This item was submitted to Loughborough's Research Repository by the author. Items in Figshare are protected by copyright, with all rights reserved, unless otherwise indicated. The paper presents a vision of a future artificial intelligence (AI) system where multiple agents can learn independently over a lifetime and share knowledge with each other. The synergy between lifelong learning and knowledge sharing has the potential to create a society of AI systems, where each agent contributes to and benefits from collective knowledge. Key aspects include the ability to learn multiple skills incrementally, exchange knowledge via a common language, use local data and communication for learning, and rely on edge devices for decentralized computation and data. The result is a network of agents that can quickly respond to and learn new tasks, collectively hold more knowledge than a single agent, and extend knowledge in more diverse ways. Open research questions include when and what knowledge should be shared to maximize learning performance. The paper reviews recent machine learning advances converging towards creating a collective machine-learned intelligence. It argues that the convergence of scientific and technological advances will lead to the emergence of new types of scalable, resilient, and sustainable AI systems. The paper discusses the integration of fields such as lifelong learning, federated learning, distributed systems, and edge computing to achieve collective AI. It highlights challenges such as knowledge sharing, communication, and hardware constraints, and discusses opportunities for future research. The paper also explores application areas such as multi-agent active sensing, space exploration, responsive medicine, and distributed cyber-security systems. The study concludes that collective AI has the potential to significantly improve scalability, efficiency, and adaptability in AI systems.This item was submitted to Loughborough's Research Repository by the author. Items in Figshare are protected by copyright, with all rights reserved, unless otherwise indicated. The paper presents a vision of a future artificial intelligence (AI) system where multiple agents can learn independently over a lifetime and share knowledge with each other. The synergy between lifelong learning and knowledge sharing has the potential to create a society of AI systems, where each agent contributes to and benefits from collective knowledge. Key aspects include the ability to learn multiple skills incrementally, exchange knowledge via a common language, use local data and communication for learning, and rely on edge devices for decentralized computation and data. The result is a network of agents that can quickly respond to and learn new tasks, collectively hold more knowledge than a single agent, and extend knowledge in more diverse ways. Open research questions include when and what knowledge should be shared to maximize learning performance. The paper reviews recent machine learning advances converging towards creating a collective machine-learned intelligence. It argues that the convergence of scientific and technological advances will lead to the emergence of new types of scalable, resilient, and sustainable AI systems. The paper discusses the integration of fields such as lifelong learning, federated learning, distributed systems, and edge computing to achieve collective AI. It highlights challenges such as knowledge sharing, communication, and hardware constraints, and discusses opportunities for future research. The paper also explores application areas such as multi-agent active sensing, space exploration, responsive medicine, and distributed cyber-security systems. The study concludes that collective AI has the potential to significantly improve scalability, efficiency, and adaptability in AI systems.