Can Test Driven Development be speeded up with Generative AI?
summary
Experiment uses ChatGPT API with GPT‑3.5 on Google Colab, 16 participants given 40 minutes to complete a TDD task using two interaction patterns and crafted prompts. Fully automated AI runs finish faster but produce far fewer tests, while the collaborative pattern yields more assertions than solo human TDD. Generative models tend to solve whole problems in one shot and hallucinate, so outputs require heavy supervision. Recommendation is to write small methods and keep tests tiny and incremental to reduce AI errors.
AI-written summary · about
chapters (8)
- 0:00Introduction to talk
- 1:56Limitations of generative AI
- 5:30AI for software development
- 6:56Test Driven Development refresher
- 7:36Problem statement and approach
- 9:22Experiment setup and results
- 11:50Conclusions and observations
- 12:59Future research and Q&A
related talks
43mBecome a 10x developer by harnessing the power of Generative AI
36mUnleashing Generative AI: Improving Developer Productivity in SRE
01h 26mAutomation of the Automation using GenAI
46mTDD & generative AI - a perfect pairing?
40mDavid Fogl: Leveraging Generative AI for Enhanced Software Testing
45mTDD & generative AI - a perfect pairing?
45mTDD with AI?
32mHow to Make AI Work for You: A Deep Dive Guide into ChatGPT for IT Professionals