An AI Tool to Identify AI-Written Text
September 19, 2019
When distinguishing human writing from AI-generated text, the secret is in the predictability. MIT Technology Review reports, “A New Tool Uses AI to Spot Text Written by AI.” We have seen how AI can produce articles that seem to us humans as if they were written by one of us, opening a new dimension in the scourge of fake news. Now, researchers have produced a tool that uses AI technology to detect AI-generated text. Writer Will Knight tells us:
“Researchers from Harvard University and the MIT-IBM Watson AI Lab have developed a new tool for spotting text that has been generated using AI. Called the Giant Language Model Test Room (GLTR), it exploits the fact that AI text generators rely on statistical patterns in text, as opposed to the actual meaning of words and sentences. In other words, the tool can tell if the words you’re reading seem too predictable to have been written by a human hand. … GLTR highlights words that are statistically likely to appear after the preceding word in the text. As shown in the passage above (from Infinite Jest), the most predictable words are green; less predictable are yellow and red; and least predictable are purple. When tested on snippets of text written by OpenAI’s algorithm, it finds a lot of predictability. Genuine news articles and scientific abstracts contain more surprises.”
See the article for that colorfully highlighted sample. Researchers enlisted Harvard students to test GLTR’s results. Without the tool, students spotted just half the AI-crafted passages. Using the highlighted results, though, they identified 72% of them. Such collaboration between the tool and human interpreters is the key to warding off fake articles, one researcher states. The article concludes with a link to try out the tool for oneself.
Cynthia Murrell, September 19, 2019
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
