How subsidiarity guides open source AI development

MediaLaws is sharing the call for papers on “Disinformation after Generative AI and Synthetic Data” for Information Polity Journal, published by SAGE. Contributions should aim to unpack the method, historical and regulatory context of GenAI and disinformation. Interdisciplinary is welcomed. For more info on topics and timeline, click here: https://lnkd.in/dR8C7BYq

At All Things Open Conference, Katie Steen-James and Nick Vidal from the OSI will provide an update on the latest policy developments and conversations at the intersection of Open Source and AI. https://lnkd.in/dNjWevcf

Policy, openness, and AI are becoming increasingly intertwined. As a community of technical writers, these conversations are vital to shaping the frameworks that guide how we build, document, and communicate technology. We should all look forward to hearing the Open Source Initiative (OSI) perspective, an essential discussion for anyone creating and sharing knowledge in this fast-evolving AI era. The choices being made today around AI transparency, licensing, and open collaboration will directly influence how we: • Access information • Attribute sources • Maintain ethical and inclusive documentation #AllThingsOpen #OpenSource #AI #TechWriting #WriteTechHub

The few things I learned from this module- 1. What is LLM and Generative AI? 2. How are the growing interests by scientists, collaborators and editors increasing with the time? 3. what are the pros and cons of using Generative AI? 4. Upto what level and how Elsevier is allowing it's writers to use AI under human observation? 5. What are the regulations are being made in China, European Union and America.

I wrote up my review of Karen Hao's recent book, Empire of AI, which is all about OpenAI, Sam Altman, and LLMs more generally. Very well written and researched. Quite scary conclusions. She makes a convincing case that modern AI companies are closer to European countries a few centuries back in their attempts at empire-building and colonizing vast swaths of the public. https://lnkd.in/egEDfZQg

"We need a very clear story of what a British trajectory on AI specifically looks like," Kanishka Narayan MP on the role the UK can play in the global AI race. ⏯️ http://bloom.bg/42SiZTy

I wonder what percentage of these AI-generated articles are really from humans who are increasingly writing like AI. And the article this figure comes from acknowledges that human-written articles might be undercounted because of paywalls. But still, the slop is rising.

Introducing Blog Spotlight, a new series featuring our favorite posts from Cleo’s technical blog! 🔦First up: Memory as a step toward more human AI. We built Cleo 3.0’s memory system to solve an issue most humans are pretty familiar with: forgetfulness. Using semantic insight retrieval and structured metadata, Cleo can now recall what matters from past chats, without dragging your entire conversation history into every prompt. It’s a crucial part of how we’re teaching AI to understand your full financial story. Read the full piece → https://lnkd.in/ej2xQE9P

I've been quietly (maybe not so quietly) working on something since June. Our Tavus Sparrow-0 turn detection model is pretty good. It considers semantics and lexical structure to determine when a turn is done. It's fast too, 20 ms inference latency. But our goal, when we set out was to create truly human AI. So, we're taking another crack at it. Here is an early preview of what's next.

In the article it was mentioned that a Surfer AI detector was used to determine whether text was written by a human or by AI. Out of curiosity, I decided to try it myself: I pasted in some text I had generated with AI (using my own prompt). The result was fascinating - the detector was 100% certain it was written by a human. Now I wonder: what are the actual proportions of AI-written versus human-written content on the web today?