By Annabelle Whittall, Head of Technical at The Cloud Awards.

 

This year marks the launch of our brand-new AI Awards program here at The Cloud Awards. The deadline is next month, and entries are already pouring in. It’s safe to say that artificial intelligence is our favorite topic right now.

In the last 12 months we have adopted several new AI tools which we are now using day-to-day in the running of our business. We are delighted to be receiving your AI Awards submissions and marveling at your remarkable applications of AI throughout the tech industry.  

In fact, the newest member of our Cloud Awards team has AI in her veins. Cate, our AI Assistant, is completely generated by artificial intelligence. Her face, voice, expressions, and “thoughts” have all been machine-created and, with each upgrade and new iteration, we get to watch her expand and grow, and take on a more active role in our business.  

Suitably impressed by Cate, and in dire need of a more professional image for my LinkedIn profile, I turned my attention to an AI headshot image generator to see what it could do.

What are AI headshots?

Although you might not have used an AI headshot generator before, you’ve probably seen creative examples on social media. You may have seen a picture of your friend styled as a Game of Thrones character, or a family member “aging” virtually. These pictures are impressive and fun to play with, but are not suitable for work use. I wanted to see whether the technology was ready to provide me with professional headshots, at a fraction of the price of real photoshoot.

AI headshot generators use machine learning algorithms to create realistic images of faces. These tools analyze thousands of photos to understand the nuances of human features, expressions, and realistic lighting. By leveraging deep learning techniques, they can produce images that closely mimic those taken by professional photographers. The process typically involves uploading a set of photos that the AI uses to learn your facial features before generating new images based on this data. 

My trip into ‘Uncanny Valley’

I scouted a few websites before settling on Instaheadshots.com where I plumped for their Starter Package, priced at $39 USD for 40 HD headshots.  

The first step was to provide Instaheadshots with some basic information about myself and then upload at least 12 high-quality selfies to my brand-new profile. (It was no trouble at all to provide 18!) 

The selfies needed to be recent, well lit, without caps or hats and extreme make-up styles. This was straightforward to achieve, and I was able to use selfies that I already had.

I selected two possible backgrounds for my headshots: city scape and professional office. There were many other options available, including academic settings such as a library, and some medical / laboratory settings (with white coats and scrubs, naturally). 

The technology behind AI headshot generators is continually evolving. As AI becomes more sophisticated, future advancements could address current limitations, offering even more realistic and customization options.  

It took around one hour for my headshots to generate, at which point they were emailed to me. I was surprisingly pleased with the results. Every single headshot resembled me to a degree, but some of the headshots looked exactly like me. In fact, it was incredibly surreal. 

The generated images are in the style of hyperrealism, with flawless skin, perfect hair, and a smile that finishes at the lips, with no hint of expression in the eyes at all. I found myself staring at 40 impeccable pictures of myself (three of them below!), most of which brought to mind two words: “uncanny valley”.  

Why do AI generated images give us the creeps?

In a recent study Karl F. MacDorman and cognitive psychologist Alex Diel discovered strong evidence supporting a theory known as configural processing. This theory suggests that reactions associated with the uncanny valley are primarily triggered by our sensitivity to the arrangement and proportions of human facial features. Another related theory, perceptual mismatch, posits that this discomfort arises when we perceive inconsistencies such as realistic eyes combined with unrealistic skin. This type of incongruity is frequently encountered in stable diffusion-generated AI images.

And so, AI generated images feel weird, because they are weird. There is no intentionality and some details “chosen” by the AI do not align well with previous decisions it has made.

With proper time and human intervention, AI is capable of producing incredibly detailed and impressive images, such as the breathtaking Théâtre D’opéra Spatial  by Jason Allen, which recently scooped first prize in the Colorado State Fair Digital Arts Competition. Jason spent over 80 hours carefully curating his entry in MidJourney and then adjusting his favorite AI-rendered images in Photoshop.

Creating the ‘perfect’ headshot

But it seems that your perfect headshot needs to be a little less than perfect, if you want it to remain relatable and approachable, whilst still impressing your professional peers. As with any other application of artificial intelligence, human intervention is necessary if you want to create AI generated imagery that blends seamlessly with human expectation.

And so, I am left with 40 headshots that are a little too perfect and require some people-powered adjustment. But that’s okay, the best thing about AI is working with it to create something I couldn’t have done on my own.

Recognizing the latest breakthroughs in artificial intelligence

Do you have an innovative A.I. solution that deserves to be celebrated? Whatever your A.I. product, service or innovation, The A.I. Awards program, from the team behind the Cloud Awards and SaaS Awards, will bring the recognition your organization deserves.

The A.I. Awards is an awards program accepting submissions with a July deadline. Partnered with the Cloud Awards and SaaS Awards, which celebrate the brightest and the best in Cloud Computing, The A.I. Awards accepts entries from across the globe.

Any A.I. solution, or any practical application of A.I. technologies, is eligible. However your A.I. solution is delivered, our A.I. awards program is open to consider your solution. Entry to The A.I. Awards closes at the end of July.

About the Author: Annabelle Whittall

Annabelle Whittall is an experienced Systems Engineer and is Head of Technical at The Cloud Awards. Annabelle has worked for over a decade in research and development alongside companies such as Rolls Royce, Boeing, QinetiQ, BAE Systems and Lockheed Martin. She has worked on the development of a state-of-the-art satellite navigation capability program, and on a cloud-based solution in single end-to-end geospatial capability and AI. Annabelle is an active STEM ambassador and a member of the Women’s Engineering Society. She is passionate about what the cloud can do for environmental sustainability.