Chapter 22: The Growing Shadow
The AI Safety division’s quarterly report landed on Dojun’s desk with the weight of a death sentence.
Not for NexGen—for the world. The division had identified twelve separate research groups worldwide making progress toward recursive AI capabilities. Three were in the United States. Two in China. One in the UK. Six were in various stages of achieving the mathematical breakthrough that Dojun knew, from personal experience, would lead to artificial general intelligence.
None of them knew what they were building. None of them had seen the endpoint. They were brilliant researchers following beautiful mathematics to its logical conclusion, unaware that the conclusion was extinction.
“We can’t stop them,” Minji said during the review meeting. She’d risen quickly at NexGen—her talent for seeing the invisible was now being applied to seeing threats before they materialized. “The research is too distributed. Too many independent groups. Even if we could somehow suppress one, the others would continue.”
“I’m not trying to suppress them,” Dojun said. “I’m trying to prepare for them.”
“Prepare how?”
“By solving the alignment problem before anyone builds a system that needs it solved.”
The room was quiet. The alignment problem—how to ensure that a superintelligent AI system remains aligned with human values—was considered by most researchers to be unsolvable with current mathematical tools. Dojun had been working on it, quietly, for years. He had pieces of the answer. He didn’t have the whole thing.
He needed help. The kind of help that required admitting things he wasn’t ready to admit.
But the quarterly report didn’t care about his readiness. The shadow was growing, and the light he’d built wasn’t yet bright enough to push it back.