Although I do like Anthropic's AI Fluency course: anthropic.skilljar.com/ai-fluency-f...

💯!

Got you. ‘Stochastic parrots’ in particular has become a mantra against engagement with AI & it’s frustrating/seems to blind people to tools that can be incredibly useful.

🙏🙏🙏 Huge thanks to BCS Women and BCS Publishing for hosting the event.

Fair point, but the original tweet did make it sound like GPT-5 had actually solved those problems, which it hadn’t. The constant AI boosterism is wearing thin. At this point, every “breakthrough” just makes me ask, what’s the catch this time?

...Martin Delahunty interview on AI in publishing, article on the future of PDFs by Hetzscholdt and ChatGPT-4o, Timo Hannay interview on Midnight at the Casablanca, and finally Hannah Shelley’s “Google Scholar Is Doomed”.

...Paradigm AI-powered spreadsheet, Academic Search on Bluesky by Aaron Tay, fake citations involving Liudmila Zavolokina, Wiley–Anthropic content deal video, em dash satire piece, Roald Dahl’s “The Great Automatic Grammatizator”, Rick Anderson on Subscribe-to-Open challenges...

...FT CMS auto-generated image descriptions, Science Advances study on AI-influenced biomedical language, membership inference vs dataset inference, WASP 2025 workshop on AI in scientific publishing, publication alert systems NEP and BIMs, responsive vs static digital publishing models...

... OSF Preprints submission suspension, Near Missives newsletter launch, KGX3 engine from Thomas Kuhn Foundation, posters.science launch, Copilot Mode in Edge with multi-tab RAG, AlphaXiv blog-style summaries for arXiv papers, Pangram GenAI detection in Oversight, Veracity tool from GroundedAI...

My overwhelming initial thoughts are somewhat cynical - this model is the result of someone being asked to reduce costs dramatically. They've hobbled the search, and you only drop into more sophisticated models when deemed absolutely necessary... (ok, often this was a waste of compute, but still).