Monday, May 18
AI's Trust Deficit Deepens
The public's patience with AI boosterism is running out in visible ways. Eric Schmidt (opens in a new tab) was booed off the stage at the University of Arizona commencement after pushing AI optimism on graduates already anxious about the job market, a scene that TechCrunch notes is becoming a reliable pattern at 2026 ceremonies. Meanwhile, the Elon Musk-OpenAI trial (opens in a new tab) is centering not on technical disputes but on whether foundational promises were ever made in good faith — a question that cuts to the credibility of the entire sector. Trust, it turns out, is the resource AI companies are burning fastest.
China's answer to US GPU export controls just arrived at 1.54 exaflops: the 'LineShine' supercomputer (opens in a new tab), built entirely on 2.4 million Huawei-designed Armv9 CPU cores with no Nvidia silicon in sight. The system proves that export bans accelerate indigenous development rather than simply delay it. That pressure point gets more acute given that a 45,000-person labor strike at Samsung's memory chip plants (opens in a new tab) now threatens the supply chain that the rest of the AI industry depends on — squeezing hardware availability from both the production and geopolitical ends simultaneously.
Apple is redesigning Siri with auto-deleting chats (opens in a new tab) as a core privacy feature, a direct acknowledgment that conversation history is now a liability as much as a product benefit. At the same time, Microsoft is retiring Teams' Together Mode (opens in a new tab) — a pandemic-era feature that never found a permanent audience — signaling a broader pruning of AI-adjacent features that failed to stick. A former Microsoft VP (opens in a new tab) put it bluntly: the company missed the AI wave the way it missed mobile, and Copilot's quiet rollback from Windows 11 suggests the internal diagnosis matches the external criticism. The giants are quietly retreating from bets that looked obvious two years ago.