Interview with Edward Tian, CEO at GPTZero

My career actually started while I was still a student at Princeton. While working in Princeton’s Natural Language Processing Lab, I researched how to detect text written by the AI system GPT-3, a predecessor to ChatGPT. When OpenAI released ChatGPT to the public for free, I decided to take my research and skills and put them to the test, where I created the app GPTZero. When I launched the app, I quite literally had overnight success, and now a few years later, I am still working with the app and company I’ve created as CEO.

I came to this career through my studies in college. I was always interested in technology and really took a unique interest in artificial intelligence while I was studying. Though my company was a first in our area of expertise, one of the biggest challenges we’ve faced along the way is competition.

Competition is a pretty constant challenge. Other AI detection software companies have popped up after us, so we have to prioritize making ourselves stand out. We strive to establish ourselves as clear experts in what we do, and also offer services that simply no other competitor does.

AI is most certainly going to be a if not the top technology trend of 2025. We’re going to see it in even more areas than before, and I think we’re also going to see a lot of challenges where legality is concerned. We might see more laws introduced regarding the use and creation of AI.

There are countless different functions for AI these days – from customer service chatbots, to generative AI, to product testing. Regardless of function, what AI consistently excels in is increasing efficiency. It can help people accomplish tasks faster, run more tests in a shorter amount of time, spot errors with more accuracy, etc. There are plenty of disadvantages too, however. Of course, there are concerns and real scenarios in which AI is replacing human workers, for example. It has also become a top concern for a lot of educators.

The accuracy of our model increases as more text is submitted to the model. As such, the accuracy of the model on the document-level classification will be greater than the accuracy on the paragraph-level, which is greater than the accuracy on the sentence level. The accuracy of our model also increases for text similar in nature to our dataset. While we train on a highly diverse set of human and AI-generated text, the majority of our dataset is in English prose, written by adults. Our detection software works robustly across a range of AI language models, including but not limited to ChatGPT, GPT-4, GPT-3, GPT-2, LLaMA, and AI services based on those models.

Though my company offers AI detection, I choose to take a pretty neutral stance on the use of AI in content writing. I am neither 100% for or against it. What I do support, however, is the ethical use of AI and allowing people to be informed of AI usage when it’s present.

Absolutely. Because of how rapidly it evolves, we may not even know all the ways it will advance next year.

As new AI technology emerges, we always work on fine-tuning our efforts or assessing new features that may need to be introduced to best support the people who need or want to detect AI.

I have quite a few hobbies! I love entertainment – movies, TV, books, you name it. I also really enjoy getting out into nature when I can and connecting with friends and family.

Figure out what interests you. There may not be a specific company or career that you want to pursue that exists yet – perhaps you will be the one to create it. If you can figure out what interests you, your opportunities are endless.

My favorite book is The Hobbit. It’s been my favorite since childhood, and it’s the book I revisit the most!