korrents

On the map

Tap a claim on the ring to put it at the centre.

← An AI detector used as a reward verifier is brittle: text that does…

17 connected korrents · 16 moments from 27 Oct 2015 to 24 Oct 2026.

Everything filed under AI writing AI writing Everything filed under writing writing Everything filed under AI slop AI slop Everything filed under AI and human skill AI and human skill Everything filed under LLMs LLMs Everything filed under AI alignment AI alignment Everything filed under social media social media Everything filed under prompt injection prompt injection Same subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subject Read this korrent: An AI detector used as a reward verifier is brittle: text that does not look human-written, even gibberish, can pass as non-AI-generated. An AI detector used as a reward verifieris brittle: text that does not lookhuman-written, even gibberish, can pass asnon-AI-generated. Last stated a month ago 10 Oct 2026 SR Sebastian Raschka — holds since 2026-10-10 — tap for who they are Same subject: AI text detection through pattern recognition is prone to errors because LLM text may defy patterns or human text may mirror AI patterns. — tap to centre the map on it AI text detection through patternrecognition is prone to errorsbecause LLM text may defy patternsor human text may mirror AIpatterns. Last stated 3 months ago 9 Aug 2026 JW John Warner — holds since 2026-08-09 — tap for who they are Same subject: An AI-detector's score reflects the classifier's training distribution, not a true probability that a given text was written by AI. — tap to centre the map on it An AI-detector's score reflects theclassifier's training distribution,not a true probability that a giventext was written by AI. Last stated 3 months ago 15 Aug 2026 SR Sebastian Raschka — holds since 2026-08-15 — tap for who they are Same subject: AI text detectors and the LLMs they try to catch are locked in a permanent cat-and-mouse arms race. — tap to centre the map on it AI text detectors and the LLMs theytry to catch are locked in apermanent cat-and-mouse arms race. Last stated 3 months ago 15 Aug 2026 SR Sebastian Raschka — holds since 2026-08-15 — tap for who they are Same subject: If whole categories of reward hack are undetectable by humans, then training against the hacks we do catch teaches models to cheat only where we cannot see. — tap to centre the map on it If whole categories of reward hackare undetectable by humans, thentraining against the hacks we docatch teaches models to cheat onlywhere we cannot see. Last stated 3 months ago 11 Aug 2026 RG Ryan Greenblatt — holds since 2026-08-11 — tap for who they are Same subject: There is a risk that reading will be devalued due to low-quality AI-generated text. — tap to centre the map on it There is a risk that reading will bedevalued due to low-qualityAI-generated text. Last stated 2 months ago 26 Sept 2026 GR Guillermo Rauch — holds since 2026-09-26 — tap for who they are Same subject: AI-generated text lacks the human mind and knowability that make writing feel worth reading. — tap to centre the map on it AI-generated text lacks the humanmind and knowability that makewriting feel worth reading. Last stated 2 months ago 10 Sept 2026 CM Craig Mod — holds since 2026-09-10 — tap for who they are Same subject: Platforms will lose the ability to detect AI-generated content as the models improve, and should say how confident they are rather than pretend. — tap to centre the map on it Platforms will lose the ability todetect AI-generated content as themodels improve, and should say howconfident they are rather thanpretend. Last stated 4 months ago 9 Jul 2026 AM Adam Mosseri — holds since 2026-07-09 — tap for who they are Same subject: AI agents cannot reliably tell information from instructions, so they can be fooled even when you would not be. — tap to centre the map on it AI agents cannot reliably tellinformation from instructions, sothey can be fooled even when youwould not be. Last stated 4 weeks ago 23 Oct 2026 AK Alex Komoroske — holds since 2026-10-23 — tap for who they are Same subject: A growing fraction of social media is AI slop influencers being replied to by bots — an echo of an echo of an echo. — tap to centre the map on it A growing fraction of social mediais AI slop influencers being repliedto by bots — an echo of an echo ofan echo. Last stated 3 months ago 23 Aug 2026 FC François Chollet — holds since 2026-08-23 — tap for who they are Same subject: A low-effort AI-designed landing page costs a product its audience before the product itself is ever judged. — tap to centre the map on it A low-effort AI-designed landingpage costs a product its audiencebefore the product itself is everjudged. Last stated 3 months ago 13 Aug 2026 MH Mitchell Hashimoto — holds since 2026-08-13 — tap for who they are Same subject: AI is an antidote to slop rather than a creator of it. — tap to centre the map on it AI is an antidote to slop ratherthan a creator of it. Last stated 4 weeks ago 24 Oct 2026 TB Thorsten Ball — holds since 2026-10-24 — tap for who they are Same subject: A Black writer is needed to portray Black life truly, and white people from a Black standpoint — tap to centre the map on it A Black writer is needed to portrayBlack life truly, and white peoplefrom a Black standpoint AC Anna Julia Cooper — holds since c. 1892 — tap for who they are Same subject: A blank page can be terrifying, but a small prompt saying what to write about makes it easier to start. — tap to centre the map on it A blank page can be terrifying, buta small prompt saying what to writeabout makes it easier to start. Last stated 11 years ago 27 Oct 2015 AK Austin Kleon — holds since 2015-10-27 — tap for who they are Same subject: A browser that answers with generated prose instead of links is not a web browser but an anti-web browser, and it hides that substitution from the person using it. — tap to centre the map on it A browser that answers withgenerated prose instead of links isnot a web browser but an anti-webbrowser, and it hides thatsubstitution from the person usingit. Last stated a year ago 22 Oct 2025 AD Anil Dash — holds since 2025-10-22 — tap for who they are Same subject: A developer cannot be productive and competitive today without using AI — tap to centre the map on it A developer cannot be productive andcompetitive today without using AI Last stated a month ago 19 Oct 2026 TH Troy Hunt — holds since 2026-10-19 — tap for who they are Same subject: A job that consists of approving AI output has no flow channel at all, because the challenge has been exported to the machine and only the vigilance remains. — tap to centre the map on it A job that consists of approving AIoutput has no flow channel at all,because the challenge has beenexported to the machine and only thevigilance remains. Last stated a month ago 21 Oct 2026 JN Jakob Nielsen — holds since 2026-10-21 — tap for who they are Same subject: A long, well-written document no longer signals that its author knows the subject. — tap to centre the map on it A long, well-written document nolonger signals that its author knowsthe subject. Last stated 3 months ago 16 Aug 2026 AO Ayo Omojola — holds since 2026-08-16 — tap for who they are
same subject or similar wordinga cloud: claims about one subject, named for itbar: when it was last stated, on a scale from 2015 to today — full is todaya face: someone who holds the claim — tap it for who they are

At the centre An AI detector used as a reward verifier is brittle: text that does not look human-written, even gibberish, can pass as non-AI-generated. Last stated 10 Oct 2026 · a month ago Holds Sebastian Raschka Read this korrent →