THE NEW CHALLENGES OF THE MACHINE LEARNING

Click farms are developing in South East Asia as real digital consensus factories. Thousands of people, mobile phones and SIM cards, incessantly click on banners in order to generate false traffic. Fraudulent clicks sold in fake downloads packages to pick up your product in the rankings so to reach real customers.

A current challenge of machine learning is to develop algorithms that are able to interpret and identify these fraudulent clicks, to identify and then report the products of companies that use these “borderline” marketing strategies.

An American foundation has offered $ 1 million to create a predictive model that can analyze radiographs and automatically identify the presence of lung cancer, faster and more objectively than the human eye.

Machine learning allows the identification of certain objects within a digitized image (eg. a CAT) in order to identify them and analyze their main peculiar characteristics.

The analysis of aerial or satellite images, supported by appropriate predictive algorithms, can map the presence and path of icebergs to the North Pole, the spread of certain types of animals on a territory, but also, more prosaically, the cars present in the parking lot of the competing distributor to reveal its turnout.

But also social life could benefit from new technologies. In the NLP field (Natural Language Processing) an algorithm has been developed that is capable of automatically identifying and blocking racist, sexist and homophobic sentences within a forum.

However, one wonders if at the end the man will not feel controlled by these algorithms or will feel obliged to respond to non-human standards, given that very often now is an algorithm that advises us on social media people interesting for us, choose the products of our taste and select the candidates for our job positions.

(Summarised freely from “In the Olympus of Data Scientists” by Giulio Pons – MAP n.5 / 2019)