Why we use Artificial Intelligence at Mangrovia
Trust, transparency and quality of information
Information is power. Too often, however, the word “power” is declined in the form of command and control, domination over something else, and various forms of violence. Instead, we prefer to recover the meaning of Danilo Dolci, who emphasised its character of possibility, that possibility of change that arises when one shares a method and a horizon.
We believe that having more information linked together, which becomes knowledge, enables the possibility of changing your point of view, acting differently, and continuing to search for solutions best suited to the contexts we experience in our relationships, both individual and social.
The Mangrovia you are reading, which you listen to as a podcast with Zenit, was born from this idea. And from the idea that to solve epochal problems of planetary scope, such as climate change, perhaps we should not always insist on the same patterns of thought and action, but try to give voice to people and territories that are generally not heard and imagined, to understand what they have to teach us.
The language of the arts is the language of imagination that we as human beings have honed over all these years of presence on earth. In Sineglossa we have been trying for many years to expand, disseminate, and use this language to inspire and reflect together on the most important issues of our time.
That is why, when SCG asked us to take care of the content of Mangrovia, we thought of an editorial project that was not afraid of the intersection of so many roots.
We’ve observed a variety of Artificial Intelligence applications, such as various techniques for collecting and analyzing large data sets, being used in the cultural sector. However, we recognised an additional responsibility in the field of journalism.
As always, the choice to use AI in editorial processes is not neutral, but directly impacts the quality of the information and dissemination we try to do, the economic sustainability of the whole project and, above all, the trust of the readers.
First, do no harm
We recognize that building and utilizing these systems bring up significant concerns. These include issues related to the data used to train algorithms, the control processes, ensuring respect for human and workers’ rights, and minimizing environmental impact. Additionally, especially in the realm of information, it’s crucial to consider how sources are handled and to protect personal and corporate data.
At the same time, we are equally aware that Artificial Intelligence represents an extraordinary opportunity to experiment with collaborations between humans and software that enable both to make the best use of each other’s skills, not as a substitute, but as an enhancement.
Given all these evaluations, and previous studies, we followed the ancient adage “primum non nŏcēre“: we turned to an Italian start-up, Asc27, subject to Italian legislation and certified for the quality control of its processes, which provided us with guarantees on the treatment of training data and the use of algorithms, as well as compliance with current regulations. In particular, we ensured that the software suite to be used was fully controlled and auditable by the producing company.
We have gone, and are going, in small steps: we have decided, for the time being, not to automate any of our production workflows, but to make use of the collaboration of this suite for a very specific task, the research of stories to be told.
We trained Asimov through fine-tuning with a dataset aggregated by the entire editorial staff, composed of sources that could guarantee us the necessary quality and fundamental pluralism, defined as a system of voices coming as far as possible from marginalised places. The selection, verification and writing remain, however, always in the hands of the human editorial staff.
We have chosen to be transparent: we will tell you how and when we use it, because we would not be comfortable doing otherwise. This too is information that generates change and new possibilities for doing things together.
The cover image is by Rick Payne from Better images of AI library and it is free to use under the CC-BY 4.0. It shows the many different ways in which people might visualise the many different facets of AI. It shows the different applications, concepts, processes and technologies which 2 simple letters can mean. These post-it notes were drawn by the audience and team at the launch of the Better Images of AI Guide for Users and Creators , held at The Alan Turing Institute.