Mathematics for Safe AI

Document type: Opportunity space

Author: David 'davidad' Dalrymple

Date published: 23 November 2023

Version: 1

DOI: https://doi.org/10.67475/mfsai8f7a-s28f

 

Abstract

We don’t yet have known technical solutions to ensure that powerful AI systems interact as intended with real-world systems and populations. A combination of scientific world-models and mathematical proofs may be the answer to ensuring AI provides transformational benefit without harm.

Download the full text [PDF - 4.64Mb]Download the full text (accessible) [PDF - 632.61Kb]

Mathematics for Safe AI was one of ARIA's original opportunity spaces. Following an evolution of our research portfolio in August 2026, it now sits within the Trust Everything, Everywhere opportunity space. The beliefs and observations that defined Mathematics for Safe AI still hold, and its remit continues in full within Trust Everything, Everywhere with respect to both future programmes and seed funding. No active funding was affected by this change.