In the last year, we have written several articles about the use of artificial intelligence in a variety of settings. AI technology can automate some processes that can free humans up for other tasks, which sounds ideal. However, because AI runs on algorithms fed by existing information, it can amplify the biases that are present in the data. For more information about AI bias, check out this article. Another concern is that of teachers and college admissions departments who fear AI writing assistants could be creating the essays assigned in their classes. Thankfully, AI scientists are working on ways to identify whether a writing sample was created by a human or a machine.
AI, Know Thyself
It probably isn’t a surprise to learn that the best way to detect whether a writing sample was created by an AI is to ask an AI. There are two new tools, GPTZero and Originality.ai, designed specifically to detect whether a human wrote a text sample. As an added bonus, Originality.ai also checks for plagiarism, making it a very popular tool for teachers.
But if AI technology has built-in flaws, how accurate will it be when used to detect itself? While no tool is perfect, especially when built on imperfect information, both of these tools are promising in their results.
Comparing Detection Accuracy
On one of my favorite tech sites, Your Nerdy Best Friend, Beth Z. created an experiment to see how accurate the AI detections were, and how consistent the tools were with each other in their findings. I found the results very interesting. If you’re interested in seeing how they did, head to the experiment’s post on her blog.
It seems artificial intelligence tools taking hold in every industry, with new applications for AI being created each day. While this whirlwind change can be dizzying, it’s nice to know there are developers out there concerned with the ramifications of this technology who are willing to develop counter-tools to keep AI use in check. Are you using any AI tools? Let us know what you are using and what you think of it in the comments.