The doomsday fantasy of effective altruists: The “AI in 2027” report imagines wild scenarios. Kevin Roose is less impressed. (New York Times)
Hi, I'm Ole Reissmann, a journalist who builds things. I'm the first Director of AI at SPIEGEL. Before that: podcasts, news product development, platform strategy. I write about AI and journalism and send a newsletter you might enjoy.
OpenAI’s latest Ghibli meme trend brazenly exploits Miyazaki’s longstanding criticism of AI. (Brian Merchant, Blood in the Machine)
Vibecoding at Frankfurt AI Forum
With the help of Claude, I vibecoded an interactive summary function into existence. At Frankfurt AI Forum, the newest event of the WAN-IFRA, the World Association of News Publishers, Financial Times' Liz Lohn talked about their successful AI summaries and how they give users the impression of realtime generation. I jumped at the opportunity to...
Is this already ‘agentic’? Google’s NotebookLM now searches and incorporates relevant web content.
Preparing a feast only to have it eaten by ghosts who don’t leave a tip: Wikipedia plans to regulate bot access and wants new attribution guidelines for web, apps, voice assistants, and LLMs. (Wikimedia)
We see clunky AI applications and resist the hype – this article makes the argument that we’re not caring enough about it. (Joshua Rothman, The New Yorker)
Dystopian vision or pragmatic future? Newsquest’s AI-powered content creators stir fears of eroded journalistic standards. (Bron Maher, Press Gazette)
SEO is shifting to crack AI search: focus on repetition and ‘answer-first’ structured content over rankings. (Threads)
AI & Journalism Podcasts That Don’t Suck (I Promise)
Here are my artificial intelligence and journalism podcast recommendations, based on actual listening. Explained: The conspiracy to make AI seem harder than it is! by Spotify's Gustav Söderström. For anyone looking for a good introduction — not a series, no storytelling, just 1.5 hours of solid explanation: This is the podcast you want. It's from...
No Permission, no pay: How my Book became AI training fodder
I was at the Leipziger Buchmesse (Leipzig Book Fair) to talk about AI. First off: I was robbed. Supposedly, Meta (and probably others) trained its AI using LibGen, a shadow library with millions of books and articles. A book that Christian Stoecker, Konrad Lischka and I wrote 13 years ago is part of the LibGen...
Anthropic looks under the hood of Claude 3.5 Haiku, using circuit tracing to see how it works across different kinds of tasks.
SEO, but for AI search engines: Is it really that simple, optimizing for semantic search rather than keywords, just match the pattern of the search query? (Threads)
Inside Google’s Two-Year Frenzy to Catch Up With OpenAI: Late nights, layoffs—and lowering guardrails (Paresh Dave and Arielle Pardes, Wired)
AI search has had little impact on referral traffic to date, based on Comscore’s desktop user data in the U.S. (Nick Hagar, Generative AI in the Newsroom)
Teach AI your face in a few clicks: A beginner’s guide to finetuning FLUX
Low-Rank Adaptation, LoRA, allows for fine-tuning without the need to retrain the entire model. You can teach FLUX a specific style or a specific aesthetic, even a face. A while back, there was a big trend of uploading a dozen photos to some shady app, maybe paying a few euros, and getting a set of...
