Can Artificial Intelligence make use of Common Sense?

AI systems are increasingly sophisticated and capable of carrying out increasingly difficult tasks. But they fall on problems that a 4-year-old child would solve by using his own common sense.

Credits Pixabay
The ' artificial intelligence in recent years has learned to do many things: recognizes objects, helps physicians in the diagnosis , the driving cars , beat the best human chess champions and go. Just to name a few.
But these same systems, however sophisticated, are not capable of answering apparently trivial questions such as "If I put an object in this drawer, tomorrow will I still find it there?" Or "How can I know if a bottle is full or empty"?
In short, artificial intelligence is like a very small child who learns to drink from a bottle but does not know how to drink from a cup. It is not capable of abstracting the problem starting from the solution.

HIGHLY INTELLIGENT?

It seems incredible, but even the most advanced supercomputers are not equipped with what is called common sense, that is the ability to respond to trivial problems based on previous experiences.
The reason is purely technological: teaching these skills to a machine is extremely difficult and to date no one has yet succeeded.
But managing to solve the problem is extremely fascinating and several groups of researchers have taken up the challenge.

HOW MUCH COMMON SENSE COSTS

At the forefront of the battle for artificial common sense is the Institute for Artificial Intelligence , the computer science lab wanted by Paul Allen , co-founder of Microsoft.
Allen has invested over $125 million in the Alexandria project , a new initiative that aims to teach common sense to AI .

Those who have already tried to tackle this challenge have used different approaches: Gary Marcus , founder of the Geometric Intelligence Company , has been inspired by the way children learn, while researchers at Imperial College London have focused on symbolic systems.
This technique provides that the AI, during the learning phase, has available a human being able to describe objects and situations.
But up to now neither road has led to appreciable results.

A Short Video

THE TECHNOLOGIES IN THE FIELD

Alexandria will however be something different. As stated by the company will integrate computer vision systems with software that will simulate human reasoning.
The researchers will also look for new solutions for the development of a "collaborative common sense", that is a set of rules and behaviors collected online among a large number of human beings that can be interpreted into code and made available to machines and robots.

Big Thanks to @enginewitty for this cool Banner