It is very interesting to see how DSPy is working towards to Prompt Programming, I feel like Prompt Engineering is pain task with different AI models and maintainability
The risks of Prompt programming include: 1. DSPy can't guarantee to generate appropriate prompts; 2. can't handle complicated situations; 3. don't know how to fine-tune.
Although fascinating and exciting, the amount of additional work you have to do to get results that are roughly 10 points better according to the slide on 35:08 is quite discouraging. You have to desperately need those gains to justify all the work.
Great presentation. I found dspy complex before and wrote it off, but this really helped me understand it! Will try it out again
It is very interesting to see how DSPy is working towards to Prompt Programming, I feel like Prompt Engineering is pain task with different AI models and maintainability
The risks of Prompt programming include: 1. DSPy can't guarantee to generate appropriate prompts; 2. can't handle complicated situations; 3. don't know how to fine-tune.
Although fascinating and exciting, the amount of additional work you have to do to get results that are roughly 10 points better according to the slide on 35:08 is quite discouraging. You have to desperately need those gains to justify all the work.