Sponsor - Click to visit; Right Click for samples, personalization, and more offers
Sponsors - Click for samples, personalization, and more offers

How the Heck Does AI Work?

And ways the promotional products industry can harness its capabilities

9/22/2023 | Steve Shannon, Bits & Bytes

Continuing with our focus on AI here at Bits & Bytes, let’s learn a little bit about how AI actually works and then look at some ways we in the promotional products industry can potentially harness its capabilities.

 

Generative AI

“Artificial Intelligence” as a term refers to a wide range of different machine learning models and technologies, but the one most applicable to our industry is called generative AI – AI that produces content such as text, images, video, audio, and even code. Popular AI models like ChatGPT (for text) and DALL-E (for images) are both examples of generative AI. These models accept input commands from a human user and then draw on enormous stockpiles of data from which they were trained to produce what they “think” is the best output – output that can be indistinguishable from what a human might create.

When we say “think”, what we’re talking about is using predictive algorithms to surmise what the next word or image should be in the output. AI models use the initial human input plus hundreds of additional parameters in the process of creating the desired content to calculate what comes next, using its training data as reference. The bigger and better the training data, the more versatilely the AI can behave.

 

Potential Uses

Generative AI can be used to save time from doing mundane or repetitive tasks, or as a creative springboard to spark new ideas. 

Let’s say your company is rolling out a new product, and you need to come up with a quick sales blurb to use as a description of the product on your web site. If you have experience writing these kinds of upbeat-sounding text blocks, you know it can sometimes be a challenge to make them sound hyped without being trite (“Truly, this is the refrigerator magnet you never knew you needed!”).  Lucky for you, generative text AIs excel at generating boilerplate or descriptive text like this super fast. 

For example, here’s what I got when I asked Bing Chat (which uses a modified version of OpenAI’s ChatGPT model) for a sales blurb for a new refrigerator magnet shaped like a UFO: 

 

Introducing the UFO Refrigerator Magnet, a unique and eye-catching accessory that will add a touch of mystery to your kitchen decor. This magnet is designed in the shape of a classic flying saucer, complete with intricate details and a metallic finish.

Made from high-quality materials, this magnet is built to last. Its strong magnetic backing ensures that it will stay securely attached to your refrigerator or any other metal surface. The compact size allows for easy storage and portability, making it perfect for use in your home or office.

This UFO Refrigerator Magnet is not only stylish but also highly functional. It can hold up multiple sheets of paper, notes, or photos with ease. The powerful magnet ensures that your items stay in place, even when the door is opened and closed frequently.

This versatile magnet is suitable for use in various settings. Whether you’re looking to add some personality to your kitchen or office, this UFO Refrigerator Magnet is sure to impress. It also makes a great gift for friends and family who appreciate unique and quirky decor.

Experience the wonder of the unknown with the UFO Refrigerator Magnet. Add this charming and functional accessory to your collection today!

 

Now the length and overtness of that description may be overkill for your purposes, but as a starting point it’s still pretty good, and it only took 8 seconds to generate. You could use it as a template or just as easily take bits and pieces from it to use instead to save you time.

Next, let’s say your company wants to create a flyer to advertise this new product, and you need to come up with a professional looking header image to use for it. You could start from memory and create it from scratch assuming you have the required skill set, or you could sift through an endless scroll of premade examples for inspiration. Or, you could enter what you specfically want into a generative image AI model and see what it gives you. Below is what Bing’s image creator (powered by DALL-E) produced when I asked it to create “a UFO header image for a business email flyer” for our imaginary refrigerator magnet:



That’s not half bad in and of itself for 6 seconds worth of work, but an experienced graphic designer could certainly tweak that to make it more appealing.

You get the idea, but let’s see one more fun, personal example from yours truly.

I’ve been solo developing a classic point-and-click computer game for the last few years, and while I’m no great digital artist I have a lot of experience with pixel art. When I get stuck creatively while designing a scene, I often do exactly what I described above – I search through endless swaths of images online to find inspiration. However, lately I’ve been experimenting with DALL-E to try to cut down on the time I spend searching. Below is a recent pixel art creation of mine that will appear in my game and the AI-generated image I prompted from DALL-E that inspired it:



Considerations

Generative AI models work best when their training data and parameters are fine-tuned through hundreds of thousands of tests and tweaks from human engineers. That’s why AI models are operated by only the biggest tech giants right now, since training and maintaining an AI model takes massive amounts of storage and computing power, not to mention big bucks. 

But despite their pedigree, AIs are far from perfect. Generative models can produce incorrect or biased text, sexist language, and even hatespeech. AI-created images often have issues accurately depicting human hands and words on signs. There’s also the nebulous legal issues dealing with copywrite and plagiarism, which we’ll cover in next month’s Bits & Bytes. 

All told, generative AI works best when used as a supplement – not as a replacement – for human creation. As long as you keep that in mind, you should feel free to experiment with AI to bolster your creativity and productivity.

Steve Shannon has spent his entire professional career working in tech. He is the IT Director and Lead Developer at PromoCorner, where he joined in 2018. He is, at various times, a programmer, a game designer, a digital artist, and a musician. His monthly blog "Bits & Bytes" explores the ever-evolving realm of technology as it applies to both the promotional products industry and the world at large. You can contact him with questions at steve@getmooresolutions.com.
Next up from Bits & Bytes...

CAPTCHA vs. AI

Has AI rendered CAPTCHA obsolete?
Steve Shannon

Computer Productivity Tips

Help improve your organization
Steve Shannon

Is Your Computer Running Slow?

It might be time for an upgrade
Steve Shannon
Latest from PromoJournal...

Microfiber Waffle Caddy Towel from Pro Towels

Make your towel your own
PromoErrday

Niche To Meet You, Infinity Her

Headwear designed for women!
Niche To Meet You

Waffle Weave Beach Towel from Pro Towels

Imprint that goes from edge to edge
PromoErrday