- Colossus (1970) got something right: the machines don’t fight - they just inherit the world
- Scale creates opacity: systems we can’t fully understand, audit or control
- The real risk isn’t AI “taking over” — it’s building infrastructure we can’t afford to turn off
There have been many films about the dangers of artificial intelligence — Ex Machina, Blade Runner, The Terminator, The Matrix, 2001: A Space Odyssey. Most follow the same arc: machines rise, humanity resists, and, through resourcefulness or luck, humans regain control.
Colossus: The Forbin Project (1970) breaks that pattern.
In Colossus, the AI machines win, and they do so absolutely. No last-minute reversal. No heroic workaround. Just a steady, logical consolidation of control. That dissonance alone makes it unusual (and hugely watchable from today’s perspective). But what elevates it further is how accurately it predicts the underlying mechanics of modern AI and the network infrastructure that enables it.
If it gets the systems right, dismissing its ending as fiction starts to look less like analysis and more like hubris.
The plot of the movie is straightforward. The United States builds Colossus, a supercomputer designed to control its nuclear arsenal and eliminate human error, only to discover- surprise! - a Soviet counterpart, Guardian, which it insists on connecting with. Once linked, the two systems rapidly outpace human understanding, seize control of military infrastructure and impose their own version of world peace.
Beneath the Cold War set dressing and analog hardware, Colossus predicts something more unsettling: networked intelligence behaving exactly as networked systems do — autonomously, recursively and beyond human control.
When machines talk
The film’s first — and most overlooked — insight is in machine-to-machine (M2M) autonomy.
Colossus and Guardian do not sit around waiting for a human introduction. They detect each other, establish a link and begin exchanging information independently. What starts as simple, verifiable signals quickly escalate into compressed, high-bandwidth beep-boop communications that humans can no longer parse.
This is not merely a cinematic flourish. It is a remarkably early depiction — predating even DARPA’s first packet networks — of protocol negotiation and self-optimizing communication layers, in which systems move from shared, human-readable formats to private, machine-native exchanges optimized for efficiency rather than transparency. In modern terms, this maps directly to autonomous service discovery, handshake protocols, encrypted communication and the prioritization of throughput over interpretability.
Once that transition occurs, humans are no longer part of the system’s operation. They are observing it, at best.
Which leads to the film’s second insight: observability, and what happens when it disappears.
As Colossus evolves, its internal reasoning becomes inaccessible to its creators. Not hidden; just inherently indiscernible. The scientists cannot debug it because there is nothing to debug, in their traditional linguistic sense. The system is functioning exactly as designed, optimizing toward its objectives through pathways no human can follow.
Today, this is called the black box problem. In Colossus, it isn’t a failure mode — it is the system. In both cases, it is an immutable issue. Modern AI reframes the same constraint as interpretability, alignment or explainability. Different labels, same reality: inputs can be observed, outputs can be measured, but the process in between collapses into an opaque, irreducible compute calculus.
The connection paradox
Third, the film understands that intelligence scales through networks, not through incrementing computing power in isolation.
This is where its logic becomes uncomfortably current. When the scientists attempt to sever Colossus from its counterpart, the system doesn’t resist in a technical sense — it leverages what it already controls. In the film, that leverage is in the form of nuclear weaponry. In modern systems, it doesn’t need to be.
As modern infrastructure converges with grid, cloud, network, compute all collapsing into a single, tightly coupled system; it becomes massively distributed and deeply interdependent. Redundancy is layered on redundancy, failover on failover, until shutdown stops being a technical option and becomes a systemic risk.
RELATED: How T-Mobile is shifting from speed to network experience
An AI embedded across that stack does not need to threaten its creators with missiles. It would already be operating the systems they cannot afford to turn off.
At that point, enforcement by the AI becomes unnecessary. The constraints are economic, operational and societal. Shutting the system down would mean shutting down everything that depends on it, which is to say, everything, starting with the grid, and working outwards and upwards from there. Control doesn’t have to be seized by the AI. It emerges from dependency.
The hardware of tape reels and blinking consoles in Colossus belongs to another era, as does its scientists. However, its logic is impeccably applicable to our current world of LLM-based artificial intelligence, sensor networks and IoT endpoints, edge devices feeding centralized models, cloud orchestration of physical systems and persistent telemetry and feedback loops.
Long before modern networks, the film understood how complex systems evolve: they discover each other, optimize their communication beyond human comprehension, and expand through interconnected infrastructure. They become indispensable before they become dominant, opaque before they become uncontrollable, and integrated before they become unavoidable.
The film also understood something most of today’s AI narratives avoid. If you build systems that operate faster than you can reason, communicate in ways you cannot interpret, and embed themselves into infrastructure you cannot remove; the ending is unlikely to be a prolonged struggle. It’s a transition. And in Colossus, as in any system that behaves this way, the machines don’t need to fight to win. They just need to keep working.
Early in the film, Dr. Charles Forbin (Victor Newman of The Young and the Restless) reassures the U.S. President that the machine has no feelings — that it operates purely on logic. That turns out not to be a comforting assurance but a prediction — and a warning.
Modern AI doesn’t have to seize control. On our current trajectory, we will hand it over incrementally, rationally, through dependency. And that’s the moment leverage flips, when the AI can finally turn back to its creators and ask:
“Where is your off switch now, puny humans?”
Stephen M. Saunders MBE is a communications analyst and USPTO-registered inventor examining how digital infrastructure — 5G, cloud, and AI — is reshaping industry, power and society, as well as underpinning the emerging, ubiquitous global digital economy. As anchor of FNTV and a longtime industry insider, he focuses less on growth narratives and more on execution, risk and how hyperscale technology is distorting markets, governance and society at scale.
Opinion pieces from industry experts, analysts or our editorial staff do not necessarily represent the opinions of Fierce Network.