Guidance by the United States Patent and Trademark Office, released on Feb 13: Federal Register Read more... 15 Feb 2024 - less than 1 minute read
A poster on LinkedIn generalizes from his experience with the Gemini Advanced system to all vision models: Vision models are weirdly prone to prompt injection - they are more likely to take (even contradictory) instructions from an image to follow them. Read more... 10 Feb 2024 - less than 1 minute read
description: “Examines AI-assisted legal co-writing workflows, highlighting document drafting, compliance checking, collaboration features, and ethical considerations.” layout: post title: “Legal co-writing” date: 2024-02-24 last_updated: 2024-02-24 tags: [copilot, openai, claude] — Read more... (Updated) - 2 minute read
Peter Gostev elaborates on LinkedIn about how Google Bard includes 5 hidden chat messages at the beginning of each conversation “to get in the mood”. Read more... 28 Jan 2024 - less than 1 minute read
description: “Reflects on unexpected performance regressions in Claude 3, analyzing benchmark discrepancies, user frustrations, and potential regression causes.” layout: post title: “Claude 2.1: Mysterious Disappointment” date: 2024-01-19 last_updated: 2024-01-19 tags: [Anthropic, Claude, Bedrock, Classification, RAG] — Read more... (Updated) - 2 minute read