A blog by Professor Sheri Markose in which she answers questions on how she has arrived at her Genomic Nash Equilibrium with Novelty Production
Let us start with the 2019 keynote talk On the Digital Foundations of Intelligence: How We Became Smart and Protean: you must get asked this a lot, but as an Economist how did you pick this topic?
For at least two decades, I have been aware that mainstream Decision Sciences and Game Theory have been remiss in regard to restricting rational choice to what is already known in a pre-specified set and nothing new can happen. Novelty and innovation are rampant in economic systems, but the only model we have for this is statistical white noise. Even worse, in Behavioural Economics, the one defining behaviour of humans that we are endlessly adaptive and protean, capable of ‘thinking outside the box’, and innovation in arms races is simply not addressed. Indeed, extant Game Theory says novelty and surprises cannot exist in a Nash Equilibrium of a game. So, how are we capable of novelty production? Novelty comes in two forms, within an organism as in novel antibodies and in so called extended phenotypes, to use a term from Richard Dawkins, for artifacts that lie outside organisms.
So why the Digital Foundations of Intelligence?
I was tipped off early about this in my training in Game Theory at the London School of Economics by a famous Game Theorist called Ken Binmore. In his 1987 paper called Modelling Rational Players, Binmore raised the spectre of Kurt Gödel and said: “What about Gödel’s Liar?” Gödel’s Liar is a software agent which can negate what it can predict. Effectively, this is like a contrarian agent, the virus or malware agent which is adversarial. According to Binmore this is what necessitates the exit from a given action set and produces Novelty and Surprises. The implication that only advanced software or digital systems with Gödel self-referential and Liar-like negation operators will produce novelty is already in the literature in what is called the Wolfram-Chomsky Schema. Gödel with Alan Turing and Emil Post laid the foundations of Computation Theory and it is Turing’s idea of machine executable code which underpins the digital economy. I started teaching Gödel Logic and the Liar/Contrarian Strategy to Masters and PhD students at Essex when I set up the Centre For Computational Finance and Economic Agents to alert them that the world cannot be closed and complete and radical uncertainty will arise in the form of Gödel Incompleteness.
So when did you get the breakthrough that the Gödel logic could help model intelligence?
About 2015 an Italian PhD student of mine, Simone Giansante, pointed out that one of the conditions of Gödel digital systems has been discovered by neuroscientists at the Parma university in the brain of primates and they call it Mirror Neurons ! ! It is an equation from a test book by Hartley Rogers on Recursive Function Theory (another term for Computation Theory) which shows that Gödel systems are not about machines running their own codes but a 1-1 offline record of machines running their own codes. My antennae went up. I invited Vittorio Gallese to Essex to give a talk, along with John Kelso who had found that the human MNS can also recognise negation of a prediction in a social cognition experiment. This is the basis of Theory of Mind capabilities involving deceit and counterfactuals. I began to think, it must be exactly like it says on the tin of Gödel logic. In 2017 I published a paper in which this evidence was used to prove there is a Nash Equilibrium of a game in which novelty and surprise follow as in Gödel logic. Mikhail Prokopenko (Director of University of Sydney Complexity Centre) and the Editor of the Frontiers Computational Intelligence journal is on the same page as me. He invited me to become an Associate Editor of the journal and of course invited me to Sydney to give a Keynote talk in December 2019.
You have coined the term Genomic Intelligence, Why?
Genomic Intelligence is to underscore the point that our intelligence evolved from what Walker and Davis have called the ‘algorithmic takeover’ of Biology four billion years ago with the digitization of inheritable information in the genome. By chance in around 2017 I stumbled on the epochal discovery by Nobel Laureate Barbara McClintock that viral software called Transposons and Retrotransposons (which, respectively, scissor paste and copy paste) is the vehicle of evolvability and have recently been found in brain plasticity. This underscores the truism that only software can change software and along with benign changes, this viral software also has malign effects which must be kept in check. So, from the start of life there has been a digital game between self-codes of a host and viral agents. This appears to have led to a unique digital genomic information processing system, which is highly self-referential, geared to complex self-other interaction, especially with hostile malware/viral agents, and having access to open ended search for novelty à la Gödel formal systems. Self-codes are ‘theorems’ in the genomic system.
You say Genomic Intelligence made a step change with the so-called Big Bang of Immunology in the lineage of jawed fish. How so?
In about 2018, to my complete amazement, I found that the Adaptive Immune System from about 500 million years ago also forms an offline/virtual record in the Thymus Medulla of 85% or so of expressed gene codes that produce morphology and somatic identity of eukaryotes exactly like the Mirror Neurons record online self-actions. And why is this done? Both the immuno-cognitive systems make out the other, especially the hostile other, from self-codes or self referentially. The AIS and the human brain are the only two systems in which offline virtual simulations in the order of 1020-1030 can be done based on self-codes. In the former to see how they may be hacked by nonself antigens, and in the latter for social and higher order cognition.
You say that far from being funky, esoteric constructions in the foundations of Mathematics, Gödel Sentences are ubiquitous in immuno-cognitive systems, Why?
Yes, in the 2021, based on my Keynote talks in Sydney and CMU, I published a paper in Entropy Special Issue on Bio Complexity. I explicitly show that the Adaptive Immune System can generate Gödel Sentences as fixed points of novel viral software that allows a gene code to self-report it is under attack. Novel antibodies follow only if Gödel Sentences can be formed. From get go genomic intelligence in multi-cellular life evolved ‘to think outside the box’ with astronomic capacity of 1020 and upwards. Such powers developed millennia later in the human brain that make us greatly empathic (as the other is a projection of self in the mirror neurons), phenomenally protean and of course Machiavellian, as our intelligence co-evolved with viral software. The question as to why genomic code has not been hacked to bits over 5 billion years though it has been beset by bio-malware is because it is organized like a block chain distributed ledger where Gödel Sentences can alert the system of any malware activity that can subvert previous blocks or add novel blocks that are inconsistent with the former. All this has far reaching implications for our understanding of biology, the digital world and AI.
If these issues are of interest to you, Sheri is organizing a Webinar on 5 Nov 2021 for Frontiers AI and Robotics:
Narrow and General Intelligence: Embodied, Self-Referential Social Cognition and Novelty Production in Humans, AI and Robots
Professor or Economics, University of Essex
Sheri Markose received a PhD in Economics from the LSE in 1987. She joined the department in September 1986 after a position as research fellow (1982-1986) at the London Business School Centre For Macro-economic Forecasting. Her research interests, in applied economics, are in financial market modelling under extreme non-Gaussian events, computational mechanism design which uses agent based computational economic (ACE) models to 'wind tunnel' test proposed market protocols, electronic payments and cashlessness, interbank settlement systems, financial contagion and systemic risk. Her longstanding research interest and contributions to the Godelian formal mathematics of incompleteness and non-computability has enabled her to develop a theory of markets as complex adaptive systems and Nash equilibria in which strategic innovation and surprises occur. She was the lead researcher on the Foresight Office of Science and Technology 2006 IIS project on designing Smart Market Protocols for Road Transport Congestion. From 2006-2010, she directed research at Essex as part of the EC FP6 Euro 4 million RTN on the Computational Optimization Methods in Statistics, Econometrics and Finance (COMISEF) project which led to the development of multi-agent financial network models for systemic risk modelling.
Sheri was Director of the Institute of Studies in Finance from 2000 and then became the founder Director since 2002 of the Centre for Computational Finance and Economic Agents (CCFEA) where she has pioneered postgraduate research and teaching in agent-based computational economics (ACE) and markets as complex adaptive system. Starting in 2013 October within the Economics Department, Sheri has designed a new post Great Financial Crisis MSc Computational Economics, Financial Markets and Policy, which will train students in cutting edge skill sets such as financial network and systemic risk modelling, computational stress test platforms for robust macro and micro policy design and real time financial markets.
From February 2011-2015, Sheri has been appointed as senior consultant to the Reserve Bank of India Financial Stability Division to help establish ICT based financial network oriented modelling platforms for financial stability analysis. She was an academic advisor (Feb- Aug 2013) for BIS/BCBS OTC Derivatives Reform Report. In the 2017 Banque de France Financial Stability Review, Sheri and co-authors have provided an updated assessment of systemic risk from global derivatives markets following the 2009 G20 OTC derivatives reforms. Sheri was awarded the 2017 Eubank Prize by the Rice University, USA, “For integrative synthesis and data driven leadership toward understanding systemic risk in global financial markets.” Sheri has been a member of the European Science Foundation Review Panel in the area of socio-economic risks since 2014. Her 2017 publication in the American Institute of Mathematical Sciences Journal of Dynamics and Games on How Can Digital Agents Innovate?, highlighting the role of the Gödel sentence that enables a code to self-report that it is under attack to produce Type 4 dynamics in the Wolfram-Chomsky schema. This has been called ‘exciting’ by Noam Chomsky. In 2017, Sheri was invited to become an Associate Editor of Frontiers in Robotics and AI: Computational Intelligence https://www.frontiersin.org/journals/robotics-and-ai/sections/computational-intelligence#editorial-board
Sheri has been invited to give keynote talks On the Digital Foundations of Intelligence at 2019 Bio Inspired ICT (BICT) Conference at Carnegie Mellon University and again in 2019 Dec at the C3 Complexity Symposium of Sydney University.