Learning to Talk to Machines with Speech Acts
summary
Builds a tiny Clojure language with instaparse grammars and bottom‑up transforms to evaluate numbers vectors and operations, then layers speech‑act primitives: assertions, requests, beliefs (commitments that fire when beliefs become true) and a REPL. Hooks the language to hardware: uses OpenCV to train the AR Parrot Drone to spot a boat, issues drone commands (takeoff, ascend, dance, land) when beliefs flip, demos the system working live, code on GitHub (Babar, clj‑drone).
AI-written summary · about
chapters (17)
- 0:00Start
- 3:28Ready to Code
- 3:49How to build a language?
- 10:27What does it mean to speak?
- 10:58Speech Acts
- 12:02Elephant 2000
- 13:19Answering Queries
- 13:29Accepting Requests
- 15:25Convince
- 15:48Requests with Beliefs
- 18:19Talk to me
- 20:26Don't miss the boat
- 20:43Boat recognition
- 20:54Watch for the boat
- 23:54Demo!
- 24:53How to to Catch a Drone
- 28:24Closing/ Recap
related talks
41mA Clojure fusion of symbolic and data driven AI
33mBuilding Conversational Speech Annotation Tool in Clojure
27mReal World AI Integration
21mLLMs + Clojure = Who needs frameworks?
24mBuilding Intelligible Systems
36mLearning Clojure and CLJS by playing a card game
40mFun with Functional Programming in Clojure
38mIntroduction to Logic Programming with Clojure