Replies: 2 comments 1 reply
-
|
Thanks for sharing the prompt @albvar. Really great application of AI too! That is a lot of directions for the model to follow. Too bad the new OpenAI o1-preview reasoning models are not yet available to the API, it'd be very interesting to see how it handles it, it wouldn't be 100%. Also, how large are the VBS files? That too will be challenging for the model. Off the top of my head, breaking this down into multiple steps would be how I'd approach it. Something like, pass in the VBS, with the assumptions and the first three guidelines. Save the results. Then take those results, pass it in, with the assumptions and the next three guidelines. So maybe there would be seven passes. This is defintely where this is "art." Figuring out how the model best works with a complicated VBS + the assumptions + guidelines. The model is an intern that has read everything on the planet. Following directions over a complicated scripts made need more hand holding. Would love to see an example of a VBS. Happpy to discuss approaches. Super use case! |
Beta Was this translation helpful? Give feedback.
-
|
Thanks for that. Good question on the Then you can model the assistant with PSAI to further automate. Good StuffLooks like the VBS is involved. Never tried what you are doing. Early on AI models were not doing PowerShell very well. Also, I don't think there are a lot of places that hold VBS code, like GitHub, so the model training may not be the best. Then again, it looks like VB code, so. AI models have been trained on COBOL, and folks are using it to migrate to Java. You're in real experimental space. I'm also wondering if you could fine-tune a model with a couple hundred examples of VBS -> PS might really help. 80/20 see how far prompting can take you. I would also consider trying all the code+prompt in Claude - cheap to get credits. Sonnet is a but better thatn 4o. Also - we are only a few weeks away from seeing of OpenAI drops GPT-5. Prompt engineering muscle is worth building at this point is. |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
I have several hundred VBS files that need conversion, but I've found using Invoke-OAIChat (gpt-o model) with instructions and a conversation prompt to be somewhat unreliable. The results tend to be hit or miss when it comes to following the instructions accurately. Is there a more effective approach I should consider?
I am sharing my prompt below:
Beta Was this translation helpful? Give feedback.
All reactions