Tuesday, September 23, 2014

Inroads in perception

Posted on 08/21/14 by Frank Tobe

Google's recent acquisition of Emu Messenger is just one of many items in recent news about improvements in perception and artificial intelligence (AI).

Emu, not to be confused with the Australian ostrich-like bird nor the European Monetary Union, is a small Palo Alto start-up comprised of some serious software talent with experience in machine learning, natural language processing and mashing up different data, databases and systems at Siri, Apple, AOL and Google. Perception in this case is the feeling that your words and intentions are understood and acted upon. No financial details were disclosed about the acquisition however the Emu app will be shut down next week.

Another form of perception, computer eyesight (AKA machine vision), took a giant step forward when the winners were announced for this year's Large Scale Visual Recognition Challenge. The NY Times reported the winners: the National University of Singapore, the Oxford University, Adobe Systems, the Center for Intelligent Perception and Computing at the Chinese Academy of Sciences, as well as Google in two separate categories.

Fei-Fei Li, director of the Stanford Artificial Intelligence Laboratory said, “What really excites us is that performance has taken a huge leap.”

Machine vision has been a challenge in automation and robotics but in recent years has become integral in countless applications including computer gaming, medical diagnosis and factory robotics. Carmakers too have added the ability to recognize pedestrians and bicyclists and trigger safety actions. Factory robots need much improved perception systems in order to monitor the progress of their tasks and the tasks of those around them. [A quick Google Images search produces a collage of various uses, from milking to adaptive cruise control.]

Enhanced algorithms, data libraries and faster and cheaper computing are all contributing to the increased accuracy and speed of the systems recognizing objects and identifying them by type and in 3D space, nevertheless, at their best they are still no match for human vision and perception.

« Return to The Robot Report

Avatar

About the author: Frank Tobe

Frank Tobe is the owner and publisher of The Robot Report. After selling his business and retiring from 25+ years as computer direct marketing and materials and consulting to the Democratic National Committee and major presidential, senatorial, congressional, mayoral campaigns and initiatives all across the U.S., Canada and internationally, he has energetically pursued a new career in researching and investing in robotics.

Read more »