Paul-Olivier is (according to Paris-Match) one of the world's leading experts on data management. He is an entrepreneur in the field of artificial intelligence, through his company hestia.ai and their product Argo, a sovereign platform building artifical intelligence systems through compositionality. He has extensive experience investigating and reverse engineering data ecosystems, from a data, technical and financial standpoint. Public landmark achievements include:
- Uncovering then leading, as a citizen, the investigation into the Facebook-Cambridge Analytica scandal, leading to the largest one-day stock market crash in history for a single company (119 $Bn), a Netflix movie (The Great Hack), the bankruptcy of 30+ separate companies, and numerous testimonies in public hearings by lawmakers around the world (UN Human Rights Council, UK Parliament, Swiss Parliament, European Parliament, Council of Europe, World Trade Organization,…);
- Helping journalist Judith Duportail get her data from Tinder (she got 800 pages!), and written an op-ed about the experience;
- Reverse engineering Uber’s routing, matching and pricing algorithms in Geneva and France.
Former life and intrinsic motivation
Paul-Olivier was a research mathematician (Stanford PhD under Dan Bump, Oxford JRF, ETH Zurich Heinz Hop Lecturer, IHES visitor, University of Zurich Swiss National Foundation Professor). He worked on the project Combinatorics of partitions and number theoretic aspects, whose agenda was to build bridges between combinatorics and number theory, circumventing the need for random matrix theory in so-called Random Matrix Theory conjectures (quantified conjectures relating to the Riemann Hypothesis), and attempting to get to the core intuitions that physicists have when they think about number theory problems. This has led to a preprint entitled Combinatorics of lower order terms in the moment conjectures for the Riemann zeta function, never accepted through peer-review as it merely rephrased existing conjectures. While this reformulation was meant to attract more diverse insights and intuition into the problem - a step that might be considered worthwhile on its own - that takes time. In consequence, Paul-Olivier refocused his attention on how mathematicians collaborate and how to overcome the “One Brain Problem” that was very characteristic of math research. That led him to be part of what were the largest collaborations of (pure) mathematicians at the time: the Open Source Mathematical System Sage and the L-Function and Modular Forms Database. He also initiated the OpenDreamKit project. This project aimed to combine Data, Knowledge and Software into coherent tooling to assist research mathematicians (a flagship EU Research Environment). His key insight in structuring the project was that Florian Rabe and Michael Kohlhase’s research and tooling on flexiformalization would be applicable not only to interlink formal theorem proving systems, but also computer algebra software actually used for research mathematics. Based on all these learnings, he also submitted an ERC-like grant proposal with the Swiss National Science Foundation in 2014 entitled Engineering Discovery in Mathematics. This did not get funded, as sort of expected: it was considered too speculative by the reviewers. Instead, Paul-Olivier left academia and focused on changing the societal perspective on personal data (no small task, see above).
Where is this going
LLMs and machine learning bring about two revolutions. The first is this idea of mutualized deduction pathways. The second is increased accessibility of this tooling through natural language. Research mathematicians have picked up on this, and like many other scientists are starting to push in the direction of “augmented researchers” with strong confidence that it will benefit their work (e.g. Llemma, AstroLlama). In mathematics, this takes place especially through the use of formal theorem provers, particularly around LEAN. As a research mathematician, Paul-Olivier sought to build bridges between multiple areas of science: number theory, combinatorics, random matrix theory as practiced by physicists and random matrix theory as practiced by mathematicians. He is now focusing his passion as an entrepreneur on the entirety of human deductive powers, and organizing an ecosystem preserving what he calls meaningful intelligence. This term can be afforded a mathematical definition, which actually helps in propulsing his unpublished preprint forward, and finding meaning in what he has decided to apply his intelligence to: Argo.