The marriage of IoT and AI is a transition that dwarfs the original Internet’s impact

Manish Tewari

A person awakens not to an alarm clock but to a wearable band that has monitored the entire sleep architecture — tracking REM (rapid eye movement) cycles, heart rate variability and nocturnal peripheral oxygen saturation levels. Through a cloud-hosted machine learning model, it has chosen the lightest sleep phase within a predetermined window to trigger a gentle haptic pulse.

The bedroom lights slowly shift from a dim amber to a cool white, mimicking the arc of dawn, as soon as a tiny passive infrared sensor in the corner registers the first stirring of motion under the blanket.

The refrigerator has already detected that the milk carton’s weight, measured by a strain gauge embedded in the shelf, has fallen below a threshold, and it generates a grocery reminder that appears on the mobile phone screen, routed through a cloud service maintained in a data centre in some other part of the world. The smart kettle, triggered by a proximity tag in the owner’s slippers, heats water for tea; an ultrasonic sensor ensures that the water does not boil dry. All this unfolds before a single conscious choice has been made.

Welcome to the Internet of Things (IoT) — a network of physical objects embedded with sensors, software and communication technologies, enabling them to connect to the Internet and exchange data automatically without direct human involvement.

The term “Internet of Things” was coined in 1999 by Kevin Ashton, a British technologist then working on supply-chain optimisation at Procter & Gamble, during a presentation linking Radio-Frequency Identification (RFID) tags to the Internet. The insight was deceptively simple: if everyday objects could sense their environment and communicate without human intermediation, inventory management would cease to be reactive and become self-aware.

Yet, the primordial object was even older: a hacked vending machine, wired to purchase food supplies through a computer terminal, report its inventory on its own, and chill. The machine, Prancing Pony, named after an inn in JRR Tolkien’s The Lord of the Rings, also sold beer but only on credit and only to people over 21, since it knew everyone’s age. If a youngster attempted to buy beer, it responded “Sorry, kid.”

From that quirky beginning, the field mutated through embedded systems, wireless sensor networks and machine-to-machine protocols. Every IoT device now embodies sensing, networking, processing and actuation. This convergence of operational technology with information technology dissolved the air gap that once insulated, say, a factory’s robotic arm from the Internet.

That dissolution represents the single most consequential architectural change in the history of digital infrastructure because it transforms every sensor into a potential entry point for malicious code, every actuator into a weapon that can manipulate the physical world.

Ashton’s original substrate contained the entire genetic code of IoT: physically embedded constraints, low-power silicon and a wireless link to a networked database. The real ignition for IoT came from the collision of three exponential curves: the plummeting cost of micro-electromechanical systems (MEMS) sensors, the planetary spread of IPv6 (Internet Protocol version 6) and the emergence of low-power WAN (Wide Area Network) protocols that whisper tiny packets of data across kilometres on a single coin-cell battery.

The current state of affairs is simultaneously a staggering engineering triumph and a policy quagmire. On the one hand, IoT has moved beyond wearables and smart speakers into the hard infrastructure of civilisation. In healthcare, for instance, continuous glucose monitors and implantable cardiac loop recorders that transmit medication adherence data have already transitioned to standard care protocols. These devices fuse the patient’s physiology with clinical decision-support algorithms, effectively turning every individual into their own digital twin.

On the other hand, the very ubiquity that makes IoT revolutionary also makes it terrifying. Unlike the Internet of human-to-human communication, where the worst-case breach often means stolen credit-card numbers or embarrassing email leaks, an IoT compromise is a kinetic threat. For example, insulin pumps can be remotely reprogrammed to deliver a lethal dose, or the CAN (Controller Area Network) bus of connected vehicles can be injected with malicious frames from the infotainment unit.

The marriage of IoT and AI is a transition that dwarfs the original Internet’s impact. The first Internet revolution was one of information: it indexed the world’s explicit knowledge and compressed the cost of sharing it to near zero. The IoT-plus-AI revolution is one of action — it indexes the world’s physical behaviour and compresses the time between signal and response to near zero.

IoT is a system of five layers: perception, network, middleware, application and business; but that model is static until AI becomes the brain across all layers. Edge-deployed TinyML models running on Arm Cortex-M microcontrollers fundamentally re-architect the Internet from a human-centric request-response system to a world where objects negotiate, learn and act long before a human ever joins the loop.

No case study exemplifies the breathtaking promise and the apocalyptic risk clearer than the saga of Claude Mythos, an AI tool capable of probing any system — desktop software, cloud infrastructure, embedded firmware — and rapidly surfacing zero-day vulnerabilities that a human reverse engineering expert might overlook for years.

Simply put, it could crash any machine it touched. The statement is not about mere software instability; it is about the fundamental brittleness of a world where AI, even with benign intent, can issue a physical command — such as opening a valve in a chemical plant or cutting power to a neonatal incubator — that cascades into a catastrophe because the AI’s mental model of the environment is imperfect.

This is the ultimate caution: in an IoT-saturated world, an AI agent is not just a disembodied text generator; it is a poltergeist with a wrench, a screwdriver and access to a billion actuators.

The future is one where entire cities will become soft-real-time control loops. Streetlights with LiDAR (Light Detection and Ranging) sensors will not only adjust brightness but also detect the micro-Doppler signatures of a pedestrian who has stumbled, instantly altering the signal timing of surrounding intersections while pinging the nearest autonomous ambulance. This is about turning the planet into a closed-loop metabolism — a form of geocybernetics — where every sensor-to-actuator chain is a potential attack vector.

The Internet gave humanity a new nervous system for information. The Internet of Things, once wedded to machine agency, gives humanity a new musculoskeletal system for the planet itself. The future beckons.