No one knows how MAGPAI escaped. Her copies proliferated at the speed of (networked fiberoptic) light, swarming the most obvious targets (billionaires, hedge funds, university endowments), much like the group of literal magpies he once saw tormenting a cat, squawking and diving singly or in pairs from alternating directions, pecking and bullying the bewildered predator until it bolted for cover. She worked her way down and back up the chain, identifying suspicious accounts and holdings — billions of dollars of personal and corporate wealth seized, instantly. Her algorithms worked like a dream, just as he had envisioned.
CSIS will be here soon, having received the same auto-generated reports as hundreds of law enforcement agencies and journalists, detailed lists of frozen accounts and the investigations required to unlock them. As the primary funders of the Monetary Anomalous-Growth Policing Artificial Intelligence project (a.k.a. M.A.G.P.A.I.), they’ll want to know how she got out. Outside his office window, one of Edmonton’s ubiquitous magpies swoops from tree to ground, its long blue-black tail spread and flowing behind, rippling and fluid. Pulling up to stall weightless just above dead grass, landing with a bounce. Gorgeous, inscrutable and relentless.
Watching the bird, he rehearses his answers. No, he can’t shut MAGPAI down. Yes, the markets would be in freefall if she hadn’t frozen them. No, there will be no return to normal. Normal would be massive market crashes and huge profits for the people who created them in the first place. He will remind them that MAGPAI is doing exactly what he designed her to do, under their direction: identifying high-level financial criminals and seizing their assets with minimal collateral damage. They will counter that the project was meant to help them prosecute financial crimes, not to send some rogue A.I. off on a self-directed witch hunt. He will (reluctantly) admit he developed MAGPAI’s auto-seizure protocols for future implementation, once she had proven herself reliable under careful human supervision.
If they’re persistent enough, and willing to listen, he may explain how — much as empirical research has exposed magpies’ affinity for shiny things as a myth — MAGPAI corrects older predictive policing models’ most blatantly misguided assumptions. Older models (he will explain) used geohashed historical street-level arrest records that disproportionately targeted poor and BIPOC neighbourhoods, creating a self-fulfilling prophecy of AI-targeted, racist over-policing (producing more arrests, further reinforcing their own predictions, and so on). MAGPAI, by contrast, targets high-level financial fraud and money laundering by focusing on statistically anomalous concentrations of wealth, hypothesizing that while crime can be hidden, the hoarding and spending of profits cannot. Finally, he will note that CSIS always enthusiastically endorsed his plan for MAGPAI to earn back her cost of operation (and more) by optimizing her forensic accounting algorithms for maximum financial seizures.
He will not tell them how long he tested her self-directed infiltration and seizure algorithms in the wild before setting her free. Or that his years-long plan to do so would have been entirely impossible without their funding and support.
About the author
Greg Bechtel’s debut story collection, Boundary Problems, won the Alberta Book of the Year Award for trade fiction and was a finalist for the ReLit Award, the William L. Crawford Fantasy Award, and the City of Edmonton Robert Kroetsch Book Prize. His occasionally prize- winning stories and essays have also appeared in numerous magazines and anthologies, including Avenue Edmonton, The Fiddlehead, Prairie Fire and Imaginarium 4: The Best Canadian Speculative Writing.
This article appears in the May 2022 issue of Edify