Sunstone Institute publishes open analysis on thirty of the largest threats to humanity, from climate to conflict to democracy. The impact we have largely depends on our choice of topic, and today that choice is made purely with human judgement.
The question we are working towards is whether it can be made systematically: for a given threat, what is actually being done about it, where is the gap largest, and can new public analysis change anything? That requires representing threats, interventions and the evidence between them in a form you can query and reason over. This talk covers the groundwork: an ontology relating threats to intervention mechanisms, LLMs used in a deep-research fashion to draft domain maps we then curate into SKOS vocabularies, claim extraction with NER and ontologies, and HornDB, an OWL 2 RL reasoner we wrote in Rust.
A lot of this work is still not done: comparing threats across domains as different as health and democracy is an open problem. I am giving this talk partly to describe our approach and partly to connect with people who have worked on related problems.
Speaker: Stig Bakken, CTO Sunstone Institute
