Artificial intelligence has now entered every sphere of our lives. Autonomous robots are self-dictated, have emotional intelligence and can analyse their surroundings and react accordingly. Robots like Smartibot, Vector by Anki and Evie are some of the coolest robots in the market. They can perform walk, perform chores and even play games.
AI has been in vogue since its ability to mimic human-like game moves was discovered. Scientists at the Indian Institute of Science (IISc) are working on a robot that can play Pictionary and other social games like chess with humans. They have already constructed a computational model regarding this project. There’s a demand in development in AI as well, modified AI systems require effort to construct but since the demand in the market is high, AI bots sell easily.
For researchers at IISc, it has been a difficult task since Pictionary can be a complex game for a bot to understand. The person who draws uses visual medium while the people who guess use speech medium. The Video Analytics lab at IISc designed used deep-learning to make a neural network which can match up with the sketch and guess like a human. To introduce a model like this, researchers built a comparison between realistic and human drawings of these objects. This helped the robot to give out human-like object guesses. The scientists had an objective to produce agents which can are relatable to humans and can be accepted in their social settings.
Game Of Bots
The process through which the AI bots are trained to play games is called Reinforcement learning. According to reports, a bot is given an objective to do and if it performs the task well it is given a reward. The first time, the bot uses random techniques to complete the objective, after it figures out the correct way to earn the reward; it follows that technique again to improve its “moves” in the following games. After multiple attempts, the bots conquer the art of the gameplay since their strategies develop to perform only the best “moves” to win and receive rewards.
There have been several events where scientists have staged up competitions between AI-bots and human e-players. Open AI is an AI system which is backed by tech-titan Elon Musk, participated in a game called Dota 2, they played against 5 e-players. After this event in August 2018, several scientists are creating their own models for robots to make them autonomous.
Bots have dethroned their human opponents in some historic game matches like AlphaGo defeated a player in a 5-day gaming tournament in 2016. And Google’s Deepmind defeating several players by teaching itself a golden strategy in “Breakout”.Though the achievements are many, the bots can’t replace humans. Its action is dependent on rewards, if there’s a bug in the game or even a glitch, the bot is not able to recognise the anomalies. It will perform the simplest of tasks just to win rewards, it has no idea of how the video game works.
Enjoyed this story? Join our Telegram group. And be part of an engaging community.
Provide your comments below
What's Your Reaction?
Jignasa pursued her bachelor’s degree in Biotechnology and is currently a trainee journalist at IIJNM. Her mind is usually preoccupied with art, music, food and travel.