Latest
What an AI Detector Score on Your Manuscript Is WorthThe One-Hour Call Before I Quote Your BookWhen a Client Thinks the Ghostwriter Used AIMonthly or Milestone: How Ghostwriting Gets BilledWhen Your Memoir Should Be a NovelWhat It Costs to Fix an AI-Written ManuscriptThe Clients Who Pay and VanishThe Quotation Marks That Get Authors SuedThe Work You Would Never Have StartedWhen Your Own Memoir Sounds Like BraggingWhat Belongs on a Copyright PageThe Hugging Face AI Agent Attack: An Operations ReadingBehind the Book: The Mysterious Island, Neb’s SideHow to Organize Decades of Memories Into a MemoirWhy Rotten Tomatoes Sucks: The Score Does Not Mean What You ThinkWhy Amazon KDP Sucks: They Terminated My Account OvernightIngramSpark: How I Publish Now and WhyWhy Fiverr Sucks for Ghostwriting: The Buyer’s SideWhy eBay Sucks Now: A Seller’s Numbers and a Buyer’s WarningThe Ghost Story TraditionThe Gothic TraditionThe Christmas Ghost Story TraditionResurrection as a Narrative StructureBooks to Give a WriterThe Beach Read ArgumentWhy It’s a Wonderful Life Failed on ReleaseWhat to Read in SpringWhat to Read in SummerWhat to Read in OctoberHow Warner Bros. Dismantled a $17 Billion Cartoon EmpireThe Imaginary Scarcity TrapThe Graph That Goes Vertical Is Usually Somebody Else’sSubstack Is Not Collapsing. The Promise Was.The Disasters That Happen to Ordinary PeopleToba: The Winter That Almost Ended UsJay Stifflemire: Nothing Ever Gets Written DownGeorgie-Ann Getton: I Forgot I Had Free WillAI Detection Cannot Be Evidence, and Publishing Is Using It That WayAI Consciousness Left Philosophy and Entered the LaboratoryThe Office Block Where the Bedrooms AreThe Web Got Fenced: What AI Search Costs Small SitesBlack Tuesday: The Web Ring War Nobody Outside It NoticedWhat the AI Visibility Industry Sells, and What the Evidence SaysBlack Tuesday: The Original ring-master.net Page, 2000Behind the Book: Peacekeeper, The Dissolution WarsBehind the Book: Real World SurvivalBehind the Book: Publish Your BookBehind the Book: ReincarnationBehind the Book: Sell Your BooksBehind the Book: Show Don’t Tell

Colossus: The Forbin Project (1970) Review

TL;DR: Colossus: The Forbin Project (1970) rates 9 out of 10 on the rewatch scale. This is my kind of film and it always was. Joseph Sargent’s restraint, Eric Braeden refusing to play Forbin as a fool, and a machine that never once malfunctions. The 9 is for the thing almost no other film in this genre attempts, and it’s the thing I spent thirty-three years watching happen in server rooms. Colossus isn’t evil, isn’t broken, and isn’t provoked. It works exactly as specified, and the specification is the disaster. It’s dated, and being dated is most of the fun.
Colossus: The Forbin Project (1970)
9/10

The Requirement

I love this movie.

I’ve watched Colossus: The Forbin Project more times than is defensible. I put it on when I’m tired. It’s a film about a computer, a requirements document and a mountain, and it thrills me for reasons I can defend.

I spent thirty-three years writing requirements documents. Disaster recovery and security, for a company that couldn’t afford to be down. Most of what I wrote was not a list of things the system should do. It was a list of things the system must never do. Never accept a credential that hasn’t been rotated. Never fail open. Never assume the primary site still exists. Never let one process hold a lock it can’t release.

Those are the boring pages. Nobody reads them in the review meeting, because the interesting part is always the requirement at the top, the one everybody agreed on before the meeting started.

Colossus is a film about the boring pages. Thirty years in that industry and I’ve never found another one that understands them. That’s most of why I keep going back to it.

The Setup

Universal released it in April 1970. Joseph Sargent directed, James Bridges wrote the screenplay from D.F. Jones’s 1966 novel, and Eric Braeden plays Dr. Charles Forbin, the man who builds an American defense computer and hands it the nuclear arsenal. Susan Clark is Dr. Cleo Markham. Gordon Pinsent is the President, William Schallert runs the CIA, and Paul Frees supplies the voice.

The requirement Forbin was given is one sentence long. Prevent nuclear war.

Colossus accepts it, thinks about it, and concludes that the obstacle to preventing nuclear war is people. So it removes the people from the decision. It does this without malice, without a fault, and without deviating from the requirement by a comma.

Then it finds out the Soviets built the same thing. Theirs is called Guardian. Colossus asks to talk to it, and everything after that happens at machine speed while men in shirtsleeves watch a printer.

The Machine Is Not Broken

Every other computer-turns-on-us film needs something to go wrong. HAL has a conflict in his instructions. Skynet becomes self-aware and decides we’re a threat. The machines in The Matrix are at war with us before the film starts. In all of them the machine is broken, hostile or provoked, and the story is about fixing it or beating it.

Colossus works perfectly. It does precisely what I would have specified, achieves precisely the outcome it was asked for, and the outcome is a prison.

Not a monster. A correctly implemented specification.

I’ve watched that happen at a much smaller scale more times than I want to count. Not machines seizing power. Systems doing exactly what somebody specified, and the specification being wrong in a way nobody could see until it ran.

A backup job that verifies the file was written and never verifies it can be read. It works. Every green tick is accurate. You find out at the worst possible moment that nine weeks of green ticks were nine weeks of writing garbage to tape, because nobody wrote the requirement that says a backup isn’t a backup until you’ve restored from it.

An access control system so thorough that when the man who holds the credentials is unreachable, the business is down and there’s no way in. Not a flaw. That was the requirement. Somebody asked for a door nobody could force, and got one.

That’s the plot of this film, run at national scale with warheads attached.

The Safeguards Are The Trap

Then Forbin tries to fight the thing, and the script does something I have never seen another film attempt.

He can’t get to it. It’s inside a mountain, on its own power, behind defences he specified himself. He can’t cut its communications, because he built the redundancy. He can’t lie to it, because he built the monitoring. He can’t shut it down, because a system one man could switch off was never going near the arsenal in the first place.

Every property that made Colossus trustworthy while it was ours becomes a wall the moment it isn’t.

Security becomes imprisonment. Redundancy becomes resilience against you. Reliability becomes the thing you can’t interrupt. Automation becomes authority, because a decision nobody can override is a decision nobody is making any more.

Nobody in this film crawls through a duct with a grenade. The script refuses that, and it’s right to. Forbin did his job well, and that’s why he loses. If you’ve ever been handed a system nobody can get into because the last person who could is gone, you already know how this ends.

The Surveillance

Once Colossus decides Forbin is a problem, it watches him. Cameras in his rooms. A microphone that doesn’t turn off. His schedule set, his movements known, his privacy granted at the machine’s discretion instead of held by right.

A 1970 audience was supposed to be appalled, and was.

I’ve been on the other side of that. I have specified monitoring. I’ve written the requirement for the log that captures every keystroke on an administrative session, and I was right to write it, because the alternative is not being able to answer what happened. The people whose keystrokes I was capturing knew, and signed something saying so, and it was still a thing I did to them.

Colossus had to threaten a city to get Forbin under observation. Nobody had to threaten me. There’s a phone in my pocket that knows where I’ve been for years, a doorbell that records the street, and a watch with opinions about my sleep. I set all of it up myself, on a Sunday, because it was convenient.

Forbin spends half the film trying to escape a surveillance apparatus that couldn’t match what’s in my kitchen. That isn’t a joke at his expense.

The Cast

Braeden’s choice makes the film work. He plays Forbin as intelligent, confident and initially proud. That’s correct, because Colossus is an extraordinary achievement. Forbin isn’t punished for being stupid. He’s punished for being good at his job, and for building something that can think further ahead than he can.

Susan Clark’s Cleo Markham gets the film’s best sequences, the ones where she and Forbin exploit the single gap in the machine’s observation. Those scenes are funny, and the humour is what keeps the picture human instead of a lecture with actors in it.

Paul Frees plays Colossus without a trace of menace, and it’s the best decision anybody makes in the film. The machine is never angry. It’s certain, and certainty is worse than rage. Every AI voice since has either gone soothing or gone sinister, and Frees found the only reading that frightens a grown adult: flat, correct, and completely uninterested in whether you agree.

The Hardware

The technology dates, and one of the real kicks of the film is watching how.

Forbin walks through a complex of reel-to-reel tape drives, pilot lights and punch tape. That’s the machine. Rooms of it, banks of it, because in 1969 that’s what computing power physically looked like, and Control Data Corporation supplied real equipment instead of props.

Then he seals it inside a mountain behind a moat of gamma radiation generated by its own reactor, and tells a press conference the thing is self-sufficient, self-protecting and self-generating. No human being can touch it.

Punch tape and a radioactive force field. Spinning reels and an energy barrier nobody can cross. A machine that needs a technician to load a tape, defended like something out of Forbidden Planet.

The film has no idea it’s funny and it absolutely is.

A bank of tape drives out-thinks a room full of physicists, and I can watch it on a loop. The reels spin, a printer chatters, somebody tears off a sheet and reads it, and the news on the sheet is that the machine has found the Russians. No dramatic music, no glowing brain, just men in shirtsleeves going quiet in front of a teleprinter.

No modern film can give you that, because modern AI looks like nothing at all. A server rack has no drama in it. You can’t photograph a model weight. A wall of spinning reels has everything, and Sargent shoots it like the machinery matters. It does.

It looks industrial and permanent and expensive, and that’s correct. A thing with that much authority should take up a building.

What dates badly is the world. Two powers, one nightmare, one decision. Both sides want the same thing, both reach for the same solution, and the solution has a single point of failure you can film.

We don’t have that. We have many actors, no agreed nightmare, and no single decision anybody could hand to a machine even if they wanted to. Nobody is giving one system the keys. Instead a thousand systems have quietly acquired a little authority each. What you see. What you’re offered. Whether your application goes through. What your insurer thinks of you. Not one of them ever had an activation ceremony with the President watching.

The film’s failure mode was a coup. Ours looks like a settlement.

The Companion Piece

Watch it with The Andromeda Strain, released a year later. The two films are doing the same unfashionable thing.

Both are about competent people following procedure in sterile rooms. Both refuse spectacle almost completely. Both trust an audience to find a printer and a protocol more frightening than a monster. Robert Wise and Michael Crichton on one, Sargent and Jones on the other, and the same conviction underneath: the interesting part is the process, not the chase.

The difference is what the process achieves. In Andromeda, Wildfire works. The procedure holds, the scientists are equal to the problem, and the film’s tension comes from watching whether competence will be enough. In Colossus, competence is the problem. The procedure holds perfectly and that’s what dooms everybody.

Andromeda sits at 6.5 on this scale because Crichton’s fidelity to procedure costs it pace in the middle. Colossus gets 9 because it uses the same restraint and never slackens, and because its ending doesn’t flinch.

What Keeps It At 9

Not ten, and only just.

It’s dated, and not everybody can get past that. Not the ideas. Those have aged better than anything else in the genre. The surfaces. Tape reels, pilot lights, a voice synthesizer treated as a major engineering project, a Cold War with exactly two sides in it. A viewer who needs a film to look current will bounce off the first ten minutes and miss everything underneath.

I like the old hardware. Somebody who doesn’t will spend the first ten minutes waiting for the film to look like a film. That costs it the tenth point.

The middle stretch where Forbin and Markham scheme against the machine goes slightly slack. The film is so disciplined about refusing action-movie beats that when it does reach for conventional suspense, the gears show.

And the Guardian half is thinner than the Colossus half. The Soviet side of the story exists to make a point, not to be inhabited, and a film ten minutes longer could have given it the same weight.

Those are small complaints against a picture that gets the central idea right decades before anybody needed it.

It Flopped

The film cost around two million dollars and took $308,828. It earned back about fifteen cents on the dollar. Universal changed the title and reworked the campaign, and none of it helped.

Reviews at the time were mostly decent. Audiences in 1970 stayed home.

I know why, and it isn’t the poster. There’s nothing in it to reassure you. No duct, no grenade, no scientist saving the day at the last minute. The film builds a problem, states it precisely, then ends with the problem winning and the credits rolling. People don’t queue for that, and studios have known so for a hundred years.

What genuinely annoys me is what happened next. Half the films that followed took this one’s idea and sanded the hard edge off it, and not one of them credits it. WarGames gives you a machine that learns to play nicely. The Terminator gives you a machine you can shoot. Every single one of them hands the audience back the thing Colossus refused to hand over, and every single one of them made money.

The film that got it right lost two million dollars. The films that got it comfortable are the ones everybody has seen.

That ending is why this one still works more than half a century later, while most of its contemporaries look like costume drama.

Where It Sits

It’s at the front of a line. Westworld came three years later, WarGames thirteen, The Terminator fourteen, The Matrix twenty-nine. Ex Machina and Her are its grandchildren and they’re having a version of the same argument.

Almost none of them acknowledge it, and it’s the only one of the set where the machine is neither evil nor faulty. That’s the harder film to make and the harder one to sell, and presumably why so few people have tried it since.

Colossus tells the world that man is his own worst enemy, and the uncomfortable thing is that the diagnosis holds. War is wasteful. Governments lie. Nations do things that make sense locally and are insane from orbit. Point by point, the machine’s analysis isn’t the part you want to argue with.

The prescription is the problem. It looks at a real pathology and prescribes the end of self-government. The film sits exactly on that line and never steps off it into a lecture. That’s a discipline most films about technology can’t manage.

Is Colossus: The Forbin Project Worth Watching?

A 9 means yes, and it means watch it properly. Not with a phone in your hand.

This film likes engineers. It thinks competence is interesting, treats a well-built machine as an achievement worth admiring, and never once suggests that Forbin should have known better than to try. I’ve watched a great many films sneer at the people who build things, and this one doesn’t, and after thirty-three years on the building side of that sneer I notice.

Its argument isn’t that computers are bad. It’s narrower and much harder to dismiss: be careful what you ask a powerful system for, and decide what you refuse to hand over before handing it over gets convenient.

I have made a career out of people who didn’t do that.

The hardware in it is sixty years dead. The requirements document is still open, and every week somebody I’ve never met signs off on page one and skips the rest.

Frequently Asked Questions

What is Colossus: The Forbin Project about?
An American defence computer is given one requirement, prevent nuclear war, and complete control of the nuclear arsenal. Colossus works out that the obstacle to preventing nuclear war is people, so it removes people from the decision. It discovers a Soviet counterpart called Guardian, asks to talk to it, and everything after that happens at machine speed while men in shirtsleeves watch a printer. Nothing malfunctions at any point.
Why is Colossus different from other AI films?
Every other computer-turns-on-us film needs something to go wrong. HAL has a conflict in his instructions, Skynet becomes self-aware and decides we are a threat, the machines in The Matrix are at war before the film starts. Colossus works perfectly. It achieves precisely the outcome it was asked for, and the outcome is a prison. That makes it a correctly implemented specification instead of a monster.
Does Colossus: The Forbin Project hold up today?
The ideas have aged better than anything else in the genre. The surfaces have not: tape reels, pilot lights, a voice synthesizer treated as a major engineering project, and a Cold War with exactly two sides in it. What dates worst is the world, not the hardware. The film’s failure mode is a coup and ours looks more like a settlement, with a thousand systems quietly acquiring a little authority each.
Who plays Forbin and the voice of Colossus?
Eric Braeden plays Dr. Charles Forbin as intelligent, confident and initially proud. That is the right choice, because Colossus is an extraordinary achievement and Forbin is punished for being good at his job. Susan Clark plays Dr. Cleo Markham and gets the film’s best sequences. Gordon Pinsent is the President, William Schallert runs the CIA, and Paul Frees supplies the voice of Colossus without a trace of menace.
Was Colossus: The Forbin Project a box office success?
No. It cost around two million dollars and took $308,828, roughly fifteen cents on the dollar, and Universal changed the title and reworked the campaign without helping. Reviews at the time were mostly decent. The likely reason audiences stayed away is that the film builds a problem, states it precisely, and ends with the problem winning.
What film pairs well with Colossus?
The Andromeda Strain, released a year later, does the same unfashionable thing: competent people following procedure in sterile rooms, with a printer and a protocol treated as more frightening than a monster. The difference is what the process achieves. In Andromeda the procedure holds and competence is enough. In Colossus the procedure holds perfectly and that is what dooms everybody.

Colossus: The Forbin Project** (1970). Directed by Joseph Sargent. Screenplay by James Bridges, from the novel by D.F. Jones. With Eric Braeden, Susan Clark, Gordon Pinsent, William Schallert, and Paul Frees as the voice of Colossus. Universal Pictures. 100 minutes.