THE DAILY MAIL SAYS computer generated text has ‘tells’ that make it stand out from human writing. Researchers at Harvard and MIT have identified how computers rely on predictable words. Language used by humans tends to be more unconventional, either in individual word choices or how they are arranged in a phrase or sentence.
Their findings could help flag fake texts. The researchers parsed texts and coded the results using colours. Human writing shows up as a balanced mixture of colours. Computer-generated text shows up as a higher ratio of green — the colour assigned to predictable words.
The researchers say their system works on smaller texts and is not intended for a large scale. The project concentrated on text and not video.
‘According to results of their own research, (the system) helped to improve the detection-rate of forged text from 54 percent to 72 percent’
SEE RELATED STORY
This AI tool is smart enough to spot AI-generated articles and tweets TNW | July 29, 2019 | by Ravie Lakshmanan |‘Initiatives like (this) can be valuable not only in detecting fake text, but also identifying Twitter bots that have been used to disrupt electoral processes in the US and elsewhere.’