why computers will rule nations. and why they’ll fail miserably.
Once upon a time, genteel academics and tech savvy utopians dreamed that the internet would allow us to create a global village and bring communities across the world together as people who’d never meet connected to share their cultures and ideas. Instead, we’ve been exposed to the open sewer of human depravity shoved into our faces by algorithms designed to keep our attention long enough for the platforms we use every day to show us ads for drinkable tortilla chips or sarcastic t-shirts throughout our doom-scrolling. We knew that at our core, we were deeply flawed creatures, but holy crap did social media bring that out to the shock and disgust of an audience now measured in the billions.
Now, you’re probably thinking that this is another tale of fake news and the threat of terrorism by aggrieved white people in the West because this is what people usually focus on after broaching this topic. But the root problem is far greater. Racism and bigotry are in no way exclusive to the West, and even in minority communities whose members all know full well what it’s like to face discrimination, there’s still animosity. Black convenience store customers cite Asian American proprietors watching them like hawks. Bisexuals chafe at being told to “pick a side” in LGBTQ+ spaces. In short, no matter how you look at it, humans keep finding new ways to be horrible to each other, even when they’re also victims of the same treatment.
With all that in mind, it’s no wonder pundits will periodically pop up to suggest that we turn to the cold, hard logic of computers to run societies for us since humans are clearly too defective to be trusted with doing it anymore. Proposals include letting AI run our markets, judge criminals, and even become de facto totalitarian leaders of the future. The thought is that unlike biased, easily corrupted humans with limited brainpower, computers are logical machines that use the power of math and have no emotions to sway them into doing something irrational or deliberately and purposefully harmful. As an elevator pitch, it makes sense. But given the state of AI today and how it’s trained, this is all wishful thinking bound to make things much worse.
why artificial intelligence can’t save us from ourselves
Of course, artificial intelligence can absolutely help with complex problems, including economic questions we struggle to answer. But it does its best work as an assistant allowed to explore and interrogate extremely well defined problems because AI learns by taking enough random guesses as to how to solve a quandary until it stumbles on something that works. It’s not magic. In fact, there’s an entire shadow industry involved in painstakingly training AI models to make them even halfway decent at their jobs, often feeding them gigabytes of real world data. Can you guess what happens if that data reflects decades of human biases, irrational bigotry, and discriminatory policies? That’s right, you get an extremely biased AI.
And the problem doesn’t stop there. Once you trained an AI to emulate discriminatory, biased policies, you can now claim that because the process is executed using cold, hard math, it’s impossible for it to be biased, and if RoboJudge 8000 sentences more minorities to longer and harsher sentences for the same crimes as members of the ethnic majority, well, that’s just how the universe is. Try to argue that the numbers may be flawed and prepare for every reactionary pundit to hyperventilate about “the Marxist assault on math and logic.” It will be hypocritical, as the right has a history of picking fights with math itself, but little things like truth and consistency don’t matter when your only goal in life is to “own the libs.”
Just to add another bit of bad news, these are not hypothetical scenarios. They’re happening as you read this, and computer scientists are raising countless red flags as they see poorly trained and deficient AI models not only failing to mitigate existing problems, but making them worse by coating the obviously bad outcomes with a patina of statistics and calculus, while allowing new and more elaborate ways to discriminate and harm. Sure, computers have no emotions, nor do they hate or prejudge anyone. The concept is as alien to them as another universe would be to us. But they learn from us, and if everything we feed them is tainted with our flaws, this is what they will emulate and enable on an industrial scale.
societies can’t outsource healing their flaws
Give machines the reigns over society anytime soon and all we’ll end up doing is create what will effectively be a cyberpunk dystopia built on today’s precursor to it. The rich will keep on getting richer, the poor will keep falling further behind, and the discriminated against will face efficiently automated and potentially escalated hardships, all of which will be excused by a shrug from the beneficiaries of this system and a quip about the math being what it is. The only way we can use machines to fix humanity’s glaring flaws is to enlist them in a very deliberate effort to detect and flag downright cannibalistic greed, bigotry, and racist policies, then block systemic actions that perpetuate them.
This is in part why wishing that computers effectively override humanity’s shortcomings is so irritating. It’s nothing more than trying to come up with a shortcut because we either don’t think we’re capable of having the necessary conversations and self-reflection, or admitting that we simply don’t want to and shifting the burden for self-improvement on a tool that can’t help us. Effectively, we have to come together, say that we know we’re all kind of garbage but we can and should do better, and agree that we should build systems that call out when we do harmful things, then listen to those systems. And in the process, we should also expect a lot of people to rebel, arguing that being shitty humans is their God-given right.
As much as we’ve been able to offload to our machines, we can’t offload building a better, more moral, and more equitable society to one of our tools and expect it to just make it happen. There is also a very good chance that we’ll never succeed in building a world where everyone truly has equal rights and fair treatment because we’re social animals but we’re also tribal creatures. But perfection is not the goal. Advancement and correction have to be our aims. As long as we can leave a better world for our descendants and make up for past ills as much as we can, we’re on the right track. And yes, machines can help us in this process if we enlist their help with open minds and critical eyes. But the only things capable of healing humanity are humans.