Neptune’s Hamlet four years in the making
Municipal units and plastic bags: what’s next?
Twinning project for deadly Nova Scotia highway seeking environmental ...
58,830 people in HRM live in poverty
Highly rated international private boarding school franchise coming to ...
Fire crews respond to Upper Musquodoboit pellet mill explosion
Vote for Nova Scotia's newsmaker, top story of 2018
2018 Christmas Lights map
PHOTOS: Cinderella brings sparkle to the IWK
In the last 10 years, on two separate occasions about five years apart, the UN has been given very serious warnings by some of the world’s top scientists concerning the danger of our growing reliance on “computerized decision-making.” Of late, it is enjoying a very wide and uncritical vogue under the term artificial intelligence.
It is somewhat flabbergasting (and a little chilling) to encounter the uncritical, indeed naive, enthusiasm of many of today's computer wunderkinder for “AI” as the magic solution to our current problems.
It is even more alarming to discover that many have never heard of that mantra made famous by computing’s early pioneers: GIGO. “Garbage in, garbage out” was perhaps a little easier to appreciate when one had to manually punch data cards or remove raw data first-hand from coding sheets.
But it is perhaps no coincidence that with both the provenance and end uses of raw data now clouded in layers of remote offshore sourcing, processing and re-selling — all beyond the overview of most legal jurisdictions — the man credited with inventing the internet recently repented of his invention.
Flight 610 - Battling a computer for survival
The tragic recent crash of Indonesia’s Lion Air flight 610, which killed all 189 aboard, was a startling illustration of the dangers of losing sight of data sources and of dependence on automated critical controls.
It seems likely, based on present information, that the accident was caused by a very determined anti-stall piece of software which processed data from a faulty sensor. When one views the minute-by-minute line graphs of the battle for control between the pilots and this new Boeing software (apparently not documented in the flight manuals used by the pilots) one can almost hear a version of that refrain from Stanley Kubrick’s classic film, 2001: A Space Odyssey, when computer Hal prevents human Dave from unplugging him: “I’m sorry, Dave, I’m afraid I can’t do that.”
For at least six agonizing minutes, the pilots waged a futile struggle to arm-wrestle control back from the automated stall-prevention system that persistently continued to push down the 737 aircraft’s nose until the fatal impact.
Don’t get me wrong. I’m no Luddite. My dad and I never missed watching a launch of the Mercury space flight program. And I was on the ice as a schoolkid for the flight marking the 50th anniversary of Alexander Graham Bell’s Silver Dart at Baddeck, returning 50 years later with my own son for the 100th. But as we now see clearly in the case of data behemoths like Facebook, among other cases of runaway technology, things literally seem to have gone off the rails.
Weaponized data a true weapon
It is not surprising that in our modern zeal for “higher, faster, stronger,” we have overdriven our societal and ethical headlights. Early reviews at MIT and elsewhere of some existing AI-powered systems, such as police dispatch, reveal prejudices and negative stereotypes toward vulnerable populations encoded within the operative algorithms. And now that this genie with its DNA consisting of one-third big data, one-third high-speed processing and one-third gluttonous lust for commercialization (a.k.a. “monetization”) is out of the bottle, it may prove extremely difficult to re-cork this potent concoction.
The urgent need for critical scrutiny, for new and increased regulation, and for more modest expectations of AI and its cousins is fuelled by a growing realization, and by a lengthening case list of its serious dangers. From the relative ease with which voting machines can be hacked, to the deaths of patients in otherwise routine surgeries at the hands of robot surgery systems, to the crash of passenger planes and self-driven automobiles following their own often inscrutable “whims,” we are walking along the edges of a dangerous cliff.
And this is to say nothing of the damage that big data, once weaponized, can do. Almost all the vital controls for today’s nuclear power plants, hydroelectric dams, oil refineries, and air traffic control systems (let alone the actual firing systems of today’s highly lethal weapons of war) rely on computers, some incorporating interruptible and unshielded Wi-Fi links.
The extent to which a well-resourced, IT-savvy state agent could cripple and damage an adversary in an all-out cyber war would make Hollywood’s scary sci-fi movies seem mild. Imagine a hundred post-Katrina New Orleans scenarios.
Despite their ongoing concerns over recent Chinese and Russian military buildups of conventional weapons, senior U.S. military commanders have recently admitted, publicly, that the scenario that keeps them up at night is the threat of all-out cyber attack.
What is to be done? As a recent BBC discussion concerning the need to corral Facebook and its analogues concluded, “It’s easier said than done.”
Increased public awareness of the dangers of our rapidly growing dependence on these so-called “smart machines” must be a start.
Heaven knows, policy-makers and legislators will have their hands full trying to draw the lines between privacy and profits, between innovation and safeguarding public security, and between local needs and international legal standards.
But let’s begin by removing some of the blinding sheen from the current idolatry over AI.
Brian Joseph, a native of North Sydney, studied physics at St. Francis Xavier University and public policy at Harvard. He served two terms as commissioner at the Law Reform Commission of Nova Scotia.