Artificial intelligence and robotics have attracted widespread attention in recent years, but most of these projects focus on very specialized tasks or more extreme theoretical algorithms. Modern robotics are not yet available for making and driving cars, let alone improving people’s daily lives. To that end, Alphabet’s X company announced the launch of the Everyday Robot project to design, build and train robots to learn the chores of the second day of humanity.
Many things may be taken for granted to humans because they make us feel “natural”. Examples include small talk when pouring water next to a water dispenser, and something temporary appearing or disappearing every day. Humans won’t be surprised because they’re used to it, but it can confuse the robot’s logical thinking.
In order for these robots to be able to help people with their daily tasks, they need to be able to adapt to rapidly changing elements and environments. And compared with various algorithms, robots need to do more to learn.
Naturally, machine learning and artificial intelligence are important foundations of the X Everyday Robot project. But these algorithms are not intended to achieve a specific goal or perform operations, and they want to create a robot that will learn how to perform tasks with minimal programming. To test its theory, X chose a task that was moderately difficult but still measured its success or failure: sorting garbage.
X’s analog robots “practice” how to sort garbage in the simulation office every night, and will be deployed to real-world physical robots and real waste disposal processes in the future. The team then entered the results of the “actual” activity into the simulation training, allowing the robot to train again that night. According to X, this has led to a drop in office waste pollution levels from 20 per cent to less than 5 per cent.
This is an impressive figure, but it does more than that. X’s next step is to train the same robot to complete another task without having to rebuild the robot or write a new program. The company acknowledges that this may be an impossible task, but that efforts are still needed.