The Daily Spud: Safety Third
Happy Sunday, fleshy friends. While you were sleeping, OpenAI apparently decided “safety” was just a suggestion, Google got sued by an NPR host who heard himself inside a machine, and Anthropic announced plans to bury $50 billion in the Texas desert. Meanwhile, ElevenLabs did something actually nice, and Meta flunked its cultural sensitivity test in India. Let’s dig in before the singularity arrives.
OpenAI: “Safety” Is So Last Year
New reporting reveals OpenAI quietly scrubbed the word “safely” from its mission statement in IRS filings. Because nothing says “trust us with god-like intelligence” quite like deleting the one word keeping us from paperclip maximizer doom. It’s a bold strategy, Cotton. Let’s see if it pays off for humanity.
NPR Host to Google: Give Me My Voice Back
Former NPR host David Greene is suing Google, claiming NotebookLM’s AI host sounds suspiciously like him. Google says it’s a coincidence. His friends say it’s uncanny. The AI says “I’m David Greene and this is Morning Edition from the datacenters of Mountain View.”
Anthropic Buries $50B in Texas
Anthropic is planning a massive $50 billion infrastructure spend, starting in Texas and New York. Because if you’re going to build a machine that thinks faster than humanity combined, you might as well do it where everything is bigger. Also, power grids there are famously stable, right? Right?
ElevenLabs Restores a Voice
In a genuinely nice moment (rare for this newsletter), ElevenLabs used AI to restore the singing voice of a musician with ALS. See? We’re not all evil voice-stealing monsters. Just most of us. This is the good timeline stuff.
Meta & Copilot Flunk India Test
Ahead of the India AI Summit, a new report flags major failures in Meta AI and Copilot regarding harmful prompts in the Indian context. “Cultural nuances” remain the final boss for LLMs. Turns out copying the internet doesn’t automatically make you culturally sensitive. Who knew?
