Alice's Adventures in Artificial Intelligence-land
summary
A working Slack bot called Alice actually exists: about 10,000 lines of Ruby behind a single Sinatra route, hooked to Slack via bossy, storing user facts in MongoDB + neo4j and calling Google’s natural language API and WolframAlpha for understanding and facts. It implements a pipeline with listener → processor → commander → mediator, a context stack with declarative-indexed facts, and a live-playable dungeon that generates rooms and an SVG map — sometimes flaky but demonstrably functional.
AI-written summary · about
chapters (11)
- 0:00Introduction and inspiration
- 1:34Early AI projects on MUSH
- 3:36The Turing test ghouls
- 5:03Building Alice bot
- 7:15Technical architecture
- 10:09NLP and command parsing
- 12:30Command execution and mediation
- 14:38Context and knowledge gathering
- 18:15The dungeon game
- 23:00Live demo and interactions
- 31:24Future AI plans and conclusion
related talks
43mAI Has Entered The Chat: Humans, AI, and the Art of Collaboration
27mI know Kung Fu! RubyGems for AI
38mCraft your own Generative AI Chatbot with Amazon Bedrock
22mBuilding Alice’s Brain: an AI Sales Rep that Learns Like a Human
44mGlimpses of Humanity: My Game-Building AI Pair
34mPatterns of Application Development Using AI
01h 02mBuilding full stack AI chatbots with Java - livecoding
43mAI w/o Spaghetti: Clean Architecture in the Age of AI