NerdNews

Meta's CTO Andrew Bosworth claims OpenAI's CEO Sam Altman is being dishonest about the $100 million signing bonuses offered to poach employees. Meanwhile, Meta is focusing on entertainment and connection with friends in its AI strategy, and a startup called Iyo is suing OpenAI over a trademark dispute.

Four key researchers from OpenAI have left the company to join Meta's superintelligence team. The researchers, including Shengjia Zhao, Shuchao Bi, Jiahui Yu, and Hongyu Ren, bring expertise in deep learning and multimodal models. Meta's aggressive recruitment efforts, reportedly including $100 million signing bonuses, aim to catch up with OpenAI and other leaders in artificial general intelligence.

Tesla has demonstrated its first fully autonomous delivery, with a Model Y traveling 30 minutes from its Gigafactory in Texas to a customer's home without human intervention. The trip included navigating through parking lots, highways, and city streets, showcasing the capabilities of Tesla's robotaxi technology.

Nvidia's GB200 NVL72 supercomputer achieves 2.7× faster inference on DeepSeek V2, a 671B parameter large language model, due to software optimizations including FP8-optimized matrix multiplication and accelerated attention kernels.

Elon Musk and Mark Zuckerberg are battling to control the AI industry, with Musk's xAI and Zuckerberg's Meta competing against Sam Altman's OpenAI. The fight is personal, litigious, and high-stakes, with billions of dollars and the future of AI safety on the line. Musk and Altman are engaged in a lawsuit, while Zuckerberg is trying to poach OpenAI researchers with large signing bonuses.

Runway, an AI startup, is launching Game Worlds, a platform that allows users to create interactive text-based games with AI-generated text and images. The company plans to partner with video game companies to use its AI tools and access video game datasets to train its AI. This move aims to bring AI-generated video games to the market later this year.

The Trump administration is reportedly preparing executive orders to increase AI development, including offering government-managed land for data centers and streamlining the permitting process for power-generating projects. The goal is to solidify the US position as a global leader in AI and promote unchecked development in the area.

Google has announced the integration of Gemini in Android Studio's Agent Mode, designed to handle multi-step development tasks. Agent Mode uses the entire project as context and can directly modify the project, executing multi-step tasks such as building a project, fixing errors, and adding support for dark mode.

Apple researchers are working on a system to stop AI from taking actions that users did not approve. The goal is to create a safer AI that understands the consequences of its actions on a smartphone. The researchers have developed a taxonomy to label mobile app actions along multiple dimensions, such as reversibility and impact on privacy settings.

The US Cybersecurity and Infrastructure Security Agency (CISA) and the National Security Agency (NSA) are calling on software developers to adopt memory-safe programming languages to reduce vulnerabilities. Languages like Rust, Go, and Java are recommended, while C and C++ are not memory-safe by default. The agencies cite the risks of memory vulnerabilities to national security and critical infrastructure.

Meta and Anthropic have won two major fair use victories in court, but the rulings come with significant caveats. The companies' use of copyrighted materials to train their AI models was deemed fair use, but the judges emphasized that this does not necessarily mean their use of the materials is lawful. The rulings also highlighted the potential harm that AI systems can cause to human creators and the need for companies to obtain permission or pay licensing fees to use copyrighted works.

Google's Doppl app uses AI to generate clips of users wearing outfits found on the web. The app requires a full-body photo and a screenshot of the desired outfit. While it shows promise, the app has issues with generating pants and footwear, and can produce unrealistic results. Doppl is available for testing on Android and iOS.

Meta and Anthropic won AI lawsuits, but the war is far from over. The Vergecast discusses the implications of these rulings and other tech news, including Tesla's robotaxi rollout and the Trump Phone.

The cost of networking in AI systems is increasing due to the need for high bandwidth memory and parallel compute. Masked networking costs, such as NVLink fabrics and die-to-die interconnects, are contributing to the rise in costs. The overall datacenter systems market has exploded, with AI systems comprising about half of total server spending in 2024.

HPE's vision for agentic AI has been met with caution from customers, who are waiting to see full functionality and are concerned about data privacy. While some, like network engineers, are enthusiastic, others are hesitant to adopt the technology due to concerns about its maturity and potential risks.

US Federal Reserve chair Jerome Powell warns that AI will bring significant changes to the economy and labor market, potentially replacing jobs, but the timing is uncertain. Powell notes that while AI's current impact is limited, it has enormous capabilities to make significant changes in the future.

Tech CEOs are quietly revealing that AI is replacing human workers, with companies like Amazon, Duolingo, and Shopify already implementing automation and reducing headcount. The rapid evolution of AI technology is enabling machines to perform complex tasks, making human employees redundant.

Denmark is considering a new copyright law that would give citizens control over their own likeness, allowing them to demand the removal of deepfakes and other digital manipulations from online platforms. The proposed law would also establish severe fines for non-compliant tech platforms and provide compensation for individuals whose likeness has been misused.

Tesla released a 30-minute video showing a Model Y driving itself from the factory to a customer's home, claiming it's the first fully autonomous delivery. However, not everyone is convinced, with some pointing out that Waymo has already achieved similar feats and questioning the level of autonomy in the video.

Tesla released a 30-minute video showing a Model Y driving itself from the factory to a customer's home. The video claims to be the first fully autonomous delivery, but experts are skeptical. The car uses Tesla's 'Full Self-Driving' system, which relies on cameras and AI, unlike other companies like Waymo that use LiDAR. While the video is impressive, it's likely a carefully choreographed stunt, and the real test of autonomy is whether the car can handle thousands of unpredictable trips safely.