Quixote Helps Robots Understand Human Morals and Values By Reading Stories
Researchers from the School of Interactive Computing at the Georgia Institute of Technology have created a device called Quixote, that will help robots understand how to carry themselves in a human interaction by reading books. The proposed collection of stories are picked from different cultures and have been used to teach children socially acceptable behavior. Mark O. Riedl and Brent Harrison have proposed this technology to the American Association for the Advancement of Sciences.
The field of robotics has been advancing at a blinding pace, and new breakthroughs are being made every day. This has led to the fear of how the evolution of robots will affect humans. Quixote is a measure to teach robots what acceptable social behavior is, and ensure their choices will not harm humans. Professor Mark O. Riedl’s earlier research, the Scheherazade system, explains how a robot could arrive at a correct sequence of actions by crowdsourcing scenarios from the internet. Quixote is a sequel to that research.
The Scheherazade will identify the best course of actions, and pass it on to Quixote. Based on the robot’s choice, the device then generates rewarding signals for protagonist-like actions and punishing signals for antagonist-like actions. For instance, when a robot waits patiently in a line at a counter, interacts politely and pays for the service, it gets a rewarding signal as opposed to when it cuts in or robs a store. These actions are guided by scenarios from the stories fed to the device, and will help in establishing ethics.
The name “Quixote" is borrowed from a Spanish novel, The Ingenious Gentleman Don Quixote, which is very appropriate for a device that teaches robots to behave like gentlemen. In order for robots to never go rogue, they also require empathy, which can be reinforced by these moral stories. The research was jointly funded by US Defense agency DARPA and Official Naval Research. Though Quixote cannot be used to moralize every robot, it is fairly suitable for robots with limited functionality that includes human interactions.
Source: #-Link-Snipped-# | #-Link-Snipped-#
The field of robotics has been advancing at a blinding pace, and new breakthroughs are being made every day. This has led to the fear of how the evolution of robots will affect humans. Quixote is a measure to teach robots what acceptable social behavior is, and ensure their choices will not harm humans. Professor Mark O. Riedl’s earlier research, the Scheherazade system, explains how a robot could arrive at a correct sequence of actions by crowdsourcing scenarios from the internet. Quixote is a sequel to that research.
The Scheherazade will identify the best course of actions, and pass it on to Quixote. Based on the robot’s choice, the device then generates rewarding signals for protagonist-like actions and punishing signals for antagonist-like actions. For instance, when a robot waits patiently in a line at a counter, interacts politely and pays for the service, it gets a rewarding signal as opposed to when it cuts in or robs a store. These actions are guided by scenarios from the stories fed to the device, and will help in establishing ethics.
The name “Quixote" is borrowed from a Spanish novel, The Ingenious Gentleman Don Quixote, which is very appropriate for a device that teaches robots to behave like gentlemen. In order for robots to never go rogue, they also require empathy, which can be reinforced by these moral stories. The research was jointly funded by US Defense agency DARPA and Official Naval Research. Though Quixote cannot be used to moralize every robot, it is fairly suitable for robots with limited functionality that includes human interactions.
Source: #-Link-Snipped-# | #-Link-Snipped-#
0