Dr. Naleli Matjelo, along with his team members, Dr Lerato Lerato and Professor Molibeli Taele, at the National University of Lesotho (NUL), have found a way to explain how Artificial Intelligence (AI) works, using simple math.
But first, what is artificial intelligence?
AI is like a smart computer that can make decisions by itself, just like humans. It can do many things, like predicting the weather, recommending your next favorite song, or even helping doctors understand your illness.
But AI has a problem. Most of the time, people don’t know how it makes its decisions, even people who created it.
It’s like a magic trick—nobody sees what happens behind the curtain!
Dr. Matjelo and his team are trying to lift that curtain to show how AI works.
What did they do?
They studied a specific type of AI called an Artificial Neural Network (ANN). Think of ANNs like pretend brains for computers. These “brains” have tiny parts called neurons that work together to solve problems. The problem is, these neurons are so fast and complicated that even the best scientists sometimes don’t understand them.
But Dr. Matjelo had an idea: “what if we use math to explain what happens inside these computer brains!”
His team created a new way to study ANNs. Instead of treating them like mysterious machines, they used math equations to show how the neurons talk to each other. These equations work like a recipe that anyone can follow and understand.
How did they test their idea?
Dr. Matjelo’s team tried three tests:
- Lots of Clean Data: They gave their ANN plenty of high-quality information. The results were good! This showed that their math equations worked well when there was enough good data.
- Less Data: When they gave the ANN fewer examples, the results were not as good. This showed that the more information you give to an ANN, the better it performs.
- Noisy Data: Finally, they added random errors to the information. This made the ANN’s results less accurate. Dr. Matjelo explained this happens because their method uses something called “derivatives,” which makes errors bigger.
Why is thier work important?
When people understand how AI works, they can trust it more. Imagine going to a doctor and the AI says, “Take this medicine,” but nobody knows why! That’s scary, right?
The same goes for banks. If AI says, “This person should get a loan,” the bank can check and confirm if the decision is fair.
What’s next?
Dr. Matjelo and his team are not done yet. They want to make their method work for even bigger and smarter AIs. They also want to find ways to handle noisy data better.
One of the coolest things they found is something called symmetry. Symmetry means that sometimes you can swap parts of the AI, and it will still work the same way. This discovery can help scientists train AI faster and more easily.
Why should you care?
Dr. Matjelo’s work shows that even the most complicated technology can be understood if we work hard enough. He is proving that AI doesn’t have to be a mystery anymore.
