ChatGPT shares internal config file names?

Community
1

Is it normal for ChatGPT to share that it’s having trouble with a specific internal configuration file names?

—-
Me:

…I’m more interested in the process than finding a work around. It appears the image is successfully uploaded and the OCR process appears to start but then fails.

The package in the image is torn. That shouldn’t result in that type of error should it?

ChatGPT:

The error we’re encountering doesn’t seem to be related to the condition of the packaging or the image itself. The OCR process is failing due to a technical issue with the Tesseract library, which is the software used to perform the optical character recognition. Specifically, the error message indicates a problem with locating or accessing the ‘ara.traineddata’ file, which is necessary for recognizing Arabic script.

This file should contain the data Tesseract needs to recognize Arabic characters, and the error suggests that either the file is missing, or the system is not correctly configured to find it. This kind of issue is typically related to the setup of the OCR software on the server and is not something caused by the content of the image.

In a typical user-controlled environment, this could be fixed by ensuring that the necessary language data files are installed and that the TESSDATA_PREFIX environment variable is correctly set. However, in this environment, I don’t have the ability to change the system configuration to resolve this issue.
—-

Notes:

  • This is not reported in the bug forum because I don’t know that it’s a bug.

  • No support needed, Im just interested in how the process works

  • I didn’t share the chat link because conversations with images currently can’t be shared

2

Your prompt technique has resulted in the AI deciding to use code interpreter Python (advanced data analysis) to attempt to do OCR. The sandbox execution environment reports its lack of proper configuration for this.

Instead, write: “Use your built-in GPT-4 machine vision abilities to analyze the contents of this image, extracting the text documentation within for me.”

If you are interested in doing code-based OCR-tessaract interpretation by using your own code, which the AI may assist you in writing: How to OCR with Tesseract in Python with Pytesseract and OpenCV?

1 Like
3

This is very useful! I’ll test that next time I try using OCR. Thank you!

4

So, is that a yes or a no? :blush:

I do appreciate your reply, it could be helpful in getting OCR to work.

However in this case I’m asking specifically, is it normal for ChatGPT to share that it’s having trouble with a specific internal configuration file names?

‘ara.traineddata’
‘TESSDATA_PREFIX’

The boundaries of the sandbox and what’s allowed out is an interesting to understand.

5

Yes, if you ask about how python code that the AI wrote failed, it will be happy to report more information or even try again.

Python coding errors and python environment is not some internal OpenAI secret, nor is the AI talking about the GPT-4-Vision skill you likely intended.