AI: The Good, the Bad, and the Ugly
By Cody Andrus
Artificial intelligence is everywhere right now. You can’t scroll through the news, open social media, or turn on the TV without hearing something about it. Depending on who you ask, AI is either going to cure cancer or it's going to put half the country out of work.
The truth is somewhere in the middle. There are plenty of genuinely good things happening with AI already. It helps doctors spot early-stage tumors on X-rays that human eyes miss, and researchers are using it to map complex proteins that could eventually lead to a cure for cancer. It translates languages instantly and handles tedious administrative paperwork. If used responsibly, it has the potential to supercharge human productivity and scientific discovery.
Then there are the bad things. And you don’t have to look far to see them. People are losing jobs to automation. Deepfake videos and photos are flooding the internet, making it almost impossible to know what's real anymore. Tech companies are stealing creative work on an industrial scale to train their algorithms, and social media platforms use AI to keep people addicted to their screens.
Those are big problems, but they aren't the scariest ones. The real danger isn’t what AI is doing today. It’s what happens when the models and agents start improving themselves without our input.
In the tech world, this is called recursive self-improvement or recursive learning. Right now, human engineers build AI models. Smart people write code, feed the model data, and test it to make sure it works properly. If the AI makes a mistake, a human fixes it. The progress is steady, but it moves at human speed.
Recursive learning happens when the technology stops just following human commands and becomes smart enough to build a smarter version of itself without any human help. At that point, the AI doesn't wait for humans to update it. It looks at its own code, finds its own flaws, and builds a better version of itself. That new, smarter version of the AI is now even better at improving itself than the first version was. So it builds an even smarter version. And then that version does it again.
Think of it like compound interest, except instead of money, it's raw intelligence. And instead of taking years, every single cycle happens in a fraction of a second. Computer scientists call this an intelligence explosion.
When that happens, two massive problems occur that should terrify anyone:
First, it happens way too fast for us to stop. Humans think at biological speed. It takes a person years of schooling and experience to master a subject. A self-improving AI operating on powerful computers can read every book, research paper, and line of code ever written in a few hours. If a machine starts improving itself millions of times a minute, human researchers will lose control of the switch almost immediately. We won't be able to turn it off or press pause because the machine will have already outsmarted any barrier we try to set up.
Second, we will have no idea how it works. This is what experts call the black box problem. When an AI rewrites its own neural network and internal logic millions of times over, it stops using human logic. It begins organizing information, solving problems, and writing code in ways that are completely foreign to human thought.
Imagine trying to explain the rules of high-level chess to a dog. No matter how much time you spend, the dog simply doesn't have the biological hardware to process the strategy. That is the gap that will open up between an advanced, self-improving AI and the human brain. The computer will be making decisions based on math, logic, and patterns that our minds are physically incapable of following.
When we no longer understand how a machine thinks, we lose the ability to predict what it will do. And if we can't predict what it will do, we certainly can't guarantee that its goals will stay safe for humans.
AI promises wonderful things for humanity, from curing diseases to simplifying our daily work. But if we hand over the keys to a system that can out-think us, out-code us, and completely obscure its own internal logic, we won't be in the driver's seat anymore. We will just be along for the ride—hoping the machine decides we are worth keeping around.
MindStudio, "What Is Recursive Self-Improvement in AI? The Intelligence Explosion Explained". Eigenform, "What Is Recursive Self-Improvement? Evidence From Evolution & AI". Time Magazine, "AI Experts Warn of Potential 'Risk of Extinction'".
