1. Goal
Try to understand the trending AI tool
- ChatGPT
- Github Co-pilot
Decide
- Should use co-pilot or not
2. Copilot
From comment to code
- Use GPT3, trained to natural language
Use the whole github dataset and fine-tune in a supervised version
3. Chat-GPT
- Use the GPT3.5 and fine-tune it on conversation dataset (human labeled)
- Train on a reward model (human labeled)
- Optimize the policy
4. GPT-3
Super large model