Bogus article. They didn't write anything to the drive, they faked writes to run up the drive's write counter to test whether the controller would declare the drive worn out despite the flash not being touched.
I live in Poland and speak Polish but tend to have my devices set to English because of how low quality UI translations tend to be. The endless confusion this causes to software is amusing. Worst/best of all is Google Maps on Android, in which random parts of the interface, including voice directions, can independently be in Polish or English, and which is which changes on its own on a daily basis.
It's the latter. Some people abuse the system by highlighting the full length of every broadcast, turning their highlights section into a complete archive of their streams, which is not something Twitch ever wanted to offer.
Very true. The most recent "leak" was of a document that is available on archive.org and has been for years[1]. But the bottoms of the pages have "NATO RESTRICTED" printed, so it was declined, and news sites once again exploded with "War Thunder player leaks secret military documents on forum!!"
The requirements of orbital launch are unyielding. If you make a car 50% heavier, it will have worse mileage and handling, but it will still get you where you need to go. If you make a spacecraft 50% heavier, it will never reach orbit.
Usually "uncensored" models have been made by instruction tuning a model from scratch (i.e. starting from a pretrained-only model) on a dataset which doesn't contain refusals, so it's hard to compare directly to a "censored" model - it's a whole different thing, not an "uncensored" version of one.
More recently a technique called "orthogonal activation steering" aka "abliteration" has emerged which claims to edit refusals out of a model without affecting it otherwise. But I don't know how well that works, it's only been around for a few weeks.
Funnily enough, of all that I've tried, the model by the best at writing porn has been not one of ones uncensored and tuned exactly for that purpose, but stock Command R - whose landing page lists such exciting uses as "suggest example press releases" and "assign a category to a document".
That might be part of it, but I think the bigger factor is cost optimization. OpenAI in particular keeps replacing their models with with versions that are much faster (and therefore cheaper to run) which are supposed to be of equivalent quality but aren't really. GPT-4 -> GPT-4-Turbo -> GPT-4o have all been big upgrades to cost and latency but arguably downgrades to "intelligence" (or whatever you want to call it)