SI recently started the workshop on Claude AI for Software Development where I teach how to responsibly use AI in creating an application from ideation phase to production.
Some of the caveat I found from learners' understanding is that they of AI Models as Super-being which can automatically understand our requirement and develop an effective application. My answer to this is - NO.
AI models are designed to handle the complex task but they need clear input from its human counterparts. They are designed to work with humans in enhancing the application shipment from years to months to weeks. Humans in the loop to provide clear requirement of their goal, with step by step instructions how to reach there, suggest some examples which makes the model understand what we are looking for by exploring those examples.
**This is called Effective Communication with your AI model - it follows the below principle.
- Define your goal - what you want and be very explicit of your goal requirements.
- Discover the tasks which can be offloaded to the AI model and what type of AI model fits the requirements, may be break your tasks into subtasks and find suitable models.
- Design the tasks with explicit examples, architecture designs, specific code strategy to use.
- Develop your app along with AI, review the tasks completed by AI, provide feedback and iterate.**