Not One Brain, But Many Connected Brains
Questions That Must Grow Together as Technology Expands
AI is no longer confined to computer screens. It is connecting with the physical real world, taking on the role of "Agent" — understanding situations, grasping intentions, and acting on behalf of people. This chapter examines three important flows in how AI meets reality.
AI Operating on Top of Reality''s Copies: Digital Twins
Digital twin technology replicates real objects, spaces, or systems in digital environments — like creating a twin world where factory machines, city traffic, or building structures are reflected in real time. Initially used for monitoring, the technology has evolved so AI operates within digital twins: analyzing vehicle flow and adjusting traffic signals, detecting factory equipment failures before they occur, optimizing logistics warehouse robot routes. This creates real-time bidirectional connections between physical and digital — sensors bring information from reality, AI makes judgments, and results return to cause physical actions.
Sim2Real training enables AI to learn strategies through rapid simulated trial-and-error before transferring to real environments. Combined with imitation learning, this enables robots to learn human tasks naturally. These technologies are accelerating the transition from fixed-route AGVs to self-navigating AMRs. Going forward, more cities, factories, hospitals, and homes will be replicated as digital twins, with AI learning, predicting, and making judgments within these replicas to influence the real world.
AI Operating from My Perspective: First-Person AI
Previous AI observed from outside — seeing only text input, not users'' actual environments. People understand situations comprehensively through expression, gesture, and atmosphere. Recent AI is gaining a first-person perspective through wearable devices: smart glasses, AR devices, and smart watches with cameras, microphones, and motion sensors transmit real-time information about what users see, hear, and do.
This enables Multimodal Processing Capability — simultaneously processing visual, audio, text, location, and biometric information. For example, smart glasses detecting that a user is looking at a product label can immediately display ingredient information; detecting extended focus on a math problem can proactively provide hints. AI transitions from a tool responding to commands to a proactive helper that understands what you are doing and assists before you ask.
Not One Brain, But Many Connected Brains: Distributed AI
Early AI systems processed everything within one device or server. The enormous increase in data types (photos, video, voice, movement, health data) and the need for immediate responses (autonomous vehicles where even 1-second delays can be fatal) demand distributed structures. Distributed AI has different devices taking different roles while operating as one system. For example, asking a smartphone for tomorrow''s outfit recommendation: voice recognition on the smartphone, weather checking on cloud servers, outfit recommendation on another server — yet the user sees it as one seamless AI response.
Distributed structures offer flexibility (no single point of failure), scalability (adding devices rather than upgrading one server), and energy efficiency (devices activate only when needed). Smart glasses walking navigation is another example: cameras in glasses capture surroundings, voice on smartphones, maps from cloud, all displayed on the glasses — multiple devices, one seamless experience.
Going forward distributed AI will be applied in smart cities, digital healthcare, disaster response, and space exploration — wherever resources and energy must be efficiently utilized.
Questions That Must Grow Together as Technology Expands
As AI spreads into digital twins, first-person wearables, and distributed environments, critical societal questions arise. First, energy and resources: super-large AI models and distributed devices are requiring ever more electricity and computation. Future AI must become more efficient — lightweight models, efficient algorithms, adaptive computation.
Second, information sovereignty: as AI accesses our field of vision, physical space, and body information, sensitive data like location, facial expressions, voice tone, and heart rate all become AI inputs. Simply strengthening security is insufficient — society as a whole must discuss and agree on what information AI obtains, how far it interprets, and how it can be utilized.
Ultimately the future of AI is not realized by technological development alone. As technology spreads deeper into our lives, we must think carefully about what criteria and balance must be maintained in living with AI. The direction going forward is not just faster and smarter technology, but technology thoughtfully and wisely permeating our lives and environments.
Ultimately, the question of how to develop AI is also the question of what kind of world we want to live in going forward.
