Hello, just wanted to ask is there a way to validate the correctness and consistency of the uml models generated by the chatgpt. Is there a tool for it ?
Perhaps. I have another video that I made a little while ago in which I demonstrated how to use GPT to generate unit tests used to validate the code it generated. And since then GPT-4 has added their beta Code Interpreter model that will unit test and execute its own code until the tests pass. (Currently only for Python.) It’s getting better and better each day. But as I said in the video, the human developer or architect is ultimately accountable for the quality of anything generative AI creates. You can use it as a stimulus for ideation, but you have to refine the results for accuracy. Also, the quality of the output really depends on the quality of your prompt. You can now feed GPT large files to give it more context. The more context, the better the results. This video (already outdated) was just a super quick demo of what’s possible. Actual UML generation should be iterative. Hope that helps!
Heheh, yup! As they say, garbage in, garbage out. The original prompt was pretty lacking. That’s why validating any output is essential. But not too shabby for a bit of ideation. And I’ve noticed GPT-4 has gotten a lot better since this video was made. (That, or my prompts have gotten better 😅)
Good job ,thanks
TY this is pretty much exactly what I am trying to do. Subbed.
Thank You
Hello, just wanted to ask is there a way to validate the correctness and consistency of the uml models generated by the chatgpt. Is there a tool for it ?
Perhaps. I have another video that I made a little while ago in which I demonstrated how to use GPT to generate unit tests used to validate the code it generated. And since then GPT-4 has added their beta Code Interpreter model that will unit test and execute its own code until the tests pass. (Currently only for Python.) It’s getting better and better each day.
But as I said in the video, the human developer or architect is ultimately accountable for the quality of anything generative AI creates. You can use it as a stimulus for ideation, but you have to refine the results for accuracy. Also, the quality of the output really depends on the quality of your prompt. You can now feed GPT large files to give it more context. The more context, the better the results. This video (already outdated) was just a super quick demo of what’s possible. Actual UML generation should be iterative.
Hope that helps!
@@impostersyndromedev Thanks for giving a direction to think. Appreciate the response.
that sequence diagram looked pretty extremely wrong
Heheh, yup! As they say, garbage in, garbage out. The original prompt was pretty lacking. That’s why validating any output is essential. But not too shabby for a bit of ideation. And I’ve noticed GPT-4 has gotten a lot better since this video was made. (That, or my prompts have gotten better 😅)