Planet For Application Life Development Presents
Technology World

Explore and uptodate your technology skills...

News Navigation: First Previous Next Last

This AI software can predict your near future
16-Jun-2018

Researchers have developed the first level of technology that can predict your near future. Artificial intelligence (AI) powered technology knows what you are going do next.

German researchers have worked on a prototype of the technology that can anticipate your next actions. Tech firms and researchers around the world have been tinkering with artificial intelligence for last few years. But no consumer technology has achieved anything other than setting digital reminders and placing online orders.

Prof. Jrgen Gall and Yazan Abu Farha from the Institute of Computer Science at the University of Bonn are teaching computers to foresee your next actions. These researchers say that they have taken an important step in creating the first self-learning computer program that can anticipate your next actions.

Researchers team first made the two different neural networks watch 4-hour long videos in which humans are preparing different salads. The 20 action-item data was used to train the system on how to prepare the salad as well as a duration for each action.

Prof Jrgen Gall said, Then we tested how successful the learning process was. For this, we confronted the software with videos that it had not seen before. Our methods are the first to predict video content of up to several minutes long.

The same AI program was shown more salad making videos and demonstrated the actions that it was watching. It was asked what actions ti was watching during first 20-30% in one of the videos. The program then was expected to predict what would happen next for the rest of the video.

The algorithm is capable of predicting the actions with 40% accuracy. Researchers are improving the accuracy of an algorithm by feeding more data. German researchers have published a paper describing the details of the research. The paper will be presented at the Conference on Computer Vision and Pattern Recognition (CVPR), which is scheduled for June 19-21 in Salt Lake City.