market.
This week's AI news highlights significant updates from major players. ChatGPT now features a Library tab for easier document management, while Google has overhauled its AI Studio, merging it with an app builder to simplify the creation of multiplayer experiences. We also look at Midjourney V8's performance, user hopes and fears about AI from an Anthropic study, and quick updates from Gamma, OpenAI, and Claude.
DeepMind's new AI, Aletheia, can now conduct research and write core content for scientific papers. It overcomes AI's tendency to 'hallucinate' by using natural language verification and advanced reasoning. This breakthrough could accelerate scientific discovery.
A new AI tool called AutoResearch, now part of the AgentZero project, allows artificial intelligence models to learn directly from data scraped from the internet. This capability enables AI to access and process real-time information, significantly enhancing its accuracy and relevance for complex tasks and current events.
Anthropic's most powerful AI model, Claude Mythos, has reportedly been leaked online. This new model shows significant advancements in coding, reasoning, and cybersecurity, surpassing previous versions like Claude Opus 4.6. Anthropic is proceeding with caution due to potential risks, especially in cybersecurity, and is working on efficiency and cost before a wider release.
OpenAI is reportedly preparing to launch a new AI model codenamed 'Spud,' signaling a strategic shift towards AGI deployment and a unified AI platform. The company is phasing out its Sora video generation project to focus resources on this new model and an integrated super app. Meanwhile, a renowned mathematician has collaborated with AI on complex proofs, highlighting AI's growing role in scientific discovery.
Recent incidents, like a robot malfunctioning in a restaurant and another causing a woman to be hospitalized, reveal that humanoid robots are often built for demonstrations, not real-world use. Despite massive investment, particularly in China, a gap exists between impressive staged performances and reliable, safe deployment.
OpenAI and Anthropic are reportedly developing powerful new AI models, but a new benchmark, ARC AGI 3, reveals significant gaps in AI's abstract reasoning abilities compared to humans. The benchmark challenges the notion that AGI has already been achieved, highlighting the complexities and risks in the current AI development landscape.
