This special session aims to provide a venue for the ALife and AI communities to explore how self-organisation can inspire and enrich neural paradigms of intelligence. Submissions to the special session follow the same review process as the main-track, and are equally included in the conference proceedings.

The recent 2024 physics nobel prize to John Hopfield for the development of Hopfield networks (Hopfield, 1982) highlights the ongoing relevance of core Artificial Life notions such as self-organisation and emergence for the neural computing paradigm that enables modern AI.

The neural computing paradigm—in its most popular deep learning form—consists of learning networks made of simplified neuron-like units. Despite its simplicity, it has proven to be highly successful at solving complex tasks, and has started to exhibit early signs of emergent intelligence (Wei et al, 2021). On the other hand, self-organisation is essential to how biological neural systems come to be (Hiesinger, 2021), and has been one of the cornerstones of ALife since Ross Ashby work on self-organisation as the basis for adaptive behaviour in neural systems (Ashby, 1952), and von Neumann’s work on universal constructors. Recently, self-organisation is emerging as a promising paradigm in the AI field (Mordvintsev et al., 2020; Ha and Tang, 2021; Risi, 2021, Variengien et al., 2021), bringing ideas from complex systems and ALife onto the neural computing paradigm.

This special session aims at providing a venue for the ALife and AI communities to explore synergies around how self-organisation can produce and enrich neural paradigms of intelligence.

Some of the questions we seek to address in this special session:

• What benefits does self-organisation bring to the neural computing paradigm, and how does it help overcome its limitations?

• How do dynamics, computation, and representation work together to produce neural intelligence? Are some neural substrates better suited for self-organisation? E.g. spike vs rate, synchronous vs. asynchronous activity, critical vs non-critical regimes.

• Can learning be understood as a self-organising process? What role does self-organisation play in lifelong learning and adaptability?

• How can deep reinforcement learning be unified with self-organisation, where rewards emerge spontaneously across multiple levels? How do self-organising systems solve the credit assignment problem? Can self-organisation drive the development of autonomous goal-setting in neural systems?

• How does evolution interact with self-organisation in shaping neural intelligence? Can evolutionary principles facilitate the emergence of self-organising neural structures? 

• Can bio-inspired self-organising principles enhance robustness and fault tolerance in neural systems? 

• What are the challenges and trade-offs of integrating self-organisation into large-scale neural architectures?

• Beyond GPU: Is custom hardware (neuromorphic, FPGAs, etc.) necessary to demonstrate the potential of self-organising neural models?

References:

• Design for a Brain: The origin of adaptive behaviour, Ashby, 1952.

• Neural networks and physical systems with emergent collective computational abilities. Hopfield, 1982.

•  Growing neural cellular automata. Mordvintsev, A., Randazzo, E., Niklasson, E., & Levin, M. 2020.

• Towards self-organized control: Using neural cellular automata to robustly control a cart-pole agent, A Variengien, S Nichele, T Glover, S Pontes-Filho, 2021.

• Collective Intelligence for Deep Learning: A Survey of Recent Developments. David Ha & Yujin Tang. 2022.

• Emergent Abilities of Large Language Models. Wei et al. 2021.

• The Future of Artificial Intelligence is Self-Organizing and Self-Assembling. S. Risi. 2021.

• The Self-Assembling Brain: How Neural Networks Grow Smarter. P. R. Hiesinger. 2021.