Overview
Monday brought a flurry of activity across the AI landscape, underscoring how the technology is permeating everything from ride-hailing to government policy. Uber’s product chief detailed a future where AI quietly optimizes every trip, hotel booking, and autonomous fleet—while rival Waymo remains a frenemy. Meanwhile, the video generation arms race heated up as PixVerse locked in a massive $439 million round, pushing its valuation past $2 billion and signaling that investors still believe the next killer AI app will be visual.
But not all news was celebratory. A new WIRED investigation revealed that the U.S. Department of Housing and Urban Development (HUD) is withholding documents about its use of AI for housing policy, citing a legal privilege that doesn’t actually exist—a troubling sign for transparency. On the research front, a novel attention mechanism promises to make large language models both faster and more memory-efficient, a potential boon for developers building on platforms like GetAI Business. And in a more philosophical vein, a new book excerpt revisits ELIZA, the 1960s chatbot that foreshadowed today’s tendency to overshare with AI.
Today's Big News
Uber’s AI Strategy: More Than Just Rides
Uber’s Chief Product Officer Sachin Kansal laid out a vision where AI works behind the scenes to improve rider and driver experiences without making a grand show of it. The company is integrating AI into its financial services ambitions, using predictive models to recommend hotel bookings and optimize pricing. On robotaxis, Kansal described Uber’s increasingly complicated relationship with Waymo—it operates Waymo’s autonomous vehicles in some markets but is also building its own AV Lab data operation. The takeaway: Uber wants to be the operating system for mobility, not a full-stack automaker, and AI is the glue that holds it together.
PixVerse Raises $439M to Scale Video Generation
Just months after the video-gen boom seemed to cool, PixVerse announced a $439 million funding round, pushing its valuation past $2 billion. The startup plans to expand its “world model” offering—a neural network that can simulate physical environments—and push into new geographic markets. The raise signals that while consumer text-to-video may be commoditizing, enterprise-grade world models for simulation, gaming, and robotics still command premium valuations.
Tech’s Richest Founders Are Grinding Again Over AI
A growing trend: ultra-successful tech founders who could retire are diving back into day-to-day work, driven by fear of missing AI’s defining moment. From former CEOs to early employees of giants, they’re launching new AI startups or returning to operational roles. The motivation isn’t just wealth—it’s the chance to shape what they believe is the most important technological shift since the internet.
New Paper: Low-Rank Attention Residuals Could Supercharge LLMs
Researchers introduced Attention Residuals, a method that replaces the fixed residual sum in LLMs with depthwise attention over previous sub-layer outputs. By using low-rank projections, the technique decouples routing from representation, potentially reducing memory and compute costs while maintaining—or improving—model quality. For anyone deploying large models in production, this is a development worth watching closely.
Government Secrecy Over AI in Housing Policy
In response to a public records request, HUD has withheld documents detailing how the Department of Government Efficiency (DOGE) used AI to inform housing policy. The agency cited a legal privilege that does not exist under federal law, raising serious transparency concerns. The incident highlights the widening gap between government AI adoption and the public’s right to know how those systems affect critical policy decisions.