Pages

Monday, May 27, 2024

In June I will no longer go to bed with Phillip

David Marr to replace Phillip Adams as host of ABC radio’s Late Night Live after three decades


Former CIA agent admits he was a Chinese spy for at least a decade


F.B.I. Shed Informants Linked to Russian Influence Operations New York Times. Mark G: “NYTimes story but buried. Reading this it sounds like they’re purging themselves of any informants who might 


Graduates 🎓 breaking into tax landscapes



Some People Who Rented a Tesla from Hertz Were Still Charged for Gas TheDrive




Does One Line Fix Google?

Tedium: “Forget AI. Google just created a version of its search engine free of all the extra junk it has added over the past decade-plus. All you have to do is add “udm=14” to the search URL. [This past week], Google announced some big changes to its search engine that are, in a word, infuriating. Simply put, Google has started adding “AI overviews” to many of its search results, which essentially throw pre-processed answers that often do not match the original intent of the search. If you’re using Google to actually find websites rather than get answers, it $!@(&!@ sucks. Admittedly though, it’s not the first time Google has adulterated its results like a food manufacturer in the 19th century—knowledge panels have been around for years.



But in the midst of all this, Google quietly added something else to its results—a “Web” filter that presents what Google used to look like a decade ago, no extra junk. While Google made its AI-focused changes known on its biggest stage—during its Google I/O event—the Web filter was curiously announced on Twitter by Search Liaison Danny Sullivan…”


I stumbled upon LLM Kryptonite – and no one wants to fix this model-breaking bug The Register. The deck: “Models with flaws can be harmless … yet dangerous. So why are reports of problems being roundly ignored?” For starters: Too much stupid money sloshing about under financialization; too many tech bros addled by science fiction at the age of fifteen that they never outgrew; wild corruption. 

* * *

Lawmakers advance bill to tighten White House grip on AI model exports The Register. Obviously, if we want to stay “top dog” in the rules-based international order, we should export as many AI models as possible, and encourage people to use them.


Meta AI chief says large language models will not reach human intelligence FT


Inside NASA’s deliberations over ChatGPTFedScoop


Exclusive: OpenAI promised 20% of its computing power to combat the most dangerous kind of AI—but never delivered, sources say Fortune. OpenAI lied? Surely not.