Table of Contents
Introduction
Welcome to the first edition of Discovery Log, a series where I'll share interesting resources and tools I've recently discovered. As AI continues to evolve very quickly, this edition focuses on two emerging LLMs: Google NotebookLM and Kimi.ai.
Note: This is the first edition of Discovery Log, keeping it short and sweet to start things off.
AI Tools
Google NotebookLM
Google NotebookLM is a powerful research and note-taking application that leverages Google Gemini to help users interact with their documents intelligently. The platform generates summaries, explanations, and answers based on your provided sources, supporting a wide range of formats including text files, PDFs, Google Docs/Slides, websites, audio files, and YouTube URLs.
The application features a three-panel design:
- Sources Panel (Left): Upload and manage up to 50 different sources.
- Conversation Panel (Middle): Engage with the LLM by asking questions about your sources.
- Studio Panel (Right): Generate various types of content including syntheses, FAQs, guides, and quizzes. You can even create audio content featuring conversations between two people discussing your notes. While still in beta, the Studio currently only supports English output.
Kimi.ai
Following DeepSeek's success, China has introduced a new model called Kimi K1.5, available completely free without usage quotas. While Kimi.ai functions similarly to other LLMs through a web interface, it particularly excels in Chinese and English language processing.
Key features include:
- Document and image upload capabilities
- Real-time internet search integration
- Competitive performance metrics
According to benchmarks, Kimi claims to outperform GPT-4, Claude 3.5 Sonnet, and DeepSeek R1. While these results are impressive, I encourage you to test it yourself and form your own opinion.
