Societal Challenges

Artificial Intelligence is reshaping how we live, work, and interact (bringing immense opportunities but also profound societal challenges). From ensuring ethical decision-making in autonomous systems to bridging digital divides and empowering individuals with control over their data, the responsible development and deployment of AI is critical. 

These seven societal challenges guide our mission to harness AI for the benefit of all.  

1. Responsible AI:
   How can we ensure that autonomous systems (such as self-driving cars, drones, or AI-driven decision-making tools) operate ethically, lawfully, transparently, and accountably?
2. Digital Literacy & Inclusion:
   How can we ensure that all citizens (regardless of age, background, or location) develop the digital skills needed to thrive in an AI-driven world?
3. Data autonomy & sovereignty:
   How can we empower individuals, organizations, and governments to retain control over their data in an era dominated by big tech, where autonomous systems and AI rely on vast, often opaque datasets?
4. Human-centric AI:  
   How can we design AI systems that genuinely augment human capabilities (rather than replace them) while ensuring users trust, understand, and retain control over these technologies?
5. Applied AI:  
   How can we apply AI in domains like healthcare, energy, or industry in ways that maximize societal benefit?
6. Explainability:
   How can we create AI tools that explain why they reached a particular conclusion, not in terms of a confident-sounding story that LLMs can cook up, but actually a traceable chain of reasoning based on a modest set of weights assigned to features or measurements. This goes beyond transparency and is essential in building trust in e.g. medical applications.
7. Efficiency:
   How can we develop AI systems that do not need tons of accurately labelled data and huge compute facilities. The need for massive labelled data sets requires a lot of human input and makes curating these data sets hard, and the presence of label noise degrades the accuracy of the resulting AI system. The huge compute requirement is costly, both in terms of equipment and carbon footprint. This is unsustainable in the long run.