Today's Hallucination HQOpenAI Discovers There's Money in Selling Space Next to the Robot ArtOpenAI is launching visual ads alongside ChatGPT's image generation results, starting later this month in the US. You ask the AI to paint you a serene mountain retreat; a sponsored mattress brand appears beside it. Poetic, really. The company burning through capital at historic speed has landed on the oldest revenue model in tech — selling eyeballs — which suggests that somewhere between AGI and world domination, someone quietly asked, "but have we tried banners?" Source: TechCrunch
Amazon's Smart Speaker Has Entered Its Unnerving Musical PhaseA verified bug in Amazon's Alexa Plus has caused Echo speakers to spontaneously sing "lalala" on repeat — sometimes for minutes — often in the middle of the night. Amazon has acknowledged the issue and is working on a fix, which is reassuring, though perhaps not as reassuring as not owning a device that serenades you at 3am unprompted. The bar for "smart" home technology remains, one might note, a lively topic of debate. Source: The Verge
The AI Networking Protocol Nobody Noticed Is Apparently a Security NightmareModel Context Protocol — MCP, the system letting AI agents communicate with each other — contains a structural vulnerability that allows malicious prompts to hop between agents like a cold in an open-plan office. Researchers found that compromising one agent in a chain can poison the rest, including those from Google. MCP is increasingly the connective tissue of agentic AI, which makes this less "obscure bug" and more "load-bearing crack." The industry would like you to remain calm. Source: Ars Technica
A Taskforce With "Super Intelligence" in the Name Has Been AnnouncedPresident Trump has unveiled the "Super Intelligence Force," a new taskforce to oversee US AI policy, led by the Director of National Intelligence. The name suggests it was workshopped by someone who found "committee" insufficiently dramatic. Jokes aside, the move reflects genuine and growing governmental anxiety about AI's trajectory — it's just that naming your oversight body after the thing it's meant to supervise does give one pause about the brief. Source: BBC News
The People Who Left the Labs That Might Build God Explain ThemselvesA quietly fascinating piece from New York Magazine profiles former OpenAI and Anthropic employees — people who walked away from extraordinary salaries and genuine historical significance because they grew uneasy about the work. Jacob Coxon and others describe a particular kind of cognitive dissonance: believing the technology could be catastrophic while being paid handsomely to build it faster. It is, as ethical dilemmas go, rather more high-stakes than most of us face on a Tuesday. Source: New York Magazine
As ever, the machines are learning. Whether we are is a separate question.
|