Teaching robots to learn how to learn

Robots will be part of our daily lives, Whether you like this idea or or not.

Most of the examples we have nowadays, remain limited to passively enaging machines who are focused on sharing information. The real challenge will be to make robots ‘actionable’: assist us by performing some sort of physical task.

For them, in order to do so, these robots have to be able to act within an unstructured environment, like our homes. More and more research is focused on 𝓉𝑒𝒶𝒸𝒽𝒾𝓃𝑔  𝓇𝑜𝒷𝑜𝓉𝓈  𝓁𝑒𝒶𝓇𝓃  𝒽𝑜𝓌  𝓉𝑜  𝓁𝑒𝒶𝓇𝓃.

What this means? Instead of working with predefined goals, let robots experience certain behaviour first and assign a goal afterwards. For instance, let them play with blocks, so they notice the different behaviour of triangle or square blocks and than tell them to push a triangular block to a certain location. The newly formed algorithm quickly calculates lots of scenario’s based on the observed behaviour of this triangular block and than performs the scenario that has the highest likelihood of success.

The underlying technology has even evolved into a phase where it’s able to watch a human perform a task and copy the exact same routine. Have a look below.