What would happen if an AI gained control of the US military’s nuclear stash and decided to preemptively win World War 3 before any perceived enemy nations could react?

Fans of cinema from the 1980s may recognize that query as the plot to the classic science-fiction film “Wargames” starring a young Matthew Broderick. It was a great but terribly silly movie that paired nicely with popcorn and suspended disbelief. Nevertheless, the question it asked remains valid.

[Note: Spoilers ahead because the movie is more than 30 years old]

In the film, the AI is eventually stymied by Boolean logic after attempting to “win” against itself at Tic-Tac-Toe. Those who understand how AI actually works might find the entire plot of the movie preposterous, but the ending is especially chuckle-worthy. At least it used to be.

Today’s computers use binary logic so, in essence, everything’s a yes or no question to an AI running classic algorithms. Even when researchers design AI that “rates” things, they usually just break the degrees between ratings down into yes-or-no questions for the AI to answer in increments.

But tomorrow’s AI won’t be stuck in the mire of classical physics. Useful quantum computers are just around the corner – they should be here sometime between next Tuesday and the year 2121.

With quantum computers, our military systems won’t be constrained to yes-or-no questions and they certainly won’t have to run boring old binary simulations to determine the confidence factor for a given operation.

[embedded content]

Prasanth Shyamsundar, a researcher at the Fermi National Accelerator Laboratory, a Department of Energy research lab for the US government, recently published a fascinating paper describing two new types of algorithms that could revolutionize quantum computing and, potentially, lead to a quantum brain for military AI systems.

A press release from Fermi describes what the algorithms do by invoking the image of an AI sorting through a stack of 100 assorted vinyl records to find the sole jazz album. Under the normal AI paradigm, a deep learning system would be trained on what jazz sounds like and then it would parse each record individually until one of them meets a pass/fail threshold for jazz.

The first of the algorithms Shyamsundar proposes would, essentially, allow that same AI to sort through the entire stack of albums at the same time.

Quantum AI isn’t smarter, it’s just fast and takes advantage of “superposition.” Where classical AI works in a black box, quantum AI could exploit superposition to operate in many black boxes at once. 

Unfortunately, that doesn’t mean it comes up with the right answer. When it’s a yes-or-no question, the odds are good. But when it’s a question that requires non-Boolean logic, such as rating 100 albums for their jazzyness on a scale of 1-10, even a quantum computer needs a different kind of algorithm.

And that’s what the second algorithm does, according to Shyamsundar.

Per a press release from the Fermi lab:

A second algorithm introduced in the paper, dubbed the quantum mean estimation algorithm, allows scientists to estimate the average rating of all the records. In other words, it can assess how “jazzy” the stack is as a whole.

Both algorithms do away with having to reduce scenarios into computations with only two types of output, and instead allow for a range of outputs to more accurately characterize information with a quantum speedup over classical computing methods.

To be clear, Shyamsundar’s work has nothing to do with military operations and the Fermi lab, as mentioned, belongs to the DoE (not the DoD). Their paper represents the groundwork towards basic functioning quantum algorithms.

But what is a military AI technology if not an innocuous, basic algorithm persisting?

The problem with today’s military logic systems – and the one in the movie “Wargames” – is that they’re all based on binary thinking.

You can run a million simulations on advanced military software using cutting-edge AI, but eventually the limitations of “pass/fail” thinking will reduce almost any conflict into an arms race that ends in either stalemate or mutually-assured destruction.

But, what if the confidence factor for a given military operation didn’t rely on binary simulations? The same quantum algorithms that can determine which album in a given stack is a jazz album 10 times faster than a binary system, and how jazzy a given album is, could easily determine which combination of feasible operational strategies would result in the highest overall confidence factor for a military campaign.

In other words, where Sun Tzu was said to be able to envision an entire battle unfolding in front of his eyes before it happened, and modern software such as CMANO can simulate entire operations, a quantum system running simple non-Boolean algorithm solutions should be able to surface strong predictions for the outcome of a multi-step war campaign.

Published April 7, 2021 — 18:39 UTC