What will the future of artificial intelligence (AI) include? How can you get a comprehensive overview of the evolving landscape of AI? Friston et al.’s research paper “Designing Intelligence Ecosystems from First Principles” (2024) describes a forward-looking vision for the field of artificial intelligence (AI) over the next decade and beyond. This vision focuses on the development of a cyber-physical ecosystem comprised of natural and synthetic elements that collectively contribute to “shared intelligence.” This concept emphasizes the essential role of humans within these ecosystems. This paper highlights a specific approach to AI known as “active inference,” which is considered a physics-based approach to understanding and designing intelligent agents. This approach shares basic principles with quantum mechanics, classical mechanics, and statistical mechanics.
Active reasoning is applied to AI design, which suggests that next-generation AI systems will need to have explicit beliefs about the world while incorporating specific perspectives under generative models. This contrasts with traditional AI approaches such as reinforcement learning, which primarily focus on selecting actions to maximize reward. In active reasoning, exploration and curiosity are considered fundamental components of intelligence, leading to behaviors expected to reduce uncertainty.
The multi-scale architecture of active inference is another important aspect. It acknowledges different time scales of learning and model selection and operates in a similar way across overlapping time periods to maximize model evidence.In this context, intelligence is inherently perspectival and actively engages with the world based on a particular set of beliefs. It’s about participating.
Communication within these intelligent systems is also a key topic. The paper argues that intelligence at all scales requires shared generative models and a common foundation, which can be achieved through a variety of methods, such as ensemble learning, expert mixing, and Bayesian model averaging. In this context, an important aspect of active reasoning is selecting the message or perspective that provides the greatest expected information gain.
Finally, the paper addresses ethical considerations, emphasizing the importance of respecting and protecting individual values in the development of large-scale swarm intelligence systems. This approach contrasts with models such as eusocial insects, where individuals are largely interchangeable. The authors advocate for a cyber-physical network of emergent intelligence that respects the individuality of all participants, human or non-human.
In summary, Friston et al.’s paper presents a visionary approach to AI development centered on active reasoning and the creation of intelligent ecosystems that integrate and respect the individuality of human and non-human actors. This approach represents a significant paradigm shift in how AI is conceptualized and developed, with implications for the future of technology and society.
Image source: Shutterstock