LLM Visualization (bbycroft.net)
Bob Schneier’s post about teasing out by what we mean by ‘trust’
Simon Willison has posted a nice explainer of how ’embeddings’ can be used to build similarity or related finders for both text and image data.
Embeddings: What they are and why they matter (simonwillison.net)
Interesting, but note that some of the researchers are from Google and would benefit from putting holes in a (competitor’s, OpenAI) product.
ChatGPT Can Reveal Personal Information From Real People, Google Researchers Show (vice.com)
Extracting Training Data from ChatGPT (not-just-memorization.github.io)
Ben Werd’s rubric for assessing technology
technology-assessment/Rubric.md at main · benwerd/technology-assessment · GitHub
Why research about AI and research using AI is difficult.
Also, AI Snake Oil substack newsletter by the authors.
Lara Hogan has a template for communicating feedback in a business setting.
Allen Pike has a nice analysis of what uses LLMs and ChatGPT can be put to in their current form, that might work for business situations.