open-endedness
In plain wordsThe ability of a system to keep producing new and interesting things without end — as evolution and human culture do — instead of stopping once it solves one fixed task.
The property of a process or system that continually generates artifacts, behaviors or problems that are both novel and interesting, without converging on a fixed goal — as biological evolution and human science and culture appear to do. Research on it, rooted in artificial life and evolutionary computation, has produced algorithms that reward novelty or diversity rather than progress toward an objective (novelty search, quality-diversity), that generate ever-harder environments alongside the agents that solve them, and that keep archives of 'stepping stones' for later innovation. Edward Hughes and colleagues defined it relative to an observer: a system is open-ended if its outputs stay novel (hard for the observer to predict) yet learnable (more predictable once the observer has seen more of them). They argued that open-endedness built on foundation modelsfoundation model, n.: A very large AI model, built once at great cost from vast amounts of text, images or code gathered mostly from the internet, that many products then reuse as their starting point, from chatbots to coding tools. is essential for artificial superintelligenceartificial superintelligence, n.: Superintelligence in machine form: AI that would outthink the best humans at nearly every intellectual task. Often shortened to ASI and pictured as the stage after human-level AI. — a view behind self-improving systems such as the Darwin Gödel MachineDarwin Gödel Machine, prop. n.: An experimental AI coding assistant that improves itself by rewriting its own code, testing each new version and keeping a family tree of variants to build on. and AlphaEvolveAlphaEvolve, prop. n.: A Google DeepMind system that combines AI language models with an evolution-like search to invent and improve computer programs, finding new results in mathematics and computing. — while noting that its unpredictability raises safety concerns.
The archive kept every odd agent variant, because in open-endedness research today's dead end is tomorrow's stepping stone.
Origin
Open-ended evolution was a goal of artificial-life research from the 1990s. Joel Lehman and Kenneth Stanley's novelty search (2008) and their book Why Greatness Cannot Be Planned (2015) popularized objective-free search, and Jeff Clune's 2019 proposal of AI-generating algorithms tied it to general AI. Interest surged after a 2024 Google DeepMind position paper linking open-endedness to superhuman intelligence.
Related
- Darwin Gödel MachineDarwin Gödel Machine, prop. n.: An experimental AI coding assistant that improves itself by rewriting its own code, testing each new version and keeping a family tree of variants to build on.
- AlphaEvolveAlphaEvolve, prop. n.: A Google DeepMind system that combines AI language models with an evolution-like search to invent and improve computer programs, finding new results in mathematics and computing.
- genetic algorithmgenetic algorithm, n.: A problem-solving method that copies natural evolution: many candidate solutions compete, the best are 'bred' by mixing and randomly tweaking them, and over many generations the solutions improve.
- intrinsic motivationintrinsic motivation, n.: Giving an AI agent an inner drive — such as curiosity about surprising things — so it keeps exploring and learning even when the task itself offers no reward yet.
- artificial lifeartificial life, n.: The study of life by building lifelike systems — in computer simulations, robots or chemistry — to learn how living things organize themselves, reproduce and evolve, and what other kinds of life could exist.
- self-playself-play, n.: Training an AI by having it play against copies of itself, so it always has an opponent at its own level and can keep improving without human examples.
- artificial superintelligenceartificial superintelligence, n.: Superintelligence in machine form: AI that would outthink the best humans at nearly every intellectual task. Often shortened to ASI and pictured as the stage after human-level AI.
Sources
- Hughes, E., Dennis, M., Parker-Holder, J., Behbahani, F., Mavalankar, A., Shi, Y., Schaul, T. & Rocktäschel, T. (2024). Open-Endedness is Essential for Artificial Superhuman Intelligence. arXiv:2406.04268. arxiv.org/abs/2406.04268
- Stanley, K. O. & Lehman, J. (2015). Why Greatness Cannot Be Planned: The Myth of the Objective. Springer.
Mentioned in
AlphaEvolveAlphaEvolve, prop. n.: A Google DeepMind system that combines AI language models with an evolution-like search to invent and improve computer programs, finding new results in mathematics and computing., Darwin Gödel MachineDarwin Gödel Machine, prop. n.: An experimental AI coding assistant that improves itself by rewriting its own code, testing each new version and keeping a family tree of variants to build on.