AI, please make me look good!

Recently, I read an article that ChatGPT has improved its ability to create images. This piqued my interest. I enjoy creating images and graphics for personal and professional use. However, the extent of my design skills is limited to passing critique on what I like or don’t like. So, I decided to try it out!

I used ChatGPT‘s public and freely available website and posed the following question:

Create a photo realistic graphic of Ben Woo of Neuralytix standing behind a podium with a Neuralytix powerpoint presentation behind him in landscape orientation (sic.)

I also uploaded the currently logo for Neuralytix to see if the Artificial Intelligence (AI) engine would incorporate it into the image. After a few minutes, ChatGPT returned the following result:

The result: very impressive, but I gained a lot of weight

I must admit, I am very impressed with the result. With just the “question” I posed to ChatGPT, the AI engine was able to create image of:

• An Asian man with steel rimmed glasses (in my opinion, satisfactorily Chinese),
• Dressed in a suit, which is my preferred dress style,
• Speaking to a proverbial audience,
• The original title I held when Neuralytix® was founded in 2012,
• The Neuralytix website, and
• A logo that is remarkably similar to a very early draft of the Neuralytix logo.

For me, the only two objections to this first pass of using AI to generate this image are:

• It looks like I have a receding hairline (which thankfully I do not), and
• The man depicted, is 3 to 5 suit sizes larger than I am in real life.

ChatGPT offered me the opportunity to provide feedback to this image to improve the accuracy of what I am looking for. While I did not take up this offer, the capability of ChatGPT to generate this image far exceeded my expectations. Although I only tried ChatGPT, I am confident that other image generating AI engines are equally capable of generating images of similar accuracy and quality.

In this singular example, it shows many of the underlying benefits of AI:

• Accelerated outcomes,
• Low (in this case, no) cost, and
• Assisting users to do tasks for which they do not (yet, or want to) have the skills.

Update

Several days later, I asked ChatGPT the same question again. It generated a different image, albeit similar, and without any references to Neuralytix. This time, I followed up with a few requests for revisions. I asked it to make me skinnier, replacing what was on the screen with charts and graphs, and put me in a setting of a large audience before exhausting the limit of revisions I am allowed for the free version of ChatGPT.

ChatGPT regenerated an image that responded to my additional criteria. However, I think one of my request for revisions was to give me a full head of hair, because in the final image, I had even less hair than the one above. And no, I’m not going to post the final outcome!