The Evolution of AI in 10 Fast Years
By Katherine McKean, Junior and President of my high school AI Club
Ten years ago, artificial intelligence was mostly background noise. It autocorrected our texts, guessed our next Netflix binge, and occasionally misidentified a cat as a muffin. Today, it’s writing essays, generating videos, predicting proteins, and having philosophical arguments with middle schoolers. The evolution from basic rule-based bots to what we now call generative AI hasn’t just been fast—it’s been very teenage. Messy, moody, unpredictable, and occasionally brilliant.
from rules to learning
Old-school AI was like a programmable calculator. You told it what to do, and it did exactly that. These were rule-based systems. Think spam filters or your Roomba avoiding stairs. Helpful? Yes. Life-changing? Not really.
But around 2015, things started to change. Deep learning—the idea of training AI on massive amounts of data so it could find patterns on its own—began to take off. Suddenly, AIs weren’t just following rules. They were learning. You didn’t have to tell them what a cat looked like. You just had to show them 50,000 photos of cats and let them figure it out. And maybe also 50,000 muffins for contrast, just to be safe.
2018: transformers and language models
The real plot twist happened in 2018 when researchers introduced transformers. No, not the robot cars. Transformers are a type of model architecture that helped AI understand context in language. This is the tech behind the scenes of all your favorite chatbots now. It allowed AI to move from clunky sentence autocomplete to full-on paragraph generation that sounds eerily human.
That’s when we got GPT, BERT, and the whole acronym crew. These models didn’t just answer questions—they told jokes, wrote poetry, and sometimes got weirdly obsessed with pirates if you gave them too much creative freedom.
2020–2022: generative art and the deepfake boom
While language models were learning to talk, another corner of AI was learning to draw. Generative adversarial networks (GANs) made it possible for AI to create fake faces, realistic photos, and eventually entire TikToks of celebrities saying things they definitely never said.
This is when “deepfake” entered the public vocabulary. Some were funny. Some were scary. All of them reminded us that reality had a new contender—and this one didn’t even need lighting or makeup.
2023: chatbots everywhere
ChatGPT launched in late 2022, but 2023 was the year it became a daily tool for millions of people—including a suspicious number of tired high schoolers who suddenly sounded way too formal in their history essays.
By summer, there was a chatbot in every browser, school, and app. Your grocery store app had a chatbot. Your bank had one. Even your fridge had opinions. Claude, Gemini (formerly Bard), and others joined the conversation. Each bot had its own personality quirks. Some were helpful. Some were like talking to a mildly distracted librarian with a Wi-Fi problem.
2024–2025: video enters the chat
And then came Sora.
If 2023 was about text and images, 2024 was the year video generation took the spotlight. Sora by OpenAI showed off AI’s ability to generate moving visuals from text prompts. You could ask it for “a golden retriever skateboarding through a sunset in Paris,” and it would give you something uncannily close.
In our AI club, we used Sora2 to storyboard short videos. One member created a visual novel set on Mars using just descriptions. Another tried to generate a “spooky hallway with flickering lights and suspicious cats.” It worked, mostly. One cat had twelve legs. But we respect the ambition.
Sora2 improved on Sora with better coherence, smoother animation, and fewer floating limbs. It’s now a go-to for prototyping animations, planning content, or just creating ultra-specific memes.
what high schoolers actually think
In our school’s AI club, opinions on AI vary wildly. Some think it’s the future of everything. Others just use it to plan their outfits. But most of us agree: it’s useful, sometimes spooky, and way better at staying organized than we are.
Here’s a sampling of what I heard in one lunch period:
“I asked ChatGPT to help me write a breakup text. It turned it into a formal resignation letter.”
“Sora2 made a video of me as a detective. I looked like an anime potato.”
“Honestly, AI is better at studying than I am. It doesn’t get distracted by Doritos.”
We’re not exactly afraid of AI. But we’re not ready to give it our group projects either.
why it matters
Most students I know don’t see AI as magic or menace. It’s a tool. A powerful one. It can help you learn, cheat, create, or distract you for three hours with a conversation about whether cake counts as bread.
The real shift is how quickly we’ve gone from “AI is for tech companies” to “AI helped me plan my prom theme.” That jump—from background helper to creative partner—is what makes the last ten years feel less like a timeline and more like a plot twist.
the speed of change
If you blinked, you might’ve missed a version. GPT-3 turned into GPT-3.5 and then GPT-4o. Midjourney went from weird blobs to actual art. Every week, it feels like something new is launching. It’s like being in a group chat where one friend keeps texting updates faster than anyone can read them.
That’s part of the excitement—and part of the chaos. It’s hard to learn something that changes this fast. But it also means students get to experiment with the cutting edge while it’s still being shaped.
what comes next?
We don’t really know. But here’s what we’re watching for in our club:
- AI that understands emotion better. Right now, it’s good at guessing—but bad at subtlety.
- Better tools for detecting fake content, especially in videos.
- Ways to combine all the modes: text, images, video, and voice—in real time.
And of course, we’re still waiting for an AI that can write group essays without sounding like five people who forgot to meet. (Which is…accurate.)
student thoughts on Sora2
Sora2 has its fans. It also has its skeptics. Some students love being able to generate visuals for their creative projects. Others worry it’s just another way to procrastinate.
One student used Sora2 to reimagine a scene from Macbeth in space. Another tried to recreate their Minecraft world as a film trailer. The results were strange and beautiful and occasionally hilarious. But more than anything, they sparked ideas.
For a lot of students, Sora2 made AI feel less like a science project and more like a sketchbook. That shift—from productivity tool to creative engine—is something even the skeptics noticed.
closing thoughts from the club
The past decade of AI has been fast, weird, and full of surprises. And if the next decade moves just as quickly, we’ll probably be asking our fridge for career advice by college. Until then, we’re testing, learning, and occasionally laughing way too hard at a 12-legged cat in a haunted hallway.
Want to bring the power of AI to your school? Check out this step-by-step guide on How to Start a High School AI Club: 6 Easy Steps for Success.
