Semantic Commit: Helping Users Update Intent Specifications for AI Memory at Scale
summary
SemanticCommit is a mixed-initiative interface that detects semantic conflicts when users add new intent to natural-language intent specifications and lets users resolve them locally or with AI. It uses a knowledge graph-based retrieval-augmented generation pipeline (custom KG + LLM edge induction) and boosts recall over naive string-replace or full-document rewrite. The prototype and a 12-participant within-subjects study (vs OpenAI Canvas) show half the participants adopt impact-analysis workflows and report greater control with no increase in workload.
AI-written summary · about
chapters (8)
- 0:00Introduction and motivation
- 2:09Intent specifications for AI
- 3:38Semantic commit concept
- 5:25Demo of Semantic Commit
- 7:30Impact analysis workflow
- 8:32User study results
- 9:58Key takeaways and future work
- 11:42Q&A session
related talks
46mBuilding AI solutions with Semantic Kernel
32mOpenSearch Agentic Memory: Intelligent Context for AI Agents
43mRedefining Human-AI Collaboration
26mSemantic Kernel: Teaching AI to Not Just Repeat, But Complete!
53mUnleash the power of AI: Automate tasks with OpenAI and Semantic Kernel
30mBuild an AI knowledge fabric for your organisation
23mPanel
36mSemantic Kernel: the magical way of integrating LLMs into you