Menu
November 2022
BSc (UNAM), MPhil (ENS), PhD CompSci (Lille), PhD Phil (Sorbonne). Elected member of the London Mathematical Society, Fellow of the Royal Society of Medicine, member of the Canadian College of Health Leaders
The book published by Springer Nature offers an alternative approach to algorithmic complexity rooted in and motivated by the theory of algorithmic probability. It explores the relaxation of the necessary and sufficient conditions that make for the numerical applicability of algorithmic complexity better rooted in the true first principles of the theory and what distinguishes it from computable or statistical measures, including those based on (lossless) compression schemes, such as LZW and cognates, that are in truth more related to traditional Shannon entropy and demonstrably inadequate to characterising causal mechanistic principles. The methods introduced in the book are at the foundation of the field I introduced and is covered by the second book, Algorithmic Information Dynamics, a type of digital calculus for causal discovery and causal analysis in software space, the result of combining perturbation and counterfactual analysis with algorithmic information theory.
I help create mathematical and computational tools to find new theoretical and methodological approaches to reveal the causes or mechanistic explanations of natural phenomena as computable models. This to transform the way in which we do and apply science, particularly in the context of living systems, cell and molecular biology.
I have held academic and industry positions in 7 countries (Mexico, U.S., France, Sweden, KSA, UK, and the UAE). Over the last 10 years, I have been associated with the universities of Oxford (Kellogg College – Structural Biology Group, Department of Computer Science), and Cambridge (Machine Learning Group, Department of Chemical Engineering and Biotechnology), The Alan Turing Institute, the UK’s national institute for data science and AI, supported by the ONR/U.S. Navy, DoD and based at the British Library in London, and the Center for Molecular Medicine (and SciLifeLab) at the Karolinska Institute (the institution that awards the Nobel Prize in Physiology or Medicine in Stockholm, Sweden). I have held senior staff and faculty positions in these institutions running the gamut from Assistant Professor, Senior Researcher, Principal Investigator, to Lab leader, as well as serving as an advisor to policy making for institutions like the OECD (on AI for Scientific Discovery).
Gregory Chaitin, one of the founders of modern computer science and complexity theory who provided a proof equivalent to Gödel’s incompleteness theorems in the 60s based on computability and information theory, described me as a “new kind of practical theoretician“. I introduced the field of Algorithmic Information Dynamics (AID), a new and exciting field devoted to the study of dynamical systems in software space using the power of Artificial General Intelligence methods.
I have published over 140 peer-reviewed papers in the top journals in the areas of physics, computer science, computational intelligence, molecular biology, complexity science, and finance, including the Philosophical Transactions of the Royal Society, Nature Communications, Physica A, Physical Review E, Nature Methods, Bioinformatics, Nature Machine Intelligence, Nucleic Research Acids, among others. In more than 90% of them I am either first or corresponding author.
In 2018, I founded Oxford Immune Algorithmics, an award-winning deep biotech spinout from the University of Oxford, graduated from CDL (U.S. & Canada) and SpinLab (Germany), operating multinationally (UK, Canada, US, India, UAE, Bangladesh) that has raised about 8M USD to transform medicine and health care by using the power of Artificial General Intelligence (AGI) to learn from and monitor each individual person’s immune system.
In the U.S., I was a visiting scholar at Carnegie Mellon (Pittsburgh, PA) and a member of the NASA Payload team for the Mars Biosatellite project at the MIT in Boston (Cambridge, MA), where I was in charge of the development of the tracking software to study the effects of microgravity on small living organisms. Also in the U.S. (Boston, MA and Urbana, IL), I contributed to the code in Wolfram|Alpha, the A.I. engine behind Apple’s Siri and Alexa, that enables them to answer factual questions.
I am also the Managing Editor of Complex Systems, the first journal in the field of complexity founded by Stephen Wolfram in 1987. Additionally, I serve as Editor for several journals including Entropy, Information, Frontiers in AI, and Complexity; and for book series such as Springer’s on Complexity.
I have also published over 8 books including A Computable Universe, with a foreword by Sir Roger Penrose (Nobel Prize in Physics 2020). My work has been featured in Wired, The New York Times, Le Monde, The Independent and the MIT Technology Reviews, which are among about other 40 tech magazines and national newspapers to do so.
I am also an invited scientific advisor and mentor at the Creative Destruction Lab (CDL) at the Saïd Business School, University of Oxford. CDL is one of the most reputable accelerators with 8 sites around the world.
One year before passing away, Marvin Minsky, widely considered the founding father of Artificial Intelligence, made the following astonishing claim describing what turns out to be exactly my own research in a closing statement at a prime venue (video on the right and excerpt below). I wish I could tell him that I have gone farther than anyone else at what he thought everybody should be doing!
One of my research aims is to reintroduce symbolic computation into statistical machine learning to alleviate current limitations in techniques such as deep learning along the lines of the criticisms put forward by people such as Sydney Brenner, Marvin Minsky and Judea Pearl and believed to be fundamental to make further progress in AI research. Here a video produced by the journal Nature to explain my research:
I got interested in neural networks from the standpoint of computability and complexity theories in my early 20s when I was writing my final year memoir for my BSc degree in math at UNAM. Today, I am helping revolutionise the field by reintroducing the theories of computability and algorithmic complexity back into AI and neural networks.
On the right, an image showing how a deep neural network trained with a large set of fine art paintings ‘sees’ me. My current research consists in helping machine and deep learning see beyond these statistical patterns in more clever ways than simple pattern matching. By introducing algorithmic probability to Artificial Intelligence I help the field to reincorporate abstract thinking and causation in current AI trends.
Known to underperform in tasks requiring abstraction and logical inference, current approaches in deep and machine learning are very limited. An example of our research in this direction is our paper published in Nature Machine Intelligence that can be read here for free (no paywall).
Composition of ECA rules 50 ◦ 37 with colour remapping leading to a 4-colour Turing universal CA emulating rule 110.
Composition of ECA rules 170 ◦ 15 ◦ 118 with colour re-mapping mapping leading to a 4-colour Turing universal CA emulating rule 110.
As reported in our paper published in the journal of Cellular Automata, we proved that these two 4-colour cellular automata are Turing universal, found by exploration of rule composition. This means that these CAs can, in principle, run MS Windows and any other piece of software (even if very inefficiently).
These new CAs helped us show how the Boolean composition of two and three ECA rules can emulate rule 110.
This also means that these new CAs can be decomposed into simpler rules and thus illustrates the process of causal composition and decomposition.
The methods also constitute a form of sophisticated causal coarse-graining learning that we have explored in other papers such as this one. In the same paper, we also introduced a minimal set of ECA rules that can generate all others by Boolean composition.
In this other paper, we also found strong evidence of pervasive universal computation in software space
On Philosophy and Epistemology of Deep Learning:
I contributed to the final materialization of the Leibniz-Chaitin medallion after Leibniz’ original design 300 years ago to celebrate his discovery of binary arithmetic
The Leibniz-Chaitin medallion story in celebration of the works of Greg Chaitin and the discovery of binary arithmetic from which, according to Leibniz, everything can be created
I have been to more than 300 cities in about 50 countries giving talks related to my
work in about half of them, and as an invited speaker in 15:
Histograms of number of cities visited (with at least 4 cities) and continents:
