Day 2

ChatGPT-4 Image Processing

2023-10-01

Tags: GenerativeAI OpenAI ChatGPT ImageProcessing CitizenDataScientist CitizenDeveloper

For Day 2, I was planning on doing a gentle introduction to Large Language Models and ChatGPT but this space is moving so fast that I’m instead going to focus on a new feature ChatGPT released.

You are now able to upload an image to GPT-4 and ask questions about it.

The sample I’m using is an old photo I had on my phone of menu specials from a restaurant.

Take note that it does an awesome job of understanding the text in the image.

Although it doesn’t recognize the price for Combo #1 it is able to associate that the multiple menu items all tie to the same price. And it is able to determine that there is no Combo #3 on the menu.

Finally it understood that “Gift Car” was likely a typo and that it should probably be “Gift Card”, which is pretty amazing.

Now let's have it convert the data it has pulled from the image into a single list.

Now let's start turning that list into more structured data.

What have we done?

We’ve taken unstructured data pulled from an image with zero training (no Data Scientist needed).

And we've converted that data into various formats, all without writing a line of code (no Developer needed).

This is just a simple example from an old image I had on my phone. Imagine all of the unstructured data floating around your organization that you could easily convert into structured data.

This is the power of Generative AI.