Reflections of AI: A Trilogy in 4 Parts
summary
Summarizes recent evidence that generative AI rarely delivers promised productivity gains: controlled study claims 55.8%, but large analyses find 4% productivity boost and only 1% of AI-generated code ships without significant rework; users report ~2.8% time savings (~100 seconds/hour). Agents can save ~30 minutes only in heavily tested, low-debt repos; retailers use RAG for better recommendations; a Danish ML breast‑cancer system detects more dangerous tumors and reduces false recalls. Advice: stop automating broken processes, build small validated pilots, test in production.
AI-written summary · about
chapters (8)
- 0:00Intro
- 1:13Agenda
- 2:201. The impact of GenAl adoption
- 9:112. The [hidden] price of Al adoption
- 18:283. Useful Al in a world saturated with LLMs
- 23:334. How to build [valuable Al] solutions
- 27:45Takeaways
- 31:03Outro
related talks
51mAssessing AI's progress
18mHumans + AI: The Future of Software Development Teams
28mProductivity: a dialogue
41mPanel discussion: AI in code development – hype vs reality
27mSoftware engineering with GenAI
35mProductivity in the age of AI
52mPanel: Productivity gains through AI?!
39mAI-assisted development: Common pitfalls and smarter practices
