We’re one step closer to the prospect of terrifying robot overlords thanks to some clever MIT researchers. Programmers at the university have trained a robotic arm to learn and perform complex tasks, including setting a table, just by watching humans.
The MIT team tested the robotic arm’s abilities by laying out a table – complete with plates, cutlery and cups – and asking the robot to replicate their actions. After observing the researchers, the little arm managed to accurately lay the table over and over and over again. Out of 2,000 table-laying attempts, good old arm-y only made six mistakes.
To make sure that the robot didn’t have it too easy, the researchers stacked certain items and hid them too. This may have disturbed less sophisticated robots, who would have been confused by this variation, but arm-y persevered and managed to nail his table-laying task.
The robot’s system relies on a new system from the researchers called “Planning with Uncertain Specifications” (PUnS). This basically helps the robot make decisions. Where most robots aren’t capable of logically sorting though complicated or contradictory inputs, this little guy weighs up various options and chooses the action most likely to bring it closer to an end-goal. That’s a very rough translation of PUnS, but this is what’s responsible for those impressive learning and replicating skills.
Related: Best robot vacuum cleaners
In terms of real-world use, this new learning-capable robot could be an adaptable little worker that’s easy to train-up. You could, for example, have one deployed in your house that you can quickly train to do domestic chores.
Ankit Shah, who’s the lead author on this project, said: “The vision is to put programming in the hands of domain experts, who can program robots through intuitive ways, rather than describing orders to an engineer to add to their code.”
This new programming is just one part of that vision, but it might one day mean we have one robot that’s capable of ironing, cleaning, tidying and hoovering. Sorry, Roomba.