Research

The science behind Genetic AI

Our methods are published, not hidden. Every claim we make about ranking, matching and multi-objective optimization traces back to work you can read and cite.

Research

Research behind Genetic AI

From evolutionary data theory to feature weighting and ab initio multi-objective optimization — our latest work on how Genetic AI learns to rank, match and surprise.

arXiv preprint — cs.NE · 2026

Evolutionary Data Theory: On the Similarities between Data Problems and Evolutionary Games

P. Wissgott

Applying the concepts and formalism of Evolutionary Game Theory to the data regime, this paper introduces the fundamental paradigms of Evolutionary Data Theory. Replicator equations, evolutionary strategies, the Bishop-Cannings theorem and the analogy to Lotka-Volterra systems are mapped to the data interpretation, with input data understood as genes and organisms competing in an evolutionary game.

arXiv preprint — math.OC · 2026

Feature weighting for data analysis via evolutionary simulation

A. Daniilidis, A. Domínguez Corella, P. Wissgott

An algorithm for assigning weights prior to scalarization in discrete multi-objective problems arising from data analysis. Weights — interpreted as feature relevance — evolve by a replicator-type dynamic on the standard simplex, and the resulting sequence is proven to converge globally to a unique interior equilibrium with non-degenerate limiting weights.

arXiv preprint — cs.NE · 2025

Genetic AI: Evolutionary Games for ab initio dynamic Multi-Objective Optimization

P. Wissgott

Genetic AI is a method for multi-objective optimization without external parameters, predefined weights or training data. Input data is converted into genes and organisms which compete for fitness in a simulation from first principles, governed by the Dominant, Altruistic, Balanced and Selfish evolutionary strategies.

Essay — danube.ai · 2025

What if Darwin made an AI?

P. Wissgott

While the world follows the hype of ever larger language models, this essay argues for a different direction: an AI built on the mechanics of evolution rather than the statistics of language. It contrasts the averaging, training-data-dependent nature of LLMs — with their bias, hallucinations and enormous energy footprint — against a lean, evolution-inspired approach that works from first principles.

Whitepaper v01 — danube.ai · 2024

Genetic AI: Applying evolutionary strategies to create a new type of artificial intelligence

P. Wissgott

The foundational whitepaper on Genetic AI. Inspired by genetic algorithms and evolutionary game theory, it develops a framework in which data features become genes and data records become organisms that interact under specific evolutionary strategies. Fast evolutionary simulations make real-time, decentralizable model training possible, with applications across data analysis and AI.

danube.ai
arXiv preprint — cs.NE · 2026

Evolutionary Data Theory: On the Similarities between Data Problems and Evolutionary Games

P. Wissgott

Applying the concepts and formalism of Evolutionary Game Theory to the data regime, this paper introduces the fundamental paradigms of Evolutionary Data Theory. Replicator equations, evolutionary strategies, the Bishop-Cannings theorem and the analogy to Lotka-Volterra systems are mapped to the data interpretation, with input data understood as genes and organisms competing in an evolutionary game.

arXiv preprint — math.OC · 2026

Feature weighting for data analysis via evolutionary simulation

A. Daniilidis, A. Domínguez Corella, P. Wissgott

An algorithm for assigning weights prior to scalarization in discrete multi-objective problems arising from data analysis. Weights — interpreted as feature relevance — evolve by a replicator-type dynamic on the standard simplex, and the resulting sequence is proven to converge globally to a unique interior equilibrium with non-degenerate limiting weights.

arXiv preprint — cs.NE · 2025

Genetic AI: Evolutionary Games for ab initio dynamic Multi-Objective Optimization

P. Wissgott

Genetic AI is a method for multi-objective optimization without external parameters, predefined weights or training data. Input data is converted into genes and organisms which compete for fitness in a simulation from first principles, governed by the Dominant, Altruistic, Balanced and Selfish evolutionary strategies.

Essay — danube.ai · 2025

What if Darwin made an AI?

P. Wissgott

While the world follows the hype of ever larger language models, this essay argues for a different direction: an AI built on the mechanics of evolution rather than the statistics of language. It contrasts the averaging, training-data-dependent nature of LLMs — with their bias, hallucinations and enormous energy footprint — against a lean, evolution-inspired approach that works from first principles.

Whitepaper v01 — danube.ai · 2024

Genetic AI: Applying evolutionary strategies to create a new type of artificial intelligence

P. Wissgott

The foundational whitepaper on Genetic AI. Inspired by genetic algorithms and evolutionary game theory, it develops a framework in which data features become genes and data records become organisms that interact under specific evolutionary strategies. Fast evolutionary simulations make real-time, decentralizable model training possible, with applications across data analysis and AI.

danube.ai

All publications

The complete list, newest first.

  1. arXiv preprint — math.OC

    Feature weighting for data analysis via evolutionary simulation

    A. Daniilidis, A. Domínguez Corella, P. Wissgott

    An algorithm for assigning weights prior to scalarization in discrete multi-objective problems arising from data analysis. Weights — interpreted as feature relevance — evolve by a replicator-type dynamic on the standard simplex, and the resulting sequence is proven to converge globally to a unique interior equilibrium with non-degenerate limiting weights.

  2. arXiv preprint — cs.NE

    Evolutionary Data Theory: On the Similarities between Data Problems and Evolutionary Games

    P. Wissgott

    Applying the concepts and formalism of Evolutionary Game Theory to the data regime, this paper introduces the fundamental paradigms of Evolutionary Data Theory. Replicator equations, evolutionary strategies, the Bishop-Cannings theorem and the analogy to Lotka-Volterra systems are mapped to the data interpretation, with input data understood as genes and organisms competing in an evolutionary game.

  3. arXiv preprint — cs.NE

    Genetic AI: Evolutionary Games for ab initio dynamic Multi-Objective Optimization

    P. Wissgott

    Genetic AI is a method for multi-objective optimization without external parameters, predefined weights or training data. Input data is converted into genes and organisms which compete for fitness in a simulation from first principles, governed by the Dominant, Altruistic, Balanced and Selfish evolutionary strategies.

  4. Essay — danube.ai

    What if Darwin made an AI?

    P. Wissgott

    While the world follows the hype of ever larger language models, this essay argues for a different direction: an AI built on the mechanics of evolution rather than the statistics of language. It contrasts the averaging, training-data-dependent nature of LLMs — with their bias, hallucinations and enormous energy footprint — against a lean, evolution-inspired approach that works from first principles.

  5. Whitepaper v01 — danube.ai

    Genetic AI: Applying evolutionary strategies to create a new type of artificial intelligence

    P. Wissgott

    The foundational whitepaper on Genetic AI. Inspired by genetic algorithms and evolutionary game theory, it develops a framework in which data features become genes and data records become organisms that interact under specific evolutionary strategies. Fast evolutionary simulations make real-time, decentralizable model training possible, with applications across data analysis and AI.

Try Genetic AI on your own data

With DarwinNGC you can already upload your own data and run Genetic AI on it. Ready to try it for yourself? - or get in touch directly.