June sat down again to watch.
“I hesitated to send this for several reasons. First, you’ll probably think it’s creepy.” Her mom smiled, and June smiled back at the screen. Her mom had been a pain in the ass, but lack of self-knowledge was not one of her problems. “I wanted to make sure you know that I love you very much, and that you are the best daughter a mother could ever hope for.”
June shook her head and wiped away a tear. A video from the grave telling June she was a good daughter, something her mom had rarely said when alive. This was Hazel Cassidy in a nutshell. June looked back at the monitor.
“Second, and this will be hard for you to hear, if I’m dead, it is very likely that I was murdered for my work.” Thanks, Mom. Big dumb guys in black suits emptying her mom’s office was one pretty good clue. June getting thrown into a car was another.
Her mother kept talking. “As you know, my work has centered around neural networks and machine learning. How to get computers to solve increasingly complex problems. The holy grail with complex systems is to develop systems that can write themselves, can grow themselves organically, if you’ll excuse the inappropriate but inescapable metaphor.”
June knew this holy grail was becoming a reality. As chips grew faster and data storage became cheaper, previously impossible problems like voice-to-text had become a free feature of every new cell phone on the market. June loved that she could dictate an article to her phone while walking to the grocery store. But she remained morally conflicted about the many areas in which government agencies used facial recognition software. Catching criminals, sure. But watching everyone else? How do you separate the two, and where do you draw the line?
On the screen, her mom sighed. “You know I’ve always been more interested in the theoretical aspects of my work rather than the commercial applications. But my work and the work of others has been used to create functional technologies capable of monitoring individuals across the planet. And like you, I became concerned about privacy. About the abilities of governments and corporate entities to gain access to private information of individual citizens, to track citizens’ actions both online and in the physical world, while keeping corporate and governmental behavior increasingly secret. This imbalance of information is a significant problem.”
No shit, thought June. Just ask Snowden or Manning or any of the other major leakers of the last decade. She didn’t know if she agreed with what they’d done, but she certainly agreed that the conversation was necessary.
“So I decided,” said Hazel Cassidy, “to try an experiment. When I began to grow a new algorithm several years ago, I gave it several priorities. I wanted it to be curious about the information on the Web, and when requested, to be able to collect and summarize information from multiple sources. I also included what I thought was a slight interest in what lay behind security firewalls.”
Now we’re getting to it, June thought.
“I had no idea what those relatively few lines of code would become. The algorithm is growing exponentially. It has become quite good at its primary function, to collect and summarize information from multiple sources. However, in the last twelve months, the software has also begun to function as a kind of skeleton key. It is teaching itself to penetrate secure systems for the information hidden inside.”
The creases on her mother’s face deepened. “It’s not very advanced, not yet. But I’ve already seen corporate memos about ‘acceptable defects’ in medical devices. I’ve read letters from senators to the lobbyists paying for their reelection campaigns. I’ve seen a Pentagon list of proposed drone strike targets. None of it is pretty.”
June had her hand to her mouth. She’d known her mother worked on the bleeding edge of technology. She’d known her mother’s opinions about the disconnect between public and private secrets for years. But she’d never had a hint that her mother’s work might directly impact this disconnect. She kept watching.