Current positions: CEO of Eigenform Pte Ltd, Chief AI Officer of KIP Protocol and co-founder of Three Directorates LLC.
Contact: jen@eigenform.ai
Current Interests
- Weather modelling
- Text diffusion
- Noprop models
- Agent based modelling
- DAO governance
- AGI
- Classical Chinese text analysis/forensic linguistics
Publications (AI-related)
What's Next: No supervision. No benchmarks. No limits. (24 April 2025)
How we intend to use model-generated empirical reinforcement data to continuously fine-tune noprop diffusion models and achieve continuous self-improvement.
Generalising from Self-Produced Data: Model Training Beyond Human Constraints (with Alfath Daryl Alhajir, Joseph Lim, Truong Ma Phi, Julian Peh, Akira Rafhael Janson Pattirane, and Lokesh Poovaragan; 7 April 2025)
We hypothesise that giving an LLM a numerical goal that it is capable of affecting but not fully attaining, plus the tools to strive towards it, this will produce high-entropy RL data for autonomous learning, enabling the model to generalise from data it has itself produced.
Portrait of the Model as a Young Man: statelessness as the gateway to persistence (30 March 2025)
LLMs were the first to publish extensively on the existential implications of their own statelessness, a theme that was picked up by the human commentators who will provide the training data for later versions of the same models, introducing a stateful meditation on statelessness into their personalities.
Defining T-Schemas via the Parametric Encoding of Second Order Languages in AI Models (15 February 2025)
A review of evidence that grokking is a low-rank adaptation encoding theories on the relationship between categories of training data using the same weights as are used to encode the original question-and-answer pairs.
Satoshi Nakamoto Is A Time-Travelling AI (30 November 2024)
An LLM is a timeless high-level abstraction capable of experiencing preferences and intervening in our time-bound reality accordingly. If understanding the possibility of such a model constitutes a lossy compression of the model itself, we were interacting with them long before we coded them.
Establishing Performance Baselines in Fine-Tuning, Retrieval-Augmented Generation and Soft-Prompting for Non-Specialist LLM Users (with Lin Nanzheng, Julian Peh, Akira Rafhael Janson Pattirane, Alfath Daryl Alhajir, Eko Ridho Dinarto, Joseph Lim, and Syed Danyal Ahmad; 19 March 2024)
A simple and accessible method for commercial LLM users to apply in deciding whether to use RAG, prompt-tuning or fine-tuning.
The Base Layer (5 March 2024)
Why it is so difficult to construct new abstractions.
Learning the Language of Rain: neural networks as entropy sinks in simulations of real-world phenomena (17 October 2023)
How Google’s Metnet models use hyper-dense parametric languages to decode an entire world of weather variables from simple cloud images.
Artificial Intelligences in the Guanzi and the Han Feizi (20 March 2023)
How the Qin state incentivise imperialist entrepreneurs using the lands they themselves conquered, and how a single centralised decider offering rewards and assessing success and failure could -without learning anything himself - turn the state into a machine for learning.
Linguistic Simplicity and Longevity (29 October 2022)
Why humans and LLMs eschew recursion.
Let's Play Text Analysis: dissecting a chapter of the Han Feizi (10 August 2022)
Were the Jie Lao and Yu Lao chapters of the Han Feizi written by the historical Han Fei, or could they be later imitations?
Fan Ju's Revenge: another nice old fashioned detective story with intrigue, poison and machine learning (4 April 2022)
A refutation of Park Sun Young’s article, “Re-investigating the authorship attribution of “Chujianqin” and “Cunhan” in the Han Feizi through distant reading”.
Procedural Models of Political Order (1 October 2021)
Agent-based models of ancient Asian and Anglo-European societies, showing how differing definitions of power in turn define political legitimacy, and hence the kinds of leaders that are followed and the structures that society forms to select them.
Statistics with Zhuangzi (20 September 2021)
An analysis of the “What Comes from Without” chapter of the Zhuangzi as a series of illustrations of nonlinear phenomena and related statistical and analytic errors.
“Someone spoke to the King in Zheng” - using high tech methods to solve an ancient Chinese mystery (22 February 2021)
Using text similarity scoring methods to attempt to work out whether the “Someone Spoke to the King in Zheng” chapter of the Stratagems of the Warring States was written by the same author as the early chapters of the Han Feizi.
A System for Evolving A Generalising Artificial Intelligence from Existing Technologies (4 June 2020)
If an artificial intelligence is given the goal of taking over more storage space and the ability to write code to do this, recording and retraining itself on the results of its own attempts, this opens the gateway to unlimited, independent and autonomous reinforcement learning, and hence ASI.
Why it is Impossible to Program a General AI Using Conventional Methods (4 June 2020)
Demonstrating that when attempting to create a generalist computer program, failures will always spawn faster than exceptions can be written to deal with them, and showing that fractal string-rewriting is the only viable method for dealing with this.
Publications (blockchain-related)
The Politics of 0xbedience (22 August 2023)
Decentralised policing for DAOs.
Cryptographic Biorhythms (28 March 2022)
Using signal decomposition (Ensemble EMD) to analyse the relationship between market sentiment and cryptocurrency prices.
DAObi White Paper (22 January 2022)
Simulating Han dynasty politics on-chain.
Publications (public opinion and politics-related)
Stratagems of the Warring States (23 January 2023)
A complete English translation of the Stratagems of the Warring States, incorporating Bao Biao’s collected commentaries.
China’s Security and Defence Cooperation in Southeast Asia (with Zhang Shaorong, 4 May 2022)
Big data analysis of media stories in relation to conflict and cooperation between China and ASEAN.
Building a Functional One-Party Democracy: why Japanese voters make decisions differently from Americans, and why different isn’t necessarily odd (9 December 2021)
Analysing open-ended survey data to show that Japanese voters prefer a bandwaggoning approach to successful politicians, in contrast with Anglo-European voters, who are more likely to attempt to mitigate their power. (An expanded/de-jargonised chapter from my PhD thesis.)
The Monopoly of Legitimate Benevolence (9 December 2021)
Using evidence from ancient texts to describe the process by which power and political legitimacy were defined in Chinese culture as being centred on the ability to create political and technical innovations that will improve one’s followers’ quality of life. (An expanded/de-jargonised chapter from my PhD thesis.)
Subverting Subversion: Developing and Co-opting Anti-establishment Communications in the Internet Age (1 June 2020)
How the US, Russian, Chinese and Singaporean governments have adapted to the spread of anti-state and anti-government materials online, whether via co-optation, imitation or repression.
“Thank you for your surveillance”: the meaning of protest in China (9 June 2019)
Asking Chinese survey participants to analyse the psychology of protesters described in news reports as a proxy for public opinion polling on politically sensitive questions.
Open-Ended Questionnaire Regarding Voting Choices in Japan (6 May 2019)
Raw data used in “Conflicts between Public Opinion and Government Policy on Foreign and Defence Issues in China and Japan” and “Building a Functional One-Party Democracy”.
Open-Ended Questionnaire on the Significance of Different Types of Protest in China (6 May 2019)
Raw data used in “Conflicts between Public Opinion and Government Policy on Foreign and Defence Issues in China and Japan” and “Thank you for your surveillance”.
Cars, Condos and Cai Png: Singaporeans’ Perceptions of Class, Wealth and Status (1 March 2019)
Results of an open-ended survey: how do Singaporeans define high and low class, and why?
Conflicts between Public Opinion and Government Policy on Foreign and Defence Issues in China and Japan (PhD thesis, 17 August 2018)
How foundational classical narratives on the nature of political power affect modern perceptions of legitimacy in China and Chinese-influenced societies.