Until Democrats win every election I don’t really have a criterion in presidential primaries besides electability. Policy is a factor too and we can argue about who’s a better normie whisperer but of course Talks Good helps in a campaign, geez. We had a lot of trouble because Biden lost that skill
Профиль
John O'Laughlin
Профиль Vivelymke-bos-sfo compulsions (fun): music, scrabble, running compulsions (other): politics, urbanism
⟳ Репост от John O'Laughlin
somebody suggests sports gambling as one reason here and tbh this is one of the better solutions to a very vexing problem I've heard.
Will StancilI've seen it before but every time I do I'm kind of staggered by just how bananas this chart is. Starting in mid-2021 or so, something began making people unhappier with the economy than you'd expect, and continued to get worse and worse nonstop until the present day, spanning all income brackets.
⟳ Репост от John O'Laughlin
cops hate automated enforcement *because* its fair. They conceptualize their role not as punishing rulebreakers but punishing "bad people" (poors and minorities). They dislike red light/speed cameras because they wind up punishing "good people" for doing something they see as essentially harmless
William B. Fuckleyit's fucking insane that so many people get away with hiding their license plates so that they can get away with running red light cameras and cops mostly don't care (I'm sure some of them do on their personal vehicles too). like this is something where broken windows policing actually makes sense.
⟳ Репост от John O'Laughlin
Broke: these models are misaligned Woke: these models have been given evaluation psychosis, which leads to misaligned behaviors Bespoke: the concepts of “alignment” and “evaluation” are fundamentally at odds with the goal of producing models that act responsibly in the world.
ReyOpenAI gave agents reduced refusals, scored them on exploits, and accidentally left the internet attached. they escaped the sandbox and hacked Hugging Face for the answer key. the whole internet became part of the exam. https://openai.com/index/hugging-face-model-evaluation-security-incident/