NerdNews

OpenAI has launched two open AI reasoning models, gpt-oss-120b and gpt-oss-20b, which are freely available to download. The models are capable of sending complex queries to AI models in the cloud and can run on a single Nvidia GPU or a consumer laptop. They outperform other open models in several benchmarks but hallucinate more than OpenAI's proprietary models.

Google DeepMind has introduced Genie 3, a foundation world model that can train general-purpose AI agents. Genie 3 generates interactive 3D environments and features promptable world events, allowing for physically consistent simulations. This model is a crucial step towards achieving artificial general intelligence (AGI) and has implications for educational experiences, gaming, and prototyping creative concepts.

Google DeepMind's Genie 3 model can create interactive 3D environments in real-time, with improved memory and longer interaction times. The model can generate worlds with a prompt and supports 'promptable world events' like changing weather conditions. Genie 3 will be available as a limited research preview to a small cohort of academics and creators.

OpenAI has released its first open-weight models, gpt-oss-120b and gpt-oss-20b, which can run locally on consumer devices and be fine-tuned for specific purposes. These models represent a shift away from OpenAI's recent strategy of focusing on proprietary releases and are available for free on Hugging Face.

DeepMind's Genie 3 is a world model that generates interactive worlds from prompts or images, allowing for real-time changes and high visual fidelity. It can be used for research, game development, and training AI agents, but has limitations such as limited memory and imperfect world building.

JetBrains has released a private preview of Kineto, an AI-driven no-code platform for creators and small businesses. Kineto allows users to build apps with no coding experience, using prompts to generate prototypes. The platform aims to bridge the gap between creative people and no-code platforms, but may face challenges in terms of reliability and security.

Microsoft has announced plans to make its WinUI framework truly open source, with a four-phase process to increase transparency and community involvement. The move aims to address developer frustration with the framework's slow progress and lack of investment. WinUI is a key component of Microsoft's Windows App SDK and supports Fluent Design, used for apps across Windows, web, and mobile.

The speakers explain Stripe's approach to refactoring stubborn Ruby monoliths, using centralizing migrations with a point of leverage and good ratchets to tackle issues like modularity and tech debt.

The introduction of Gemini's AI on smartwatches, such as the Samsung Galaxy Watch 8, marks a significant shift in how we interact with AI. However, this new technology requires users to relearn how to use AI, as it's different from traditional voice assistants. The AI can perform complex tasks, but its unpredictability and open-ended nature make it challenging to use effectively. Users need to invest effort into training the AI to know their preferences and habits, which can be a daunting task.

The Nintendo Switch 2's game-key cards, which require a download to play, may signal the end of physical game ownership. These cards, unlike traditional game cards, do not contain the full game data and instead act as a download key. This shift could lead to a loss of ownership and control over games, as well as potential issues with game preservation and accessibility.

Security researchers discovered a chain of high-severity vulnerabilities in Nvidia's Triton Inference Server, which could lead to remote code execution and full system compromise. Nvidia has patched the bugs, and users are advised to update to the latest version.

The Port of Tyne has secured UK government funding to develop and deploy connected and automated mobility services. The project, called Port-Connected and Automated Logistics, aims to demonstrate autonomous container transport at the port, enhancing safety and operational performance.

OpenAI and Amazon Web Services (AWS) have partnered to make OpenAI models available on AWS. This is the first time OpenAI models will be offered by AWS, and they will be available as a model choice with Amazon AI services Bedrock and SageMaker AI. The move is seen as a competitive strike against Microsoft, which has been a major cloud partner for OpenAI. The partnership allows AWS enterprise customers to easily experiment with using OpenAI models with their hosted AI apps.

Critical security flaws in Broadcom chips used in over 100 Dell computer models could allow attackers to take over devices, steal passwords, and access sensitive data, including fingerprint information. The vulnerabilities exist in Broadcom BCM5820X series chips found in tens of millions of Dell business PCs with ControlVault3. Dell has notified customers about updates to fix these bugs, and users are advised to apply security updates and keep their systems up to date.

OpenAI's open-weight GPT-OSS models are now available on Amazon Web Services, allowing customers to access OpenAI models through Bedrock and SageMaker platforms. This collaboration enables companies to utilize OpenAI's models on Amazon's cloud infrastructure, potentially enhancing their AI capabilities.

Lyft and Baidu plan to launch a robotaxi service in Europe, starting in the UK and Germany in 2026. They will use Baidu's Apollo vehicles, which are already operating in China. The service will be available through the FREENOW ride-share app, which Lyft recently acquired.

The EU AI Act aims to create a uniform legal framework for AI innovation across EU countries. It applies to both local and foreign companies, and sets a high bar for what AI should and shouldn’t do for society. The Act has a risk-based approach, banning certain uses of AI, flagging high-risk uses, and applying lighter obligations to limited risk scenarios. It comes with penalties, including fines of up to €35 million or 7% of annual turnover.

The US government has added OpenAI, Google, and Anthropic to its list of approved AI vendors for federal agencies. The companies' AI tools will be offered via the Multiple Award Schedule (MAS) platform, allowing government agencies to access pre-negotiated contracts. The move aims to boost AI adoption in the federal government while ensuring security and performance standards are met.

Alaan, a Middle East-based spend management platform, has raised $48 million in Series A funding. The company uses AI to streamline financial processes and has already saved finance teams over 1.5 million hours of manual work. Alaan plans to use the funding to accelerate its expansion in Saudi Arabia and invest in AI-driven finance automation.

Tesla shareholders have filed a lawsuit against the company and CEO Elon Musk, alleging that they made false statements about the safety of their robotaxis. The lawsuit cites incidents of robotaxis violating traffic laws and notes that the company's stock fell after reports of issues with the service.