Life is but an Algorithm ...

THEY AND BIG DATA DETERMINE MANY OF OUR DAILY ACTIVITIES: FROM THE USE OF SATELLITE NAVIGATION TO ONLINE SHOPPING. AS DEMONSTRATED BY BOCCONI RESEARCHERS IN 8 DIFFERENT EXAMPLES

There are three ingredients speeding up innovation, revolutionizing the way businesses operate, our way of life and the way public affairs are managed: increased availability of data (big data), the availability of increasingly powerful computers, and the availability of algorithms capable of transforming data into information and instructions for managers and policy makers.

In a popular article on artificial intelligence on his blog, Tim Urban notes that the speed of progression in innovation over the past 50 years is equal to that of the past 150 years. Additionally, over the next 5 years, we will have innovation at a rate that is even higher than the past five.

At this stage, algorithms are playing a leading role. They are at the basis of artificial intelligence today, but if we look back, we find that our lives may have been most influenced by one of the most famous optimization algorithms, the simplex algorithm created by George Dantzig around the end of the ‘50s. It is an algorithm that solves linear programming problems (the naming “programming” is derived from the use of the first calculation programs developed at that time). The algorithm made linear programming very popular for solving business problems and today huge problems (with millions of variables) can be solved in a few minutes using a laptop PC.
Has it impacted our lives? A famous brand of children’s cookies adopted it to find the best recipe to solve dietary problems (a linear optimization problem with constraints dictated by the minimum and maximum amount of ingredients and nutrients). Many of us have taken advantage of this optimization when raising children between 6 months and 2 years old. Other famous algorithms are Quick Sort, Merge Sort and Heap Sort, which allow enormous amounts of data to be searched, or the Dijkstra algorithm for graphs, algorithms for the compression and decompression of data, etc. More recently, revenue management algorithms have affected us in online purchases. These algorithms are able to calculate the optimal time during which a product can remain for sale at a specific price. The algorithm recalibrates the price based on the demand recorded and the number of sales obtained. Here, history says that the first introduction of a new algorithm by researchers at MIT had a disappointing result: a 20% decrease in sales. But a successful adjustment increased sales by 400% and the mechanism was adopted by the project’s corporate sponsor.

Other examples of algorithms that influence us: smartwatches communicate data on our health by using algorithms. Algorithms calculate the shortest path when we request information from a GPS navigation system. Algorithms are behind the human-machine interface for speech recognition. In a talk organized by the Bachelor in Economics, Management and Computer Science at Bocconi (BEMACS Talks), we hosted IBM’s David Nahamoo, one of the pioneers of speech recognition software. These discoveries are at the basis of the functioning of assistants such as Siri and Cortana, dictation software and the IBM supercomputer, Watson.
Enormous progress has also been made in image recognition with deep learning algorithms. Here, national and international research groups, including Yann Le Cunn’s and Riccardo Zecchina’s, have made exponential progress in recent years. The same can be said of classification and clustering algorithms or any automatized text analysis, with algorithms by Gary King (who was also a BEMACS Talks guest) that are behind successful startups. Industry 4.0 features the internet of things that will allow an increased level of automation in machines, revolutionizing production processes.
Development is frenetic, but there is still a need to be critical of the use of the instrument. A well-known example is that of Google Flu. The reason? Algorithms are created based on several functioning theories and, in conditions in which these theories are not verified, their forecasts become unreliable.

It is therefore necessary to increase the expertise and understanding of algorithms themselves. In that sense, the mentality of developers is often (fortunately) inclined towards open source. There are currently many online platforms that make available libraries of code for machine learning based on open access, where anyone can use these libraries. For anyone interested, a simple online search with keywords such as “machine learning websites” is needed to find a long list of websites in which leading companies, universities and research centers make public codes with machine learning algorithms and data to use the algorithms, based on open source and collaboration that also aims to utilize collective intelligence.
Algorithms affect us closely, today in more ways that we imagine. Development seems unstoppable, but we are only at the beginning of a new climb or, to put it another way, a shift in its steepness. On one hand, we have bene able to retain the fundamental view of the fact that algorithms are tools, whose limitations we will only understand after continuous development and subsequent improvements. On the other hand, we must preserve and increase knowledge in these areas at all levels with new educational programs that are able to meet the demand for new skills.

Read how algorithms enter the work of Bocconi researchers in various fields

Guido Alfani. New Eyes to Read History
Valentina Bosetti. Researching Climate and Policy with WITCH
Paola Cillo. Fashion is Hidden in the Big Data of Instagram
Claudia Imperatore. How to Tell if Your Budget Has Been Manipulated
Silvio Petriconi. Studying Banks with Python
Sonia Petrone. Between Theory and Uncertainty: Interpreting the World
Oreste Pollicino. The Law of Privacy and Business Formulas
Gaia Rubera. The Perfect Startup Pitch? An Algorithm Can Write the Model

 

by Emanuele Borgonovo