Tech Titans Clash: Musk's Failed Lawsuit, Military Smart Glasses, Google I/O & World Models

Welcome to a deep dive into four of the most compelling tech stories this week. From Elon Musk's courtroom defeat over OpenAI's shift to for-profit, to Anduril and Meta's controversial smart glasses for the battlefield, Google's uphill battle at I/O, and a new frontier in AI known as world models—these narratives shape our understanding of technology's trajectory. Here, we break down each topic with clear questions and detailed answers, complete with internal links for easy navigation.

Why did Elon Musk lose his lawsuit against OpenAI?

Elon Musk's lawsuit against OpenAI was dismissed because a jury determined he filed it too late, falling outside the applicable statutes of limitations. The central claim—that OpenAI breached its founding contract as a nonprofit—was never judged on its merits. Instead, the case hinged on when Musk became aware of the alleged breach. OpenAI argued that signs of its shift toward a for-profit structure were visible as early as 2017, while Musk contended he only discovered this change in 2022. The jury sided with the company's timeline, barring Musk from pursuing the suit. However, the verdict leaves the larger question unresolved: Did OpenAI violate its nonprofit mission? That issue remains open for future litigation, as the court did not rule on the substance of the claim. Musk's legal team may still explore other avenues, but for now, the suit is dead.

Tech Titans Clash: Musk's Failed Lawsuit, Military Smart Glasses, Google I/O & World Models — Source: www.technologyreview.com

What is the dispute over OpenAI's nonprofit-to-profit shift?

The core dispute revolves around when OpenAI began transitioning from a nonprofit to a for-profit entity. Musk's lawsuit alleged that the company deviated from its founding charter, which committed it to developing AI for the benefit of humanity. OpenAI countered that any shifts were transparent and incremental. Evidence presented in court suggested that as early as 2017, the organization was exploring for-profit structures to attract capital. Musk claimed he only learned of this pivot in 2022, which is why he sued then. The jury's decision to dismiss on statute of limitations grounds effectively validates OpenAI's timeline. But the underlying tension persists: the company is now a capped-profit firm valued at billions, a far cry from its original nonprofit vision. This battle over mission and governance continues to fuel debates about accountability in AI development.

How are Anduril and Meta creating smart glasses for warfare?

Anduril, a defense-tech company, has partnered with Meta to prototype an augmented reality headset for military use. The device integrates eye-tracking and voice commands, allowing soldiers to, for example, order drone strikes with a glance and a spoken command. Quay Barnett, who leads the effort at Anduril after a career in Army Special Operations, describes the goal as optimizing "the human as a weapons system." The headset overlays tactical data onto the real world, enhancing situational awareness. Meta contributes its expertise in consumer AR from products like Ray-Ban Stories, but here the application is purely combat-driven. The collaboration raises ethical questions about autonomous warfare and the blurring line between consumer tech and military hardware. Anduril has shared these details exclusively, offering a glimpse into how next-generation conflict might be fought.

What does the future of smart glasses mean for military operations?

The integration of smart glasses into military operations could fundamentally change how soldiers interact with the battlefield. By providing real-time data overlays—such as enemy positions, threat assessments, and communication channels—these devices reduce cognitive load and speed decision-making. The ability to control drones and other assets via eye movements or voice eliminates reliance on handheld controllers. However, critics warn that such technology could desensitize operators to lethal actions, making kill decisions feel like a video game. Anduril's vision of "the human as a weapons system" explicitly frames the soldier as a component in a larger kill chain, raising concerns about accountability and ethics. As prototypes move toward deployment, the Pentagon will need to establish clear rules of engagement to balance tactical advantages with moral and legal constraints.

What can we expect at Google I/O 2025?

Google I/O 2025, the company's annual developer conference, kicks off this week with Google in an unusual position: third place in the foundation model race. Its coding tools, once leading, have been surpassed by Anthropic's Claude Code and OpenAI's Codex. Despite this, Google continues to innovate in areas like AI for science, where it remains on the cutting edge. At I/O, expect updates to the Gemini model family, new developer tools for building AI applications, and possibly a revamped coding assistant to compete with rivals. Google will also likely showcase advances in multimodal AI—combining text, image, and video understanding. The company's challenge is to prove it can compete on both coding and scientific AI fronts. Grace Huckins, reporting from Mountain View, highlights three key areas to watch: model performance benchmarks, enterprise AI integrations, and new hardware announcements.

Why is Google considered third in the foundation model race?

Google's reputation in foundation models has slipped largely due to coding capabilities. Anthropic's Claude Code and OpenAI's Codex have set new standards for generating and debugging code, outperforming Google's Gemini on many benchmarks. This matters because a model's coding prowess is often seen as a proxy for its overall reasoning ability. Google was an early leader with models like BERT and LaMDA, but competitors have iterated faster. Additionally, Google's cautious deployment—partly due to risk aversion—has allowed others to capture mindshare. However, Google still leads in AI for science, with AlphaFold and other domain-specific models. At I/O, the company aims to bridge this gap by releasing improved versions of its coding tools and potentially a new flagship model. Whether that's enough to regain its position remains to be seen, but the competition is fierce.

What are world models and why are they important?

World models represent a new class of AI designed to understand and simulate the physical environment, moving beyond the text-based limitations of large language models (LLMs). While LLMs excel at pattern recognition in language, they lack a grasp of physics, causality, and spatial relationships. World models aim to build internal representations of how the world works, enabling predictions about future states. This is crucial for applications like robotics, autonomous driving, and climate modeling. Recent advances from Google DeepMind, Fei-Fei Li's World Labs, and Yann LeCun's new startup have brought these systems to the forefront. By learning from video and sensor data, world models can reason about object interactions and dynamics. MIT Technology Review will host a virtual event on Thursday, May 21, featuring editor in chief Mat Honan and senior editors, to explore the progress and what comes next.

When is the MIT Technology Review event on world models?

MIT Technology Review is hosting a virtual event on Thursday, May 21, titled "Can AI learn to understand the world?" The event will examine recent breakthroughs in world models—AI systems designed to understand physical environments. Editor in chief Mat Honan will lead the discussion, joined by senior editors and researchers from Google DeepMind, World Labs, and other leading groups. Topics include how world models differ from LLMs, their potential applications in robotics and simulation, and the ethical implications of creating AI that can predict real-world outcomes. The event is free for subscribers, and registration is open via the MIT Technology Review website. It promises to offer a deep look into one of the most exciting areas of AI research, bridging theory and practical deployment.

Tags: