From diverse datasets to United Nations public good tasks

Go back to programme

From diverse datasets to United Nations public good tasks

Background 

The Digital Age has opened new frontiers in Machine Learning (ML) and Artificial Intelligence (AI), promising transformative advances across various domains. To harness this potential for the greater good, a collaborative framework for data standardization and sharing is imperative. This workshop proposes to congregate leaders and innovators in AI/ML research, domain experts in critical fields, and representatives from the United Nations to cultivate a partnership aimed at forging a shared data protocol for AI. 

Rationale 

Drawing inspiration from the inaugural DMLR workshop at ICML 2023, we recognize the need for a paradigm shift from model-centric to data-centric AI development. The shift became necessary because the main obstacle of developing AI-based solutions is not the underlying AI technology anymore, but rather the compatibility and robustness in scope of real-world data (as opposed to well controlled and curated research data). This workshop will explore how a unified data protocol can enhance access to ML for public good tasks. This is based on the believe that data owners (at the same time the problem owner) and solution developers are generally available, but do not match easily. These tasks, drawn from critical application domains such as health, agriculture, and space, will be mapped to machine learning research questions, creating a rich metadata ecosystem where data can be accessed and leveraged more effectively. 

Workshop Goals 

  1. Initiate the development of a ‘Data-transform Language’ (DTL) that standardizes ML meta-language across platforms. 
  1. Leverage the potential of ML for a connected understanding of complex public good problems, drawing from the diverse datasets provided by partners. 
  1. Address transaction costs in AI/ML. 

Structure 

Keynote presentations from the curators of the AI for Good Discovery Track, providing insight into the current landscape and potential of data-centric AI for public good. 

This will be followed by sessions focusing on the three primary ‘transaction costs’ in AI/ML and proposing solutions for language heterogeneity, commercialization interfaces, and signal propagation. 

Through this workshop, we aim to pave the way for a future where AI research aligns seamlessly with global needs, informed by data-centric approaches and unified protocols. The envisioned partnership will be a step towards creating equitable and universally accessible AI, thereby advancing the UN’s Sustainable Development Goals. Join us in Geneva to contribute to this vital discourse and take actionable steps towards a data-inclusive future 

 

Share this session

Are you sure you want to remove this speaker?