The Surprising Creativity of Digital Evolution

By Joel Lehman, Jeff Clune, Dusan Misevic. 2018

62 citations recorded by the source corpus.

Natural evolution and complex adaptations often surprise scientists. However, the creativity of evolution is not limited to the natural world, transcending any particular substrate. In the context ...

Subfields: Artificial Intelligence, Experimental and Cognitive Psychology, Mechanical Engineering, Generative Models

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Jeff Clune
2. Kenneth O. Stanley
3. Joel Lehman
4. Jean-Baptiste Mouret
5. J. Urgen Schmidhuber
6. Sergey Levine
7. Risto Miikkulainen
8. Sebastian Risi

In Experimental and Cognitive Psychology

1. Kenneth O. Stanley
2. Jeff Clune
3. Kai Olav Ellefsen
4. Joel Lehman
5. John Schulman
6. Marc Schoenauer

In Mechanical Engineering

1. Jeff Clune
2. Jean-Baptiste Mouret
3. Kenneth O. Stanley
4. Joel Lehman
5. Risto Miikkulainen
6. Anh Totti Nguyen
7. Kai Olav Ellefsen
8. Antoine Cully

In Generative Models

1. Jeff Clune
2. Jean-Baptiste Mouret
3. Anh Totti Nguyen
4. Joel Lehman
5. Risto Miikkulainen
6. Frank Hutter
7. Kenneth O. Stanley
8. Heang-Ping Chan

Citations in the corpus

• Scaling Laws for Reward Model Overoptimization (2022, background): “Previous work has shown numerous examples of such behavior in a wide variety of settings [Krakovna et al., 2020, Lehman et al., 2020].”
• The Surprising Creativity of Digital Evolution: A Collection of Anecdotes from the Evolutionary Computation and Artificial Life Research Communities (2018, background): “Even in the absence of direct foresight, studies of evolvability suggest that genomic architecture itself can become biased toward increasing creative potential [62, 65, 66].”
• Explainable AI for medical imaging: deep-learning CNN ensemble for classification of estrogen receptor status from breast MRI (2020, background): “Across the artificial intelligence (AI) community, there are a plethora of anecdotes of AI models[3] learning unintended and creative solutions to tasks that rely on a performance metric guiding the model’s learning.”
• Interpretable Approximation of a Deep Reinforcement Learning Agent as a Set of If-Then Rules (2019, background): “A major component in any RL problem involves designing a valid reward function, it is essential to devise an appropriate reward function else the trained agent may result in undesired behavior [14].”