UNESCO: Reflections on the Ethical use of Artificial Intelligence

Introduction

Within the framework of its mandate in the social and human sciences, UNESCO leads ethical reflections on the development and use of artificial intelligence (AI). In 2021, Member States adopted by consensus the Recommendation on the Ethics of Artificial Intelligence, the first global normative instrument in this field. This document establishes principles and values to ensure that AI contributes to sustainable development, social justice, equity and human rights.

In this context, we present an interview with Dr. Carlos García Torres who holds a Bachelor’s degree in Social, Political and Economic Sciences, a law degree, and a doctorate in Jurisprudence from the National University of Loja. He has postgraduate studies in gender, equity and sustainable development, and earned a "Cum Laude" Doctorate in Law and Social Sciences with honors (Cum Laude) from the National University of Distance Education (UNED, Spain). His doctoral thesis was entitled  Tolerance and Liberalism in the Constitutional History of Ecuador..

How should we understand artificial intelligence as a cultural instrument?
Let us begin by reflecting on AI as a cultural instrument—that is, an artificial product created through the collection and processing of the objects generated by human culture. From this perspective, we will understand that it is an element fundamentally distinct from previous technological tools used by humanity up to now, and that its use, while lawful and necessary, requires a careful ethical approach. 

Another element to consider in developing this ethical awareness is the social, political, and economic implications of science, along with the fact that no technology is ever truly neutral. Artificial intelligence has a developmental history always linked to very specific cultural and financial environments. Today, its core revolves around certain companies based in Silicon Valley and the global interests of those companies. Added to this is the global geopolitical landscape, which includes two key players: China and the European Union.

Finally, a not insignificant element is the environmental impact of the technological infrastructure necessary for AI to function, as reflected in its effects on energy consumption, greenhouse gas emissions, and water usage.

What aspects of justice must be considered in the use of artificial intelligence?
We must consider the social, political, economic, and environmental aspects implied by the frequent and indiscriminate use of any application, model, or system involving AI. First, we must take into account data economy and existing national regulations governing the use of third-party data. It is important to consider the cost of texts uploaded to language models, as well as the caution that must be exercised when uploading texts that are protected under third-party intellectual property rights.

Second, as responsible citizens, we must reflect on the future benefits that AI can bring to scientific research and the economic development of our country. We should strive to ensure that our role is not merely that of consumers, but also of facilitators of innovative and imaginative uses, and eventually, developers of new applications in our respective fields of study.

What ontological challenges does AI pose regarding humanity?
A distinguishing feature of large language models and other AI technologies is their ability to convincingly imitate human qualities, including artistic expressions. This capacity gives rise to the compelling illusion of interacting with an entity seemingly endowed with the distinct dignity unique to human beings. This presents the problem of remembering what is genuinely human.

Several studies have shown that the everyday use of, and growing dependence on, various large language models limit intellectual potential. Responsibility regarding the use of AI is twofold: it not only affects individuals personally but also impacts future generations.

What must we take into account regarding the reliability and rationality of AI systems?
One of the most widely promoted claims by AI system developers is the idea that, with a "prompt," one can obtain the required information at any given moment, under the assumption that this information will always be correct. This assertion is only partially true, as large language models have documented error rates of up to  60 percent.

For this reason, it is crucial to act as content reviewersWE when dealing with AI-generated outputs. Only an expert in each field can detect errors and biases.

This ties in with the general perception of large language models as being rational. We must avoid treating AI responses as the ultimate models of scientific and social rationality.

The use of material obtained from large language models must be subject to continuous scrutiny and refinement.

How should we approach the issue of writing and attribution in AI-generated texts?
In general, a transparent way to produce texts is to acknowledge the use of AI and explain the generation process employed. This highlights the importance of transparency, specifically in recognizing educational, outreach, or research texts that have been created with AI.
It is important to remember that what large language models offer us are generally ideas and resources from other authors. For this reason, PROM distinguish between one’s own work and content generated by a large language model. It should be noted that some models and systems already use watermarks in generated texts, making them easily identifiable as AI-produced.

What environmental impact does the use of artificial intelligence have?
It must be considered that the enormous amount of energy and water resources required by AI models and systems makes them a latent environmental threat, to which every "prompt" contributes. For this reason, their use must always be deliberate and limited with environmental concerns in mind.

It is therefore necessary to constantly inform oneself about the contribution of AI models and systems to greenhouse gas emissions and climate change in general.

Attribution Note: The entire content of the responses belongs to Dr. Carlos García Torres. The final document has been edited with the support of artificial intelligence tools (Copilot). This text is published within the framework of UNESCO’s approach to promoting critical thinking on the ethical use of AI.

The terms used in this publication and the presentation of the data contained herein do not imply any position on the part of UNESCO concerning the legal status of countries, territories, cities or regions, or of their authorities, or concerning the delimitation of their frontiers or boundaries. The ideas and opinions expressed in this work are those of the authors and do not necessarily reflect the views of UNESCO or commit the Organization.