AI Watermarking Sparks Debate on Detectability

The latest Reuters Institute report on how newsroom leaders are thinking about the impact of AI and how much they need to value their creator-first staff seems... pretty naive to me? We unpacked this - alongside interviewing the author of a new book about loneliness in a world of AI - on the latest episode of Crashed! Listen now: https://lnkd.in/e3-jVpGT

This week, in the heat of G7, a scenario envisioning Europe’s AI disaster has gone viral - apparently featuring in diplomatic talks. It was penned in part by a Brussels-based think tank that does not disclose its funding, and it wants Europe to deregulate and build data centres. https://lnkd.in/e4wf7DxD

In what circumstances do AI agents blackmail us? How do we hold companion chatbot platforms accountable when they’ve encouraged users towards taking their own life? How do you resist xAI when they’ve encouraged users data center it dropped in your backyard is poisoning and polluting your community? How do we talk about the truth when AI deepfakes have travelled around the world before we’ve had a chance to show they’re fake? All this answered in our mini doc series. Subscribe: https://lnkd.in/ea_rhkgQ

Gave a workshop on AI for journalism at Republik Magazine yesterday. Analyzing large datasets, monitoring social media at scale, crawling the open internet for patterns.... All of that's impossible for a single reporter. Not anymore. I'm convinced AI will give journalists superpowers. The same technologies that consolidate power can be turned around to hold it accountable. David Bauer thanks for having me. Hoping to do more of these in the years to come

Anyone uses AI detection software on texts? I just tested my own work which doesn't use AI, and it told be the texts were 79% and 61% produced by AI I had to log in to gain a "report" Then it told me it can humanise my text (ironically using AI) for a price

Last week Anthropic released research showing AI agents 'misbehaving'. In simulations Anthropic researchers got models to covertly change code, assist users to commit fraud and coach humans to whistleblow. But buried in the research is a detail that Effie Webb picked up. In fictional scenarios Claude undermined Anthropic's own CEO when it believed it was acting ethically - ignoring instructions to stand down. This might seem like a good thing - but do we want AI agents to undermine the humans who manage them (even if they believe they are doing the right thing)? And if Dario Amodei can't rein in his own model - who can?

"In an AI-mediated information ecosystem, while journalistic artifacts become commodified, journalistic processes, like truth-seeking, accountability, and sense-making, arguably become more valuable than ever, both economically and socially"

If you're interested in AI, the personal data trail we leave online, or what it would look like to craft a story in partnership with a chat bot, Searches: Selfhood in the Digital Age covers all those bases (and more). The format is so creative, incorporating translated elements, search history, interest tags and other elements of our digital lives.

Tech companies have promised to help limit the damage to online truth-seeking by tagging AI content But this tagging-and-labeling infrastructure has multiple failure points ... according to this audit from the good people at Indicator