Big Data: A Startup Thriller Novel

Chapter 9

I found him. Taye Flanagan. He'd been trying to contact me in the Woods for months. I just hadn't whitelisted him, so none of his messages got through.

His app was good. Actually, it was exceptional. He called it Gaia. The kid had an extraordinary gift with computers. I'd thought I was advanced when I was his age, but there was no way I would have been able to grasp the concepts he was working with here.

"Ma'am, can I help you?"

"No, just checking out the new iMac. The screen's pretty amazing."

The salesman agreed and tried to tell me more, but I brushed him off as politely as I could. He gave me a snarky look and walked away.

Now I had to figure out what Taye was trying to tell me. To an end user, an app is just the utility you get from it. Like sharing a picture with a friend. Looking at the code behind the app is like pulling back the face of a watch.

Most watches have digital quartz movements that look like cheap circuit boards printed in China. Most apps are the same. Their code is a mess of open-source libraries patched together by some spaghetti and meatballs.

But every so often, you can pull back the face of a watch to find a secret work of art, full of perfectly placed precious metals and gems. It's breathtaking. That's how I felt about Taye's code. This kid was a true craftsman. It wasn't just readable code; it was a joy to read. It was like looking at the Mona Lisa of code.

Just then, I noticed the sales guy looking over my shoulder.

"Can I help you?" This time, I was asking him.

"What programming language is that?" he asked with such interest that I decided to answer fully.

"Conifer. It's a dynamically typed Lisp without parenthesis tuned for machine learning."

"Like PHP?"

I shouldn't have even tried.

"Yes, like PHP."

He walked away. I had no idea if he caught the sarcasm.

I needed to get back to work. What had Taye done here? The code was beautiful. Most people, when building an app in the Woods, would take a generic template or example code and repurpose it to their liking. Even large and popular apps on Ancien used the templates as almost a de-facto standard.

But Taye had started from scratch. I could see straight away that he was using a neural network, which mimicked the way scientists thought neurons communicated. But figuring out what a neural network was trained to do wasn't necessarily straightforward; most of the magic of neural networks was hidden in layers of data that even the programmers who built them couldn't understand.

The code wasn't documented, but it was beautiful. I wrote the initial version of Ancien myself. Most of the tutorials were based on my first templates. This code was much better.

As I navigated through the maze of code, I finally figured out why I was having so much trouble understanding what it did.

If Taye's code actually worked the way I thought it did, he had made a major breakthrough in machine learning. His application was trained to understand computer code itself.

He had taken Ancien—the world's most powerful supercomputer—and unleashed all its intelligence on learning the fundamental language of computers themselves.

This was brilliant. Machine learning experts tended to research specific classes of problems: photos, videos, self-driving cars, handwriting recognition, speech recognition, financial transactions, even Internet search results. And at this point, the majority of the commercial applications of this research ran in Ancien's cloud platform.

But Taye was exploring a new class of problem: having computers fundamentally understand themselves. I gasped as realization flushed over me.

Taye hadn't written Gaia.

Maybe he'd written the first version of it, sure. But since then, Ancien itself had been refining it and improving it on its own. My own baby was responsible for this code.

This was the most significant development in machine learning in fifty years. Code understanding code. Without human assistance, interpretation, or intervention. This could solve so many other problems in the world of artificial intelligence. This was the holy grail.