“Human in the loop” sounds like oversight. Our cognitive biases turn it into a rubber stamp. Damon Kiesow on how to make checking AI output hurt a little, like red-teaming and forced rewrites. (Working Systems)
“Human in the loop” sounds like oversight. Our cognitive biases turn it into a rubber stamp. Damon Kiesow on how to make checking AI output hurt a little, like red-teaming and forced rewrites. (Working Systems)
The forensic heatmap supposedly exposing a photo as fake? Also fake. Forensic cosplay to create doubt about real photos. (shirin anlen, Mahsa Alimardani, Tech Policy Press)
Anthropic asked 81,000 Claude users what they want from AI. People in lower-income countries are more optimistic than those in Europe or North America. Worry about job loss is the clearest predictor of negative sentiment. Qualitative research at this scale was hard to do before. Whether Anthropic should be the one doing it is a question the report doesn’t ask.
Google is rewriting news headlines in search results. “Copilot Changes: Marketing Teams at it Again.” That’s not a headline a journalist wrote. Google did. Without telling anyone. (Sean Hollister, The Verge)
Journalist suspended for publishing AI-hallucinated quotes: A former editor-in-chief of NRC used ChatGPT, Perplexity, and NotebookLM for his newsletter, but didn’t verify quotes. He had warned colleagues about exactly this. (Dan Milmo, Guardian)
The use of as many AI tokens as possible, as fast as possible. Enabled by agentic coding tools that spawn subagents running unsupervised. Writing an essay uses around 10,000 tokens. Tokenmaxxers blow through billions a week, burning the planet along the way. Fat token budgets have become a job perk, and some rack up monthly...
A 47-step AI tool that writes nothing: Bauer’s internal content briefing system pulls articles from competitors, looks at rankings, and identifies content gaps. The journalist still has to write the piece. “The premium is on expertise and authority. We wouldn’t ever want to do anything which dilutes that.” (John Rahim, The Media Stack)
The term for pasting raw, unread AI output into a conversation, shifting the work of reading, verifying, and distilling onto whoever receives it. It's kind of rude, really, with real costs: eroded trust and frustration for recipients as the behavior becomes more common. The fix: read it, verify it, cut it down, disclose it, and...
Schibsted's Videofy is now open source: The tool pulls a published article, writes a script, matches footage, adds a voiceover, and hands editors a finished video.
Image Verification Assistant: A joint CERTH-ITI and Deutsche Welle tool for image verification, metadata analysis, and reverse image search. Alpha stage, open source.
Three roles for AI in journalism: source, colleague, assistant. A framework that’s more useful than another round of newsroom culture war. (Stephen J. Adler, Columbia Journalism Review)