Normally, my post images are there to add character. This one illustrates the image identification capabilities of ChatGPT 4o and why they’re important!

This is an example I illustrated in May’s webinar for ACEDS (available here) where I illustrated several examples (many of them live) on how to use ChatGPT and generative AI models for legal writing. Some of those same features can be important for other tasks as well, including ChatGPT 4o’s ability to work with non-text formats, such as images, audio and video files – i.e., it’s “omni” features that represent the “o”!

The image you see above represents some items on my kitchen counter several weeks ago. Nothing remarkable about it, except that it’s a bit more cluttered than it should be.

However, from an eDiscovery standpoint, any picture image is inherently not useful as is. Other than the name of the image potentially identifying what’s in it, there’s nothing you can use to automatically determine whether the image is potentially responsive to the discovery request. Images of pictures have historically required manual review to make that determination.

Until now.

With its image identification capabilities, ChatGPT 4o can tell you what’s in the picture – and it’s remarkably accurate!

In preparing for the webinar mentioned above, I tested out several of the new features that ChatGPT 4o promised. While GPT 4o wasn’t able to do certain things like transcript audio files when it was first released, the image identification capabilities worked right away! As you can see below, I first asked GPT 4o if it could tell me what was in the picture, and the model confirmed it could do so. Then, I uploaded the picture you see above and asked it to tell me what was in the picture. Here are the results I got:

Pretty impressive, right? It recognized that it was a kitchen countertop and listed eight items. About the only thing it missed was that there were four containers associated with item #1 (two containers in addition to the peanut butter and Nutella, not one). Other than that, it recognized the items and even brands of items (Cosori, Nutella, Kitchen Aid)!

Why is that important? Because that’s a capability that can streamline image classification for eDiscovery. The image identification capabilities of GPT 4o could be used to identify and classify the images without requiring each one to be manually reviewed. That’s huge!

Of course, it’s not perfect and won’t always provide the same answer. I did the test above before the webinar in May. I re-ran the test last night with the same picture. It gave me mostly the same items (albeit in a different order), and it successfully identified all four containers on top of the bread box! However, this time, it failed to mention the electrical outlets.

Is that a big deal? Maybe – nothing is going to be perfect. But the image identification capabilities of GPT 4o could certainly still save a lot of time in classifying images. That IS a big deal!

My test last night also included a conclusion from GPT 4o that wasn’t provided before: “The countertop appears to be clean and organized.” Hah, that’s a good one GPT 4o! 😀

So, what do you think? Do you think image identification will be a key genAI feature for eDiscovery? Please share any comments you might have or if you’d like to know more about a particular topic.

Disclaimer: The views represented herein are exclusively the views of the authors and speakers themselves, and do not necessarily represent the views held by my employer, my partners or my clients. eDiscovery Today is made available solely for educational purposes to provide general information about general eDiscovery principles and not to provide specific legal advice applicable to any particular circumstance. eDiscovery Today should not be used as a substitute for competent legal advice from a lawyer you have retained and who has agreed to represent you.