Programming
Prof. Dr. Mirco Schoenfeld
All you need
You know AI, right?
You know AI, right?
You know AI, right?
You know AI, right?
You know AI, right?
You know AI, right?
Let’s find out
Today, we want to find out if AI can help us program.
Task
Let’s start simple.
- Use or download
the dataset from this morning.
- Ask an AI of your choice to generate program code
that
produces a line plot of the data, temperature
on x-axis,
pressure
on y-axis.
General remarks:
- be specific in your prompts
- be systematic in developing / refining your prompts
- document what you tried
- document results
- let the AI generate code that you can execute (e.g. in R)
Points to consider
- Does the AI generate code at all?
- Does the generated code work or does it produce errors?
- Does the code produce correct results?
- Can you read the generated code? Do you recknognize something?
- Does the AI hallucinate?
Task Summary
Did you manage to get it to work?
Does the generated code work and is it correct?
What artifacts did you observe along the way? What were common
errors?
Task
Let the AI escape the maze:
- Download the image (right click) or make a screenshot
- Let an AI generate program code that escapes the maze
- let it pick an escape strategy on its own
- let it implement your strategy
- let it implement a specific common strategy like
Remarks:
- Document prompts and results
- Can you read the code?
- Is the code executable? Is it correct?
Task Summary
Did you manage to get it to work?
Does the generated code work and is it correct?
What artifacts did you observe along the way?
What were common errors?
Task Summary
Would you trust the AI to generate code for you?
Task
With these possibilities and limitations in mind, pick your own use
case and let the AI generate program code.
Task Summary
Would you use this in your daily work?
Task
Discuss how you think about this.
- Is this use case ethical?
- Is it advised?
- Is it clever if you used it?
- What would you recommend a new colleague under your
supervision?
Task Summary
What were your conclusions?
What were your recommendations for colleagues?