For artificial intelligence with human values

The stakes are too high to leave the course of AI development to researchers, let alone CTOs. While heavy-handed regulation is not the answer, the current regulatory vacuum must be filled, and that process requires a broad-based global commitment.

MADRID – 2023 could be a reference year because artificial intelligence transforms daily life. So Brad Smith, president of Microsoft, said at an event on AI organized by the Vatican last week. But Smith’s statement was not so much a prediction as a call to action: the event (attended by industry leaders and representatives of the three Abrahamic faiths) is in line with the goal of promoting an ethical and anthropocentric approach to development. of the AI.

There is no question that AI poses tremendous operational, ethical, and regulatory challenges. And answering them will not be easy at all. Although the development of AI began in the 1950s, the general concept and its likely impact have not yet been defined.

Some facets of the immense potential of AI have been made known by significant recent advances in the field, ranging from the chillingly “human” texts of the ChatGPT tool (created by OpenAI) to applications that can shorten the discovery process by years. new drugs. But it is still impossible to foresee all the ways in which AI will transform our lives as humanity – and, more broadly, civilization.

This uncertainty is not new. Even after recognizing the transformative potential of a technology, we are often surprised by the shape of that transformation. For example, social networks, initially hailed as an innovation that would strengthen democracy, have been much more effective in destabilizing it, by facilitating the spread of disinformation. We can almost certainly predict that there will be those who will exploit AI in evil ways.

We don’t even really understand how AI works. Consider the so-called “black box problem”: with most AI-based tools, we know what data is going in and what data is going out, but we don’t know what happens in between. Where AI is used to make (perhaps irrevocable) decisions, this opacity poses a serious risk, exacerbated by issues such as the transmission of implicit bias through machine learning.

Two other risks are the abuse of personal data and the destruction of jobs. And according to former US Secretary of State Henry A. Kissinger, AI technology can undermine human creativity and insight, if information ends up trumping wisdom. Some fear that AI will cause the extinction of the human species.

With so much at stake, we cannot leave the future of technology in the hands of researchers in the area -much less in the hands of technology executives-. Although stifling regulation would not be the correct response, the regulatory vacuum that currently exists must be filled. And to do so, we need large-scale international dialogue, similar to the one that characterizes today, after many setbacks, the fight against climate change.

In fact, climate change is a good analogy for AI. It is much more useful than the usual comparison with nuclear weapons: although nuclear weapons can affect people indirectly, through geopolitical events, nuclear technology is not present in our personal and professional daily lives; Nor is it shared by everyone. On the contrary, climate change (like AI) affects everyone, and the measures that must be taken may have social costs in the short term.

The race for AI dominance is a key aspect of the US-China rivalry. If one of the two countries limits its own AI industry, it risks the other taking the lead. For this reason, in this field -as well as in the reduction of emissions- cooperation is essential. Governments, along with other relevant public actors, should collaborate to design and install guidelines for private sector innovation.

Certainly, this assertion easy to proclaim is extraordinarily difficult to execute. Between theory and practice there is today a real abyss. For now, the limited consensus around the response to AI has given rise to a hodgepodge of regulations. And attempts to reach a shared approach in international fora have been hampered by power struggles between major players and the lack of an agreed compliance mechanism.

However, there is promising news. The European Union is designing an ambitious instrument that anchors the principles that allow the establishment of harmonized standards for AI. The AI ​​Law, which is due to be finalized this year, seeks to enable the “development and adoption” of AI in the EU, while ensuring that the technology “works for people and is a force for good in society”. . The legal proposal includes everything from adapting the rules on civil liability to reviewing the EU framework for product safety; it is paradigmatic of a comprehensive approach to AI regulation.

This regulatory leadership by the EU should come as no surprise: historically, the bloc has been at the forefront of developing regulation in critical areas. European data protection legislation has arguably inspired similar measures in other countries, from the California (USA) Consumer Privacy Law to the Personal Information Protection Law in China.

But the global regulation of AI will not be able to advance without the United States. And despite the shared commitment between the US and the European Union to develop and implement “trustworthy AI”, Washington’s main concern is a question of power: supremacy in the field. To achieve this, it has set itself the goal of not only strengthening its own advanced industries (which includes minimizing bureaucratic impediments, the so-called “red tape”), but also hindering China’s progress.

According to a report published in 2021 by the Homeland Security Commission on Artificial Intelligence, the US tactic includes “bottlenecks that impose significant derivative strategic costs on competitors, and minimal economic costs on US industry.” An example of this strategy is the export controls approved by the United States last October, which target China’s advanced computing and semiconductor industries. For its part, China is unlikely to give up its goal of achieving technological self-sufficiency (and eventually supremacy).

In addition to creating favorable conditions for AI risks to materialize, this technological rivalry has obvious geopolitical implications. For example, while Taiwan’s disproportionate importance in the global semiconductor industry gives it a strategic advantage, it also exposes it once again to being targeted.

It took more than three decades for awareness of climate change to crystallize into real action; we are still not doing enough. Given the pace of technological innovation, we cannot afford to follow a similar path in the case of AI. If we do not act now to ensure that the development of this technology is guided by anthropocentric principles, we will come to regret our inaction. As in the case of climate change, it will be much sooner than we think.

The author

She was Minister of Foreign Affairs of Spain and Senior Vice President and General Counsel of the World Bank Group; She is currently a visiting professor at Georgetown University.

Copyright: Project Syndicate, 1995 – 2023

www.projectsyndicate.org