It didn’t take long for someone to create a detection app for AI generated text. It’s pretty well done, returns a probability score rather than a definitive answer. http://gptzero.me/
The game has begun, the advance, the counter, I doubt it will stop soon.
What interests me more than a computational solution to AI detection is human one. Have you ever tried to learn cooking a cuisine from a culture you haven’t visited, don’t speak the language, don’t understand the common pantry items? I have, and I was reminded of learning to cook Indian food as I started using and learning Generative Art tools. What began as a huge question mark was resolved as I practiced, made mistakes, corrected for them, and pretty quickly I recognized the flavor of individual Indian spices and machine learning models.
Learning is fascinating, we know so much, and still so little about how we learn. We also know that while we can define some commonalities in the way people learn, there are tremendous differences and influences. I think thats why the myth of learning styles refuses to die, we sense that it was at least pointing us in the right direction of understanding that learning isn’t easily defined as it is not the same for all of us.
I was surprised at how little time it took those of us experimenting with AI art to be able to recognize it’s output. Not 100 percent of the time, but good enough. It was incredibly quickly that we made the jump from no idea to recognition. Human brains are very good at pattern recognition, even when we don’t realize we are doing it. What it takes to recognize AI generated art seems to be practice, repeated exposure and experimentation.
We’ve had a few more months to experiment with Generative graphics than with text. Will more time to adapt to AI generated text lead to a similar ability to recognize when it is used? I think its likely. Will AI continue to grow its skills perhaps outpacing our ability to identify it? Also likely.
Does it matter if we can detect whether something is AI generated? I’m sure the answer is some version of it depends but I lean more towards yes than no.
Artificial Intelligence is itself created by humans and we’ve brought our biases, our patterns, our histories into that creation. Given the way we use media and how it can impact our beliefs and understanding of the world I think its important at least at this early stage that we know when AI is involved.