AI Perception of Time Goes Beyond Human Limits

An understanding of the passage of time is fundamental to human consciousness. While we continue to debate whether artificial intelligence (AI) can possess consciousness, one thing is certain: AI will experience time differently. Its sense of time will be dictated not by biology, but by its computational, sensory, and communication processes. How will we coexist with an alien intelligence that perceives and acts in a very different temporal world?

What Simultaneity Means to a Human

Clap your hands while looking at them. You see, hear, and feel the clap as a single multimodal event—the visual, audio, and tactile senses appear simultaneous and define the “now.” Our consciousness plays out these sensory inputs as simultaneous, although they arrive at different times: Light reaches our eyes faster than sound reaches our ears, while our brain processes audio faster than it does complex visual information. Still, it all feels like one moment.

That illusion stems from a built-in brain mechanism. The brain defines “now” through a brief window of time during which multiple sensory perceptions are collected and integrated. This span of time, usually up to few hundreds of milliseconds, is called the temporal window of integration (TWI). As a proxy for this temporal grid, films with 24 frames per second create an illusion of continuous movement.

But the human TWI has its limits. See a distant lightning flash and you’ll hear the rumble of thunder seconds later. The human TWI evolved to stitch together sensory information only for events within roughly 10 to 15 meters. That’s our horizon of simultaneity.

Alien Intelligence in the Physical World

AI is poised to become a standard part of robots and other machines that perceive and interact with the physical world. These machines will use sensors hardwired to their bodies, but also remote sensors that send digital data from afar. A robot may receive data from a satellite orbiting 600 km above Earth and treat the data as real-time, as transmission takes only 2 ms—far faster than the human TWI.

A human’s sensors are “hardwired” to the body, which establishes two premises for how the brain interacts with the physical world. First, the propagation delay from each sensor to the brain is predictable. When a sound occurs in the environment, the unpredictable factor is the distance between the sound source and our ears; the time delay from the ears to the brain is fixed. Second, each sensor is used by only one human brain. The human horizon of simultaneity evolved through millions of years under these premises, optimized to help us assess opportunities and threats. A lion at 15 meters was worth worrying about, but thunder at 3 kilometers was likely not.

These two premises won’t always be valid for intelligent machines with multimodal perception. An AI system may receive data from a remote sensor with unpredictable link delays. And a single sensor can provide data to many different AI modules in real time, like an eye shared by multiple brains. As a result, AI systems will evolve their own perception of space and time and their own horizon of simultaneity, and they’ll change much faster than the glacial pace of human evolution. We will soon coexist with an alien intelligence that has a different perception of time and space.

The AI Time Advantage

Here’s where things get strange. AI systems are not limited by biological processing speeds and can perceive time with unprecedented precision, discovering cause-and-effect relationships that occur too quickly for human perception.

In our hyperconnected world, this could lead to wide-scale Rashomon effects, where multiple observers give conflicting perspectives on events. (The term comes from a classic Japanese film in which several characters describe the same incident in dramatically different ways, each shaped by their own perspective.)

Imagine a traffic accident in the year 2045 at a busy city intersection, witnessed by three observers: a human pedestrian, an AI system directly connected to street sensors, and a remote AI system receiving the same sensory data over a digital link. The human simply perceives a robot entering the road just before a car crashes into it. The local AI, with immediate sensor access, records the precise order: the robot moving first, then the car braking, then the collision. Meanwhile, the remote AI’s perception is skewed by communication delays, perhaps logging the braking before it perceives the robot stepping into the road. Each perspective offers a different sequence of cause and effect. Which witness will be considered credible, a human or a machine? And which machine?

People with malicious intent could even use high-powered AI systems to fabricate “events” using generative AI, and could insert them in the overall flow of events perceived by less capable machines. Humans equipped with extended-reality interfaces might be especially vulnerable to such manipulations, as they’d be continuously taking in digital sensory data.

If the sequence of events is distorted, it can disrupt our sense of causality, potentially disrupting time-critical systems such as emergency response, financial trading, or autonomous driving. People could even use AI systems capable of predicting events milliseconds before they occur to confuse and confound. If an AI system predicted an event and transmitted false data at precisely the right moment, it could create a false appearance of causality. For example, an AI that could predict movements of the stock market could publish a fabricated news alert just before an anticipated sell-off.

Computers Put Timestamps, Nature Does Not

The engineer’s instinct might be to solve the problem with digital timestamps on sensory data. However, timestamps require precise clock synchronization, which requires more power than many small devices can handle.

And even if sensory data is timestamped, communication or processing delays may cause it to arrive too late for an intelligent machine to act on the data in real time. Imagine an industrial robot in a factory tasked with stopping a machine if a worker gets too close. Sensors detect a worker’s movement and a warning signal—including a timestamp—travels over the network. But there’s an unexpected network hiccup and the signal arrives after 200 milliseconds, so the robot acts too late to prevent an accident. The timestamps don’t make communication delays predictable, but they can help to reconstruct what went wrong after the fact.

Nature, of course, does not put timestamps on events. We infer temporal flow and causality by comparing the arrival times of event data and integrating it with the brain’s model of the world.

Albert Einstein’s special theory of relativity noted that simultaneity depends on the observer’s frame of reference and can vary with motion. However, it also showed that the causal order of events, the sequence in which causes lead to effects, remains consistent for all observers. Not so for intelligent machines. Because of unpredictable communication delays and variable processing times, intelligent machines may perceive events in a different causal order altogether.

In 1978, Leslie Lamport addressed this issue for distributed computing, introducing logical clocks to determine “happened before” relation among digital events. To adapt this approach to the intersection of the physical and digital worlds, we must grapple with unpredictable delays between a real-world event and its digital timestamp.

This crucial tunneling from the physical to the digital world happens at specific access points: a digital device or sensor, WiFi routers, satellites, and base stations. As individual devices or sensors can be hacked fairly easily, the responsibility for maintaining accurate and trustworthy information about time and causal order will fall increasingly on large digital infrastructure nodes.

This vision aligns with developments within 6G, the forthcoming wireless standard. In 6G, base stations will not only relay information, they will also sense their environments. These future base stations must become trustworthy gateways between the physical and the digital worlds. Developing such technologies could prove essential as we enter an unpredictable future shaped by rapidly evolving alien intelligences.