Bottom-up AI and the right to be humanly imperfect | IGF 2023

Knowledge Graph of Debate

Session report

Jovan Kurbalija, Executive Director, DiploFoundation

Jovan Kurbalija, the Director of the Diplo Foundation, professes the significant intersection of philosophy, technology, and artificial intelligence (AI), particularly concerning education, cultural context, governance, and ethics. He promotes profound understanding of technological advancements without becoming engrossed by its complexities, thereby maintaining a steadfast focus on the broader societal and philosophical effects.

At the heart of Kurbalija’s argument is the Diplo Foundation’s innovative development of a hybrid system. This unique construct, merging artificial intelligence with human intelligence for reporting, has been cultivated based on the Foundation’s extensive experience and session management. The potential capabilities of this system in promoting dynamic learning environments and stimulating intellectual engagement were also highlighted.

Adding a fresh perspective to the discourse, Kurbalija proposed that AI models should harmonise with each community’s distinct traditions and practices. He believes this would contribute to a more authentic, bottom-up AI model that does not limit itself to predominantly European philosophical traditions. In a similar vein, he emphasised the urgent need for high-quality data in developing diverse, flexible open-source AI models.

However, he stressed the importance of preserving individual and community-based knowledge rights, protecting against its potential commodification by AI. Kurbalija highlighted concerns regarding transparency and explainability within AI applications, allied with apprehensions about AI’s misuse in creating disinformation.

Certain aspects of AI’s current governance invoked criticism, notably the sidelining of smaller entities by larger corporations. A call was made for increased corporate responsibility due to the extant challenges related to AI usage. Despite AI’s potential in preserving small communities’ heritage and culture, a significant gap was recognised concerning the lack of initiatives that leverage AI to safeguard cultural diversity.

While acknowledging AI’s potential in aiding individuals with disabilities, caution was raised about anthropomorphising AI, reinforcing that AI should serve as a tool, not as a master. The uniqueness and imperfection of human traits were lauded as invaluable characteristics and were claimed to be essential considerations in the development of AI.

Sorina Teleanu, Director of Knowledge, DiploFoundation

The analysis unveils an assemblage of sentiments regarding the application of Artificial Intelligence (AI) in multifaceted domains such as negotiations, decision-making, educational sectors, foreign affairs, and surmounting challenges faced by smaller and developing nations.

A positive aspect of AI is enlightened in its capacity to support complex decision-making procedures and foster critical thinking within educational environments. The effectiveness of AI in enhancing decision-making and negotiation is showcased in the global digital compact simulation. The AI advisor was utilised to refine arguments and language, whilst being trained to offer details on digital policy and internet governance. Further, in the realm of education, dismissing the use of AI in schools is argued to be counter-productive. The significance of AI in stimulating critical thinking and understanding intricate policy matters is underscored, thereby highlighting its role in shaping quality education and nurturing innovation.

However, the sentiment isn’t unequivocally positive. The analysis also uncovers AI’s limitations, stressing the importance of its critical application. Instances where AI hallucinates and doesn’t always deliver perfect results have been pointed out, demonstrating that although AI could be a valuable tool, it must not be relied upon blindly.

The evaluation also delves into the struggles of small and developing nations, particularly in digital governance and diplomacy. The overwhelming volume of information and tasks, combined with limited resources and a dearth of time, often poses significant challenges for these countries, thereby requiring the use of AI for effective decision-making and negotiation.

AI’s significance in foreign affairs emerges as it economises time and provides diplomats with a foundation for negotiations. Ministries of Foreign Affairs are encouraged to develop their own AI systems to retain control over data, relying on their knowledge base and experience. The concept of ‘bottom-up AI’ is proposed, arguing that it could allow a more controlled and tailored use of AI, and return AI back to users.

The potential of AI to promote underserved communities and mitigate representation inequalities is also explored. Bottom-up AI’s development based on knowledge from these communities bolsters this argument, aided by the observed stance that AI can encourage more meaningful engagement for smaller countries.

Nevertheless, despite the proposed benefits, the need for transparency and accountability of AI systems is underscored, with apprehensions regarding the non-explainability of neural networks being raised. There is significant criticism regarding uncritically accepting statements from large AI systems and a generic tendency for blind trust.

The evaluation concludes by emphasising the importance of addressing current AI issues, such as regulation, before getting consumed with future challenges. Large firms are depicted as demanding future AI regulation whilst disregarding existing issues, prompting a call for allocating resources to counter today’s challenges before concerning ourselves with future ordeals.

In harmony with Sustainable Development Goals (SDGs) 4, 9, 10, 16, and 17, the overall analysis accentuates the potential of AI in driving innovation, assisting in quality education, reducing inequalities, aiding in institution-building, and fostering partnerships. Nevertheless, the pivotal importance of careful, regulated, and transparent usage of AI is underscored.

Audience

The discourse unveiled a plethora of critical points spanning numerous subjects. A significant challenge was identified in Brazil with regard to technology – a substantial number of NGOs are grappling with integrating technological approaches due to lacking tech literacy. This issue hampers these organisations from fully capitalising on their potential in their operations, suggesting the necessity for dedicated digital literacy programmes.

Interestingly, the proposition was raised that augmenting participation and representation in tech-related matters could bolster the advocacy of local perspectives. This argument was underpinned by the desire to categorise knowledge in a manner that respects and supports local viewpoints, shining a spotlight on an essential consideration in the democratisation of technology and inclusivity.

The discussion then veered towards concerns about the economic ramifications of automation. Technological tools such as chatbots in Brazil’s service sector usage have soared, stirring anxieties surrounding potential structural unemployment and escalating the possibility for diminished economic opportunities and job security. In view of this, there was concurrence on the need for a paradigm shift to orchestrate the origination of dignified, rewarding economic opportunities.

The discourse additionally exhibited a robust belief in innovation and its prospective benefits. Participants conveyed stout support for a bottom-up Artificial Intelligence (AI) approach and open-source methods for managing knowledge on a grander scale. The capacity of these methods to organise and categorise knowledge with sensitivity to local perspectives was seen as a hopeful potential.

However, feedback and constructive criticism were deemed essential for the amelioration of larger systems. Questions were raised about whether insights from these systems were being considered and whether prevailing systemic problems required addressing, indicating a need for rigorous examination and rectification of these systems.

A particularly thought-provoking point in the discourse was the expression of concern regarding the rapid displacement of families due to the expanding influence of modern technology. This issue particularly afflicts rural areas of Brazil, leading to a diminution of the countryside and augmentation of cities. This cultural and knowledge erosion is significant, especially in small communities.

A suggestion was forwarded in response to these challenges to utilise AI to preserve and cultivate the history and culture of small communities. This would involve AI assisting in updating and uploading knowledge about these areas, spanning physical practices, agricultural practices, stories, and mythologies.

One neutral sentiment proffered revolved around AI’s design and adaptability, specifically tailored towards individuals with disabilities. Current AI systems are often trained on ‘perfect’ data, potentially making them less adaptable to human error. Conversely, humans are able to learn from their mistakes. Consequently, developers must cultivate more adaptable AI that can accommodate humanlike errors.

In a related argument, it was posited that AI should be enhanced to aid persons with disabilities rather than marginalising them. There is apprehension that current AI protocols might inadvertently engender a standard of ‘perfection’ that could be exclusionary, particularly for individuals with disabilities. However, by ensuring AI is a tool for inclusivity rather than exclusion, an opportunity arises.

In sum, these insights prompt a reassessment of how technology, specifically AI, is utilised and incorporated into diverse sectors of society. The call is widespread for more tech literacy programmes, adaptable AI, and active involvement in technology decision-making. These transformations would contribute significantly to striking a healthy balance between swift technological progression and preserving crucial aspects of our cultural heritage and humanity.

In conclusion, Kurbalija’s discussions presented a potent outlook on AI’s broad societal impacts, issuing an urgent summons for more inclusive and ethical AI development, whilst highlighting concerns regarding transparency, accountability, and the conservation of local cultures and individual rights.

Speakers