In a world where artificial intelligence is increasingly woven into everyday life, distinguishing between human-generated content and AI-driven creations can be a tricky task. From chatbots and virtual assistants to deepfake videos and automated news articles, AI is shaping how we consume information and interact with technology. Whether you're a professional navigating AI-powered tools in the workplace or just a curious consumer, understanding how to spot AI can help you make more informed decisions. This guide will equip you with the key indicators to look for when identifying AI content and provide tips for staying one step ahead in the digital landscape.
Remember how last year we posted a blog that was entirely written by AI? Yep...AI just wrote that introductory paragraph. It's a little scary how far technology has come, isn't it? At a single glance, you can easily be fooled by an AI-generated deepfake, but with a little observation and critical thinking, you can learn AI's patterns and shortcomings so that you'll be able to more easily identify what is artificial and what is created by a human being. Not such a bad skill to have in today's chaotic world!
Once you've finished up, be sure to register for our upcoming live webcast on Tuesday, November 12: "Artificial Intelligence and Its Influence in Society." Get ready for a deep dive into the fundamentals of AI, unravel its intricate workings, and behold its myriads of applications across domains including law, medicine, and business. We mean it when we say this is one course you don't want to miss.
And if this subject interests you, keep an eye out for future blog posts all about AI!
Spotting Deepfakes. The very idea of false images can be quite unnerving. Imagine someone trying to use one in a court case and it slips under the radar! AI-generated art is also viewed unfavorably within creative communities because takes away jobs from real artists. So how do you distinguish a real from a fake? As they say, the devil's in the details, and that's where AI struggles the most. Specifically, hands, skin texture, and facial features (especially teeth) can wind up looking distorted. The background could also be warped or the lighting unrealistic.
Here's a recent example of a two deepfakes that went viral: a little girl, a victim of Hurricane Helene, clutching a puppy on a boat. Look closely. In one picture, she appears to be missing a finger, while in the second picture, she has grown an extra one. Her skin looks a little bit too smooth while her expression is so exaggerated it borderlines cartoonish. And if you compare the two pictures against each other, you'll notice that the color of the puppy and the boat changes.
AI's shortcomings with imagery are further emphasized in video. It struggles with natural human movement, changing expressions, moving colors or shadows, and more. However, AI is learning all the time and there may come a day when it's able to perfect its weaknesses. Until then, approach dramatic, viral images with a critical eye for detail.
Uncovering AI-written content. This one might be a bit trickier as AI grows more complex with each passing day. AI lacks nuance so it will often fail to grab the larger context of what it's writing. It has the tendency to get wordy, repeat the same information, or reuse the same sentence structures without worrying too much about how it will flow. As I pointed out last time, it has lots to say with little substance.
It also doesn't always get its facts right, either. Look no further than the time Google AI suggested putting glue in pizza or eating one small rock per day for digestive health. Delicious. (AI will never replicate my sense of humor, either).
Detecting AI-generated voices. That's right: AI even has the capabilities to learn voice mimicry. Sometimes it's used innocently, like posting a video of a popular character singing a silly song. Other times, it's downright frightening, as there have been reports of people using AI's voice-cloning technology to learn the voices of our loved ones and use them in a scam to steal money. If you ever find yourself on the receiving end of this scam, take a moment to breathe and really listen. AI voices have difficulty with emotion and the pronunciation of more uncommon words. They might also have unnatural pauses. The perpetrators in these scams only use a small sample of AI because they know too much might give away that this isn't a real person.
As always, listen to your gut if something strikes you as strange or unnatural! All of this is why it's more important than ever for lawyers and lawmakers to stay on top of this fast-changing technology.
Hey ChatGPT...did I miss anything? Let's ask AI directly if it has any other advice for us:
"For spotting AI-generated text, specific tools or software can help detect AI-written content. Resources like plagiarism checkers that include AI detection capabilities would be helpful."
Thanks for the tip! If you're interested in such a thing, ChatGPT recommends GPTZero, which is specifically designed to detect AI-generated text, including its own content. It pays special attention to patterns, perplexity, and variance in sentence length; something humans have higher amounts of.
I'll leave you with this conclusion written by ChatGPT, which is certainly different from the usual tone and cadence of Raising the Bar. With the help of this blog, hopefully you can continue to hone your skills in identifying what's "real" and what's "artificial!"
"As artificial intelligence continues to advance and seamlessly integrate into our daily lives, the ability to recognize AI-generated content is becoming a vital skill. From discerning fake images and videos to spotting AI-written text and cloned voices, staying informed helps us navigate an evolving landscape where the line between human and machine is becoming ever more blurred. Remember, even as AI grows more sophisticated, it still has limitations that give it away—whether in distorted details, unnatural phrasing, or awkward voice inflections. By sharpening our observational skills and utilizing available detection tools, we can better safeguard against misinformation and protect ourselves from AI-based scams. Stay vigilant, stay curious, and continue learning about AI's impact to stay one step ahead in this rapidly changing digital world."