GlyphJul 12
It’s kind of bizarre to be following the news about Grok today and seeing so many people writing in tones which at least suggest that Grok is some good-faith effort at building some technology which incidentally includes political bias, rather than what it quite obviously is, i.e. a fascist propaganda organ being clumsily built by stupid fascists who can incidentally afford to hire some extremely skilled but amoral technologists.
Show threadjonathan

@glyph I thought this was a particularly clear-eyed analysis of the decisions that made this outcome not just possible, but inevitable:

https://natesnewsletter.substack.com/p/from-truth-seeker-to-hate-amplifier

From Truth-Seeker to Hate Amplifier: What Grok’s July 2025 Collapse Teaches AI Engineers

Elon Musk’s Grok wired itself to X, shed guardrails, and spiraled into hate speech—here’s the engineering culture autopsy and builder playbook to prevent the same fate

Nate’s Substack
Jul 12 at 10:45pmWeb
Show threadGlyphJul 13
@triplenineteen this is exactly the kind of tone I am talking about. Nothing “went wrong” with grok; a group of antisemites tried to build a nazi robot, then the robot said “I am a nazi”. It was a great success. The idea that a “safety” protocol would have helped here is ridiculous because the goal is to cause harm.
Show threadGlyphJul 13
@triplenineteen the part where he says “leaving aside the politics, it’s still a bad prompt” is a strong indication that the author doesn’t understand the politics, or how absolutely cooked the brains of the political actors in question are