| Scientists Develop Technology to Imitate Handwriting Using AI |
|---|
| Source |
Abu Dhabi-based scientists at the University of Artificial Intelligence named after Mohamed Ben Zaida have successfully developed a technology that can imitate a person’s handwriting. The breakthrough was achieved by using a transformer neural network to analyze context and semantic units in a series of consecutive data paragraphs. The technology has been patented in the department of patents and trademarks of the USA [source].
While the technology is not yet widely accessible, it represents a significant step forward in an area that has long intrigued academic circles. Previous applications and robots have been able to imitate handwriting, but recent advancements in the field have greatly accelerated the development of text recognition mechanisms. However, it remains unclear whether the potential risks outweigh the benefits of this technology.
On one hand, the technology has the potential to assist individuals with disabilities in writing without the need for a pen. On the other hand, it opens the door to widespread forgery and abuse. The developers acknowledge the need for cautious implementation and suggest increasing awareness and developing methods to combat fakes. Chisham Cholakal, associate professor of the department of computer vision, likens this to creating an antivirus for a new pathogen.
Despite these concerns, the inventors plan to start applying their technology to real-world applications within a few months and are actively seeking commercial partners. RAO Muhammad Anver, also an associate professor of the department of computer vision, highlights the vast potential that this development holds, from deciphering doctors’ handwriting to personalized advertising. Additionally, the technology can be used to generate large volumes of synthetic data to improve other AI models’ ability to process handwritten text.
However, there is still work to be done. While the neural network trained on open samples is capable of recognizing and reproducing English text, as well as some success with French, the team has not yet been able to teach it to analyze Arabic Vyaz.