My 3 insights when making GenAI applications
1. *Consult a Domain Expert:*
Engage a domain expert to think out loud about all small details related to the GNI application.
Encourage the expert to articulate their thought process clearly.
Document all the details provided by the domain expert.
2. *Update Knowledge Base:*
Regularly update the knowledge base as the reality is constantly changing.
Ensure that domain experts are involved in updating the knowledge base to work alongside AI systems.
3. *Testing Prompts and Changes:*
Develop a process to test prompts and changes effectively.
Use a cloud platform for domain experts to test different models and assess the quality of generated content.
4. *Testing Prompts:*
Use ideal answers based on historical data or human input to test prompts.
Run different models to test the prompts and compare the accuracy of each model.
Select the model that works best for the specific use case.
5. *Optimizing Prompts:*
Experiment with different versions of prompts to find the most effective one.
Gather feedback on the highlights of the generated content and make necessary adjustments.
Cautionary Notes
Domain experts may be hesitant to update prompts due to fear of errors. Encourage open communication and provide support throughout the process.
Ensure that all changes made to the GenAI application are thoroughly tested before implementation to avoid any disruptions.
Tips for Efficiency
Encourage collaboration between domain experts and AI systems to leverage the strengths of both.
Maintain clear documentation of all details provided by domain experts for future reference.
Regularly review and update the knowledge base to stay aligned with changing realities and requirements.