Eye On AI
In this episode of the Eye on AI podcast, we had the opportunity to sit down with Jonathan Gillham, the founder of Originality AI. Originality AI is a company that specializes in AI detection technology, specifically focusing on identifying AI-generated content.
JONATHAN: People wish the world was clean sometimes and it was, just give me a yes or no and I want to rely on that with 100% certainty. The world is messier than that with AI now. People need to understand that detectors are not perfect, and the flip side of that is that we can't just dismiss detectors because they do work, depending on the level of accuracy that you require. I think there's a few things that the world hasn't fully understood yet, on average, and I think that's one of the key ones as it relates to AI detection.
AI Detection Challenges
Jonathan shared his journey from working in content marketing to founding Originality AI. He talked about the challenges of detecting AI content and the importance of balancing accuracy with minimizing false positives. Originality AI's AI detectors are built using supervised learning and have been trained on millions of text records to predict the difference between human-written and AI-generated content.
Adversarial Prompts and Detection
Jonathan also discussed the use of adversarial prompts in trying to beat detectors and how Originality AI's red team and blue team work together to improve detection models. He highlighted the different models they have developed, such as the Turbo model for near-impossible detection and the standard model for a more balanced approach.
Use Cases
Web publishers are among the largest users of Originality AI's technology. The tool is especially useful for copy editors and publishers who need to ensure that the content they receive meets a certain standard before publishing. While the accuracy rate for non-adversarial prompted GPT-4 content is high, it drops slightly with adversarial prompts, leading to a small false positive rate.
Jonathan emphasized that while AI detection technology can be beneficial in academia, it is important to use it for informational purposes rather than academic disciplinary actions.
Overall, the conversation with Jonathan shed light on the evolving landscape of AI detection and the challenges and opportunities it presents for content creators and web publishers.
