Session
Organizer 1: José Gontijo, MCTI
Organizer 2: Heiko Wildner, Federal Ministry for Digital and Transport
Organizer 3: José Gontijo, Ministério de Ciência, Tecnologia e Inovações
Organizer 4: Janina Kempf, GIZ
Organizer 5: Luc Wuest, GIZ
Organizer 6: Onike Shorunkeh-Sawyerr, GIZ
Speaker 1: José Gontijo, Government, Latin American and Caribbean Group (GRULAC)
Speaker 2: Davis Adieno, Technical Community, African Group
Speaker 3: Kim Dressendoerfer, Private Sector, Western European and Others Group (WEOG)
Speaker 4: Ledénika Mackensie Méndez González, Executive Director for Digital Inclusion at the Secretariat for Communications and Transport of the Mexico City Metropolitan Area
Speaker 5: Urvashi Aneja, Civil Society, Asia-Pacific Group
Onike Shorunkeh-Sawyerr, Government, African Group
Janina Kempf, Intergovernmental Organization, Intergovernmental Organization
Luc Wuest, Government, Western European and Others Group (WEOG)
Round Table - U-shape - 90 Min
1. What elements should be considered when designing AI governance frameworks? How to adequately relate them to Internet Governance mechanisms and institutions, such as IGF/ICANN? How to make the relationship between AI, the Internet and SDGs more explicit to stakeholders? 2. How to balance the need to regulate AI with the risk of losing the chances it offers for sustainable development? 3. Are the existing shared principles on sustainable AI development really shared by everyone? Or should we seek to include developing and emerging countries better in the development and implementation of those norms?
Connection with previous Messages: The proposals connects to many of the IGF2021 messages – however the strongest connex is with the following two: Adequate enabling environments (e.g. policies, legislation, institutions) need to be put in place at the national, regional and global levels to foster inclusive, just, safe, resilient and sustainable digital societies and economies Artificial Intelligence (AI) needs to be developed and deployed in manners that allow it to be as inclusive as possible, non-discriminatory, auditable and rooted into democratic principles, the rule of law and human rights.
8. Decent Work and Economic Growth
8.10
8.2
8.3
8.4
8.5
9. Industry, Innovation and Infrastructure
10. Reduced Inequalities
12. Responsible Production and Consumption
12.2
12.a
10.3
10.4
13. Climate Action
13.2
13.3
9.5
9.b
9.c
Targets: Although AI through its cross-cutting application in virtually all sectors of life and the transformative reconfiguration of economic and social interactions bears the potential to influence the achievement of all 17 SDGs positively or negatively, the biggest opportunities and threats remain in the indicated SDGs:. SDG 8 Decent Work and Economic Growth through the expected increase in productivity but also the shift in labour markets. SDG 9 Industry, Innovation and Infrastructure by enhancing the research activities globally and set up incentives for technology development. SDG 10 Reduced Inequalities by ensuring governance approaches that mitigate discriminatory effects of AI technology and enhance social protection and equality. SDG 12 Sustainable consumption and production patterns may be facilitated by the application of AI technologies. SDG 13 Climate Action by exploring and exploiting the possibilities of using AI technologies to support climate actions.
Description:
Artificial Intelligence (AI) has become a buzzword. On the one hand, it is portrayed as a set of technologies that will take over millions of jobs and exacerbate social inequalities. On the other hand, it is seen as a central tool to increase sustainability along its economic, environmental, and social dimensions, and support the worldwide efforts to achieve the UN’s Sustainable Development Goals (SDGs).
While AI’s potential to make a real contribution exists, realizing that potential while mitigating the related risks requires conducive framework conditions to be formulated. In this respect, challenges for policy makers include fostering access to meaningful data, incentivizing capacity development and more generally speaking, closing the governance gap to seize the potential of AI for SDGs. These challenges should be tackled collectively, and adequate framework conditions should be implemented internationally, in coordination with Internet Governance related institutions and mechanisms.
This context calls attention to the relationship between AI, the Internet, and SGDs. Recent studies pointed out to that AI has the potential to achieve more than 70 % of the Sustainable Development Goals (SDGs). AI is expected to affect global productivity and economic growth, social equality and inclusion as well as climate actions and environmental protection, exerting influence on all sustainability dimensions. For instance, AI bears immense potential for contributing to urban sustainability with its application in smart city solutions. Intelligent predictive traffic guiding systems or improved approaches to shared mobility can support the reduction of traffic jams, pollution and allow for new business models. Furthermore, AI solutions can support nature preservation efforts in the fight against climate change, such as in monitoring rainforest health. The energy and agriculture sectors are other examples in which AI can boost sustainability by enhancing efficiency. Regarding the energy sector, AI can allow companies to find out detailed energy usage patterns and connect them with the availability of energy-generating facilities (wind turbines, hydraulic plants, biomass plants, etc.). While in the agricultural sector, AI can support production through intelligent usage of fertilizers and irrigation.
The relationship between AI and SDGs is so strong that various programmes are devoted to exploring AI opportunities. Some examples are the Artificial Intelligence for Sustainable Development Goals (AI4SDGs) Research Program, Oxford Initiative on AI×SDGs, and AI for Good. Nevertheless, reported potential impacts of AI indicate both positive and negative impacts on UN SDGs, and AI application does not come risk-free. For instance, digital technologies, including AI technologies, are energy intensive. Especially if energy sources are not renewable, this might counteract the mitigation of climate effects. Furthermore, AI tools and techniques can be abused and misused (intentionally or unintentionally), sometimes harming the people they are intended to help. AI also poses challenges to its own functioning, such as black boxes (researchers and users typically know the inputs and outputs but cannot explain the AI decision-making process) and the replication of social biases (embedded in machine learning dimensions such as the training data, in the algorithms themselves and the interpretation of results). This reality is dangerous as it might result in unforeseeable social consequences such as discriminatory recruiting algorithms, racist and sexist chatbots, or biased legal AI. Thus, a safe and resilient Internet with corresponding governance mechanisms is a further important element to managing risks.
Amid this context, and considering the normative dimension, some pragmatic questions of liability emerge: Who will be made accountable for machines’ decisions or what mechanisms may overturn AI decision making to guarantee social justice and a commitment to sustainability? These questions underpin two points. The first one is to the necessity to ensure that AI applications are carefully and holistically designed to allow for sustainable economic, social and environmental development while safeguarding for respective risks. The second is the relevance of a resilient Internet for AI and AI governance frameworks that consider existing internet governance mechanisms and institutions. In this regard, governance frameworks are necessary to guide development of applications, regulate the use of algorithms and implement AI systems based on the fundamental principles of safety and trust. A first step in this direction is an adequate assessment of policy and legislation frameworks, to help direct the vast potential of AI towards the highest benefit for economy, environment and society. Another could be brainstorming ideas toward guidelines foreseen possible AI consequences or red line uses. Turning to a more comparative, global scale, we face further governance challenges due to differing socioeconomic realities and regulatory approaches to these problems. The knowledge intensive development of AI-technologies may favor countries that are already focusing on technologically advanced sectors, which will perpetuate inequality.
On a policy level, actions and decisions regarding artificial intelligence in one country or region could adversely affect others very quickly. We are seeing this already playing out, where regulatory approaches vary tremendously concerning requirements on risk assessment, data protection and privacy or transparency of AI applications. Therefore, international cooperation is essential to establish common sets of values and standards for the development of AI. This may be achieved through multilateral fora, e.g. IGF or the Global Partnership on AI (GPAI), or through bilateral formats, such as the International Digital Dialogues of the German Government, which it conducts with Brazil, India, and Mexico amongst other countries. Guided by a set of policy questions, this session unpacks political, legal and philosophical dimensions at the interface of AI and sustainable development.
The session aims to trigger discussions to advance a more concrete, genuinely global, and multistakeholder agenda over AI and sustainable development. Therefore, it will critically assess the benefits and threats of the emerging technology for society, environment and the economy. In addition, it seeks to display what tools are available for policy makers to address these challenges, sketching out policy pathways for adequate framework conditions.
The session aims to achieve three main outcomes: (1) to raise further awareness about the nexus of AI and sustainable development, (2) to trigger ideas of AI governance mechanism and possible concrete steps towards unleashing its potential for sustainable development, and (3) to gather information through the polls and the Q&A to frame the debate over the topic. The results of the discussion will be included into the framework of the International Digital Dialogues between Germany and countries such as Brazil, Mexico, India or Japan. The topic of common framework conditions in the realm of AI is a long-standing topic of the Work Plan, which was jointly developed by both government after consultations in multistakeholder rounds. The results will also inform the work of the Digital Transformation Centers, a network advising governments and implementing digital transformation projects in more than 16 countries of the Global South.
Hybrid Format: - How will you facilitate interaction between onsite and online speakers and attendees? The session will facilitate interaction between the onsite and online speakers and audience through active moderation. The session will include online panellists and the organizers plan to have both an onsite and online moderator specifically to engage with both communities adequately. - How will you design the session to ensure the best possible experience for online and onsite participants? Besides making sure that questions will be addressed from the audience in live and the chat, we plan to use interactive tools, such as mentimeter, so that the entire audience interacts online to some extent. This allows for active participation besides asking question. Also, it balances the engagement between onsite, where the session happens, and online, where one part of the audience interaction happens. - Please note any complementary online tools/platforms you plan to use to increase participation and interaction during the session. Audience interaction tools, such as Wisembly or Mentimeter to do quick surveys, create word clouds to steer the discussion.
Usage of IGF Official Tool.
Report
Approaches to global AI governance should be based on transparent and inclusive multistakeholder processes to render them resilient against changing political interests, and they should acknowledge the different realities in the Global North and Global South with regards to digital inclusion. 2:There is a need for AI governance structures that actively promote sustainable development. AI governance should be focused on ethical foundations an
There is a need for AI governance structures that actively promote sustainable development. AI governance should be focused on ethical foundations and safeguard human rights.
Governance for AI must focus on the entire technology lifecycle from development to application to assure ethical AI and safeguarding human rights.
The voice of the Global South must be heard in AI governance approaches to significantly decrease digital inequalities between the Global North and the Global South.
After opening remarks by the Brazilian Ministry of Science, Technology and Innovation (MCTI) and the German Federal Ministry for Digital and Transport (BMDV), the moderator Onike Shorunkeh Sawyerr (GIZ) introduced the audience questions to be answered during the opening statements of the panellists. The idea was to engage the online and on-site audience early to allow for integrating their responses into the discussion. The questions were:
Q1: What do you associate with the term “AI governance”?
Q2: In which area(s) do you see the greatest potential for AI to contribute to sustainable development?
Q3: Overall, do you expect AI to have rather positive or rather negative impacts on sustainable development?
Q4: What risks do you associate with AI?
Urvashi Aneja (Founding Director of Digital Futures Lab, India) described how her institution investigates AI benefits for different areas, including sustainable development. She also highlighted that people still think of AI as a product, and that she considers that framing too narrow. The Digital Futures Lab sees the whole life cycle of AI interventions. For example, she argued that labour conditions for building AI and energy consumption by AI are also important aspects of the discussion. She concluded by stressing the need to improve understanding of the impact of AI on sustainable development.
As the second panellist, Ledénika Mackensie Méndez González (Executive Director for Digital Inclusion at the Secretariat for Communications and Transport of the Mexico City Metropolitan Area) said that AI is a way of innovating the public sector. It requires increasing availability of data as well as transparency. In her statement, she urged governments to assure that AI respects human rights. To accomplish that, ethical challenges must be solved.
Kim Dressendörfer (Technical Solution Leader Data & AI at IBM, Germany) started her statement by expressing that as the technology evolves so quickly, the developers should open the “black box” for everyone, teaching people how to use it properly. Dressendörfer stated that AI is an opportunity to create something new and better. AI governance has multiple layers and also involves the individual developers who must consider the ethical implications of their products. In her presentation, Dressendörfer gave examples of the use of AI: Monitoring animal health in agricultural applications, assisting astronauts on the ISS, and quantifying carbon sequestration in urban forests to store more carbon emissions.
Secretary José Gontijo (Ministry of Science and Technology (MCTI), Brazil) stated that Brazil already accomplished some progress in the field of AI governance by publishing a national AI strategy and starting a discussion on AI regulation within congress. He cited the thematic chambers for the AI strategy, which bring together government, private sector, academia, and civil society to discuss transparency and the applicability of AI. Gontijo also pointed to the ongoing debate between lawyers and technical groups about the regulation of AI and how the legislation should be applied. He highlighted that AI has great potential to boost sustainability. In Brazil, for example, AI may be used in the water management or improvement of the efficiency in agribusiness, in disaster prediction, or public security. Gontijo emphasised the need to reduce the gap between the Global South and the Global North regarding the development and the usage of AI. He expressed that science diplomacy has an important role in reducing the existing inequalities, by keeping technology in the Global South in pace with the Global North and making tech affordable and available for everyone.
Following Gontijo’s opening statement, panellists responded to the first round of questions. Asked about the building blocks of meaningful regulation of emerging technologies, Méndez González highlighted the importance of exchange of experiences between states, to enhance cooperation on building blocks of meaningful regulation of emerging technologies. Resource allocation and distribution to apply these emerging technologies in the countries is crucial to make these regulatory building blocks become a reality. Méndez González expressed that a public policy for sustainable AI should be human centric and intersectional. She highlighted that the inclusion of minorities and marginalised social groups in the AI ethical debate is crucial to reduce inequalities.
Next, Gontijo emphasised the importance of multistakeholder approaches in establishing governance structures nationally and internationally. He acknowledged that it is challenging to find consensus with diverse stakeholders at the table, but once agreement is reached, it provides a strong, broadly legitimated basis. He gave examples of the policies related to the internet and new technologies in Brazil, such as the Brazilian Strategy for Artificial Intelligence (EBIA), e-Digital, the IOT plan (still in development) or the Brazilian internet bill of rights. These strategies, some of them implemented in Brazil since 2010, adopted the multistakeholder approach and it made them able to survive subsequent political changes.
The moderator asked Dressendörfer about the role of the private sector in AI governance for sustainable development. Dressendörfer is convinced that every company has the duty to work towards sustainable AI governance. From that point of view, it is important to bring the ethical discussion into developing teams, so that everyone is aware of the potential positive as well as negative impact of their work. Companies need to make sure people can use AI and that the technology boosts sustainability. She highlighted the challenges that come with the “one-size-fits-all” regulatory approach, as AI is applied differently in many different sectors. Rather, Dressendörfer advocates for transparently describing the algorithms behind AI applications. This allows for meaningful discussions on ethics, human centricity, and governance for sustainable development.
Aneja stressed that there are multiple challenges (economic, social, political, and environmental) related to AI. Politicians sometimes do not have the knowledge about the system, so they rely on the private sector. This can lead to a biased approach. Especially in developing countries, it is harder to regulate the private sector’s influence on political frameworks. She emphasised that risks should already be reduced during the development of AI applications. Building people’s capacities is also a crucial aspect to make AI operate as humanly as possible. Aneja also warned that labour issues do not get enough attention. She ended by pointing to the challenge of building technology as green as possible while dealing with ethical dilemmas as well. While this is not an easy task, reconciling these aspects is of utmost importance.
In the next part, the moderator presented the results of the audience survey. Regarding the first question, the audience associates ideas such as cybersecurity threats, fear, security and regulation with AI standards. Regarding the second question, the audience expects that AI will have rather positive effects on sustainable development. They see this potential mostly connected to global productivity and economic growth, followed by climate action and environmental protection (third question). The main fears of the audience related to AI were digital war, surveillance and misuse that leads to human rights violations (fourth question).
Dressendörfer reacted and argued that people often associate AI with dystopian movies – but she was glad to see that people have good expectations towards the potential of AI, and that AI will not completely replace humans in labour markets. Rather, AI will take over repetitive tasks and support humans to focus on complex assignments. Aneja found it interesting that people associated the use of AI with sustainable growth when still there is a necessity to associate the economic system with sustainability in general. Spreading technologies to other parts of the world is important considering that the majority of the global population still does not even have access to the internet. The hypothesis of the positive impact of AI on sustainable development still must be proven. Although many promises for the future are being discussed, harms are currently still more evident even in developed countries – for example, when looking at the labour conditions for platform workers. According to Aneja, we need more scepticism when talking about AI, as there is still a huge gap between its potential and reality.
After that, Davis Adieno (Global Partnership for Sustainable Development Data, Kenya) was introduced to the panel. He stressed that AI is already a reality, but it is in the hands of the private sector and connects the technological avant-garde rather than the masses. For the civil society, AI and technology seem rather detached from the real world. According to Adieno, our global society has other, more urgent problems, such as poverty and lack of resources. AI has critical ingredients to be an enabler of sustainable development, but for AI to be a solution for the needs of the day-to-day life of society, we need to consider its potential harms next to its benefits.
After the speech made by Adieno, the floor was opened to the questions from the audience. The majority of questions revolved around the objectives, potential and risks connected to AI governance. Aneja as well as Dressendörfer agreed that the quality and the management of data must be improved as a precursor for more meaningful technological developments – as more data does not mean better data.