PEOPLE | Jul 19, 2019

When the algorithm becomes selfish: an interview with Massimo Chiriatti

In the book #Humanless, the impact that Artificial Intelligence has (and will have) on our work, our lives and our way of relating to the world.

One day there will be a machine that will make us understand that we are no longer the beings at the center of work?”. The time will come, or has already come, to understand how algorithms and artificial intelligence have changed our lives. The book #Humanless. L’algoritmo egoista (#Humanless. The selfish algorithm) by Massimo Chiriatti tries to shed light on how much and how our relationship with machines has changed, how they have become not only tools but also work and life companions. “The book – says Chiriatti – wants to explain in a simple way the dynamics, the main choices, the risks and the opportunities of an algorithm-based society. It is designed for those who are about to choose a career path and must understand how to relate to algorithms which are no longer passive, but active; for policy makers (entrepreneurs, trade unions, politicians) who wish to understand what the creation of these platforms based on AI will lead to; for those who want to investigate the impact that algorithms, in addition to globalization, will have on the world of labor”.

Created, not generated. The algorithm, we read in the book, was created by man and it has become selfish. “I will show you some things, actually, some rather bad things, due to the many errors resulting from poorly calibrated statistical models and unsustainable business models”, the algorithm promises in the pages of #Humanless, to then reassure the reader immediately afterwards: “We will find a way to work together for the best of all possible worlds”.

So we have a bright future ahead of us, not a world dominated by machines and where man will feel in danger, right?

“The problem is not Artificial Intelligence but our biases, the prejudices which are amplified by the machines when they make their decisions and this will worsen the situation. The theme is: “How do we correct ourselves first, to then correct AI? A way to govern machines and to derive the utmost benefit from them is by knowing how to recognize biases and above all by knowing how to accept them”.

What are the real risks, which go beyond an easy and impactful narrative that too often one reads or hears about? And how can we contain these?

“The Artificial Intelligence system may not be able to explain the reasons behind its conclusions. Its method is: marginal adjustments of numerical weights, which interconnect its huge number of artificial neurons, while analyzing the training data within a timespan which is now unsustainable for humans. A method so different from the one used by humans and one which can lead Artificial Intelligence to misinterpret the data because of this intrinsic lack of context. AI therefore cannot take into account the subjective, qualitative, ethical or reasonable prejudices which guide human decisions. The greatest danger is that, by ingesting new data, it will learn exponentially faster than we do: it is therefore likely that its mistakes and deviations will spread and grow rapidly. This could inevitably develop small deviations that, over time, would lead to systemic disasters. The machine doesn’t know how it came to suggest the information; therefore we have to choose whether to use this”.

Data are given a central value, but what is their role really, since, as the book says, they “age before we do”?

“We often quote the power of data, comparing them to the wealth of oil owners, but this has been generated by the centuries-old stagnation of fossils and is in finite quantities. On the other hand, data are unlimited and are created exponentially by the action of machines and people. From the point of view of quality, any one drop of oil is the same as another. While a datum always has a history, it is not interchangeable: each one is different from the other. Data are immaterial and therefore intangible assets, to obtain value from their personalization we must process them to extract information and obtain the desired knowledge. But if they are too old, they can’t capture people’s latest behaviors. Data must be used as soon as possible and therefore primarily by those who produce them, thus creating a virtuous cycle, a win-win solution: from those who produce them, for those who process them and to those who finally benefit from them”.

The book mentions the need for man to complement machines. In what way can man always stand out and provide added value?

“It will be necessary to ensure that the result elaborated by the machines is constantly neutral or, in other words, is not influenced and does not act according to the author or the contents. In this sense, this work of control is an art rather than a science. It is human judgment that decides: there will be an increasing need for this and it will be worth increasingly more”.

What advice should be given to a young man who has to choose his course of study at this time? What future awaits us at work?

“First of all, we must not frighten or mislead him, because there is uncertainty about what Artificial Intelligence, or better, augmented intelligence, can do and in particular about what it can become. Here we try to observe the trend of the past and to imagine the immediate evolutions. I would say: observe the ratio between your relationships with people in relation (numerator) to the transactions  we currently carry out with other people and with machines (denominator). Increase the numerator and you will find the work in the eyes of the other”.

Sonia Montegiove