US Researchers Develop New Algorithms to Train Robots for Armies

US Researchers Develop New Algorithms to Train Robots for Armies, In cooperation with the University of Texas, Artificial intelligence (AI)
US Researchers Develop New Algorithms to Train Robots for Armies

US Researchers Develop New Algorithms to Train Robots for Armies

US Researchers Develop New Algorithms to Train Robots, Researchers at the US Army Research Laboratory, in cooperation with the University of Texas, have developed an artificial intelligence (AI) technique that will teach robots to perform tasks by interacting with a human instructor for the army.

US researchers have developed a new machine learning algorithm called Deep TAMER,  an extension of (TAMER) Training an Agent Manually via Evaluative Reinforcement, that are loosely inspired by the brain to train a robot about to perform tasks by viewing video streams.

The findings of the study, which were released Friday, will be first presented at the Association for the Advancement of Artificial Intelligence Conference in New Orleans, Louisiana. In their previous work, the researchers taught a robot how to behave in a situation similar to the situations where a human teaches an agent by providing real-time feedback in the form of critique.

According to army researcher Garrett Warnell, the future army will consist of soldiers and autonomous teammates working side-by-side. In future, the team will necessarily be invited to perform tasks such as search and rescue operation, surveillance, in a new atmosphere they have not seen before. In that environment, humans will be remarkably good at executing their training, but current AI agents are not that good.

Currently, there are many systems in Artificial Intelligence (AI) that makes a robot to interact with the environments to learn how to efficiently perform tasks. During this process, an agent might perform tasks that may not only be wrong but could be dreadful.

As the first step in an experiment, the researchers demonstrated Deep TAMER’s success by giving 15 minutes of human-provided feedback to teach a robot to play the Atari game of bowling. The researchers were shocked to see that the robot has done much better than both their amateur trainers, on average, better than an expert human player.

Also Read:

LEAVE A REPLY

Please enter your comment!
Please enter your name here