Bottom-up AI and the right to be humanly imperfect (DiploFoundation)

6 Dec 2023 14:00h - 15:00h UTC

official event page

Table of contents

Disclaimer: This is not an official record of the UNCTAD eWeek session. The DiploAI system automatically generates these resources from the audiovisual recording. Resources are presented in their original format, as provided by the AI (e.g. including any spelling mistakes). The accuracy of these resources cannot be guaranteed. The official record of the session can be found on the UNCTAD website.

Full session report

Marilia

The NetMundial final document has raised concerns regarding the annotation and weighting of information in political documentation. This has shed light on the challenges involved in annotating and weighting information within political environments. While this process is considered crucial for determining the significance and influence of different contributions, it is also fraught with difficulties due to the subjective nature of political decision-making and the potential for bias.

To create the NetMundial final document, various contributions were carefully read through and assigned weights. These weights were then taken into account in the final text. However, the specifics of the weighting process remain unclear.

Workers and participants have emphasized the need to be part of the working groups to ensure that their views are properly represented and given appropriate weight. They recognized the significance of having their perspectives incorporated into the final document and wanted to ensure they were given fair consideration.

The sentiments surrounding this issue are negative and concerned. The negative sentiment stems from the challenges and potential pitfalls associated with annotating and weighting information in political environments. The concerns are likely due to the subjective nature of political decision-making and the lack of transparency and fairness surrounding the process.

It is important to note that the NetMundial final document touches on related topics such as political movements and the importance of political participation. These topics are linked to the broader objective of SDG 16 (Peace, Justice, and Strong Institutions), which promotes fair and inclusive governance systems.

In conclusion, the NetMundial final document highlights concerns about the annotation and weighting of information in political documentation. It acknowledges the significance of including diverse perspectives but also underscores the challenges and potential biases involved in the process. The negative and concerned sentiments reflect the need for transparency, fairness, and inclusivity in political decision-making.

Jovan Kurbalija

Artificial Intelligence (AI) is highly praised for its ability to discover intricate patterns in vast amounts of data and make accurate predictions based on probabilities. This advancement is possible due to the simplicity of AI, which relies on patterns and probability. The application of AI in various industries, such as healthcare, finance, and transportation, has brought about significant advancements and improvements.

Competition in AI development is deemed necessary to prevent exclusive control and dominance by large companies. It is argued that when a few major players have complete control over AI development, it can hinder innovation and limit access to AI technologies and resources. To address this, there is a growing call for the growth of open-source AI models, which would promote collaboration, knowledge sharing, and inclusivity in AI development.

The ownership and provenance of AI knowledge have emerged as critical considerations. It is argued that AI should clearly point to the intellectual or writing source upon which it is built. This ensures fairness, transparency, and shared ownership of AI knowledge. An example of this is demonstrated by Diplo’s use of AI for constructive teaching and learning, highlighting the importance of attributing the sources of AI knowledge.

In the realm of education, AI is shown to have a positive impact on individual learning. By assisting students in understanding how it thinks, AI enables them to grasp complex subjects more effectively. However, there have been concerns regarding the ban of AI in education. Instead, it is suggested that academic institutions focus on promoting critical thinking while utilizing AI as a beneficial tool for enhancing the learning process.

The scalability of AI annotation and weights is considered technically feasible. Experts believe that expanding the annotation and weight systems can contribute to the accuracy and usability of AI models. This advancement has the potential to further enhance the effectiveness of AI technologies across different sectors.

However, the computerization of political processes raises significant concerns. There is a belief that politics, being a fundamentally human activity, requires a human touch. Caution is urged regarding the quantification and use of AI as an additional tool in political processes, emphasizing the need to preserve the integrity and human element of political decision-making.

Preserving knowledge through annotation is particularly emphasized for Ministries of Foreign Affairs. The way of thinking about and solving problems is considered an asset that should be preserved. By properly annotating and documenting knowledge, Ministries of Foreign Affairs can ensure the continuous development and improvement of their policies and actions.

The selection of AI systems is an important decision that the United Nations will face. The implication of this choice was demonstrated with UNDP’s supposed support for an AI system predicting the Israel-Palestinian conflict. This highlights the significance and potential consequences that come with choosing the right AI technology.

In conclusion, AI’s ability to discover patterns and make predictions based on probabilities has revolutionized various industries. Nonetheless, there are important considerations related to competition, ownership, and provenance, education, politics, and knowledge preservation. The expanded summary provides a comprehensive overview of these key points, emphasizing the essential role and potential impact of AI in our society.

Yung-Hsuan Wu

The debate centres around the outsourcing of important information highlighting to third-party tools. One speaker argues that this practice risks misinterpretation and a lack of control. They highlight that many commercial tools offer text summarisation and generation, but their reliance on other annotators could lead to inaccurate interpretation. This suggests that when an institute outsources to a third-party tool, they are relinquishing control over what is considered important, and the process becomes vulnerable to potential misinterpretation.

Furthermore, the speaker expresses a negative sentiment towards outsourcing, emphasising the potential risks involved. By depending on third-party tools, the speaker argues that there is an increased risk of misinterpretation and a lack of control over the summarisation and highlighting of information. The argument suggests that important details may be overlooked, or not given the appropriate emphasis, potentially leading to misunderstandings or an incomplete understanding of the information at hand.

On the other side of the debate, another speaker takes a different stance, arguing that institutional thinking can only be strengthened by the individuals within the institute. They assert that institutional thinking is enhanced through the active participation of its members. To support this claim, they provide an example from Diplo, which exemplifies how institutional thinking is nurtured when individuals within the institute are involved in decision-making and interpretation processes.

From the analysis of these arguments, it can be inferred that while third-party tools offer convenience and efficiency in text summarisation and generation, there are inherent risks associated with outsourcing important information. This can result in a loss of control over interpretation and the emphasis placed on certain details. Conversely, the importance of institutional thinking is highlighted as a means to strengthen decision-making and understanding within the institute.

In conclusion, both sides present valid points in the debate. The negative perspective emphasises the need for control and highlights potential risks associated with outsourcing important information to third-party tools. Conversely, the positive perspective underscores the value of institutional thinking and the role of individual participation in making informed decisions. Ultimately, this debate reveals the significance of carefully considering the pros and cons of outsourcing when it comes to the highlighting of important information.

Parminder

In this analysis, the speakers explored several significant concerns and interests regarding artificial intelligence (AI). One of the primary concerns raised was the potential bias present in pre-trained AI models. They highlighted that generative AI learns patterns from extensive amounts of data, even though it ultimately deals with smaller datasets. This phenomenon has raised concerns about the accuracy and objectivity of the patterns identified by pre-trained models. Additionally, they noted that these models can inadvertently perpetuate certain biases present in the training data, which can have serious ethical implications.

Another topic of discussion was the dominance of big AI systems over smaller ones. Despite the value that small AI or small data systems may offer, there is a general trend towards larger systems dominating the AI landscape. This has sparked concerns about the potential stifling or marginalisation of small AI systems. The implications of such dominance raise questions about innovation, diversity, and competition in the AI field.

The importance of policy interventions to preserve the identity of small AI systems was also highlighted. The speakers argued that without such interventions, small AI systems may struggle to maintain their individuality and distinctiveness in the face of larger, more powerful systems. They proposed policy measures to ensure that small AI systems are adequately supported and given the opportunity to flourish.

Another noteworthy aspect discussed was the preservation of data and source provenance in AI. The speakers emphasised the significance of retaining the knowledge generated by AI nodes and protecting the sources within the AI system. They urged that AI nodes should be designated and that they have a responsibility to safeguard the underlying sources of data. This approach ensures the integrity and traceability of the information produced by AI systems. It aligns with a collectivist notion of preservation, where the responsibility lies with the nodes themselves.

Overall, the analysis covered concerns related to biases in pre-trained AI models, the dominance of big AI systems, the need for policy interventions to support small AI systems, and the importance of preserving knowledge and source provenance in AI. The speakers concluded that addressing these concerns and interests is crucial for the ethical and sustainable development of AI technology.

JK

Jovan Kurbalija

Speech speed

144 words per minute

Speech length

6847 words

Speech time

2861 secs

M

Marilia

Speech speed

193 words per minute

Speech length

263 words

Speech time

82 secs

P

Parminder

Speech speed

182 words per minute

Speech length

552 words

Speech time

182 secs

YW

Yung-Hsuan Wu

Speech speed

168 words per minute

Speech length

213 words

Speech time

76 secs