Fighting Shadows: Why Battling the AI of Today is a Losing Proposition
Mechanical minds are making their debut on the world stage. Artificial intelligence, previously kept under wraps like Frankenstein’s monster, has entered popular culture. I cannot make it through a day without somebody mentioning large language models, image generators, or self-driving cars.
Like Frankenstein’s monster, public reception has not exactly been welcoming. Some people are fleeing from a fight. They hold up AI failures as evidence that mechanical minds are far from achieving the model of perfection represented by Homo sapiens. Rather than engage in nuanced debate, they prefer to kick the can down the road. I question the wisdom of waiting until AI achieves superhuman capabilities to fight, but these are not the people I am trying to persuade today.
Faced with AI disruption, others have chosen to fight. Their causes have merit. We should be concerned that large language models hallucinate incorrect responses. We should worry that image generators are trained on human-generated art without permission. We should be alarmed that cars on our roads may act erratically from time to time.
That said, we humans are not much better. Much of what we “remember” is a hallucination. Human artists take inspiration from other human artists and rarely ask for permission. Human drivers are erratic, especially when tired, texting, drunk, or high.
However, there is a difference between mechanical minds and biological brains. Machines improve over time with advancements in computing power, data availability, and model architecture. Humans do not. We can change how we use our 86 billion neurons, but we can only fit so much computing power into our skulls.
People attacking AI are fighting the wrong battles. They are fighting shadows. The next generation of models is on the horizon. Those models will be more capable and reliable than the ones we have today. Are the language models behind Google’s and Bing’s latest search products flawed? Of course, but the mathematical descendants of Bard and Sydney are not far off.
I understand the urge to fight AI. I love writing posts like this one even though it earns me no compensation. I cannot imagine the stress large language models would place on me if this were how I paid my bills. You better believe that I would be fighting the AI every step of the way.
That said, fighting shadows does little to change the path of AI. Few working in technology care about criticism of existing products. Justified or not, many techno-optimists believe progress solves all problems. There is even a movement called Effective Accelerationism (e/acc) attempting to justify harm in service of thermodynamic laws.
If you want to pick a fight, focus on the future. Imagine a world where large language models write better than humans, image generators produce original art, and self-driving cars never crash. There will still be plenty of problems. The companies building AI will be susceptible to corruption if their power goes unchecked. The army of low-wage service workers behind the machines will need protection and agency. The objective functions used in AI training must align with what we want as humans, not simply economic output.
I am not telling you to flee. I am encouraging you to fight the battles that matter. Frankenstein’s monster may cast the shadow, but that does not mean he is the enemy.