“In your estimation, what’s the threat level? How likely is this program to be dangerous if it gets loose?”
“Any self-aware system—any AGI—will quickly try to improve itself. That would be one of its primary goals, and it would let nothing stop it from achieving this end. The program would make itself smarter, then, in turn, this more intelligent system would seek to make itself even smarter.”
“And on and on.”
“Also, any AGI would quickly acquire the same biological drives we do. The most important being self-preservation.”
“It wouldn’t want to be turned off . . . or die.”
“And it would do anything to stop that from happening. Secure any resource, thwart any threat, continually honing its creativity to accomplish this. And it wouldn’t even consider just immediate threats. With such immense computational power and an immortal life span, it would look for dangers beyond the horizon, far into the future, and devise strategies to stop them, even threats thousands of years from now. Worst of all, it would be continually looking at us, to judge if we’re a threat now or in the future. And if it deems we’re a danger—”
“Game over, like you said.”
“But this is also why Mara’s work is so important. She’s trying to build a friendly AGI, something that can protect us against a dangerous AGI that might arise later—correction, will arise. Beyond commercial corporations and government-funded labs, there are hundreds of stealth companies out there working on this, hell-bent to be the first, forsaking any worry about what might be unleashed.”
“How close are we to this happening?”
“Very close.” Jason waved an arm over the chaos of papers. “Google’s DeepMind program recently discovered the basics of quantum physics all on its own. A pair of AI translation programs began to talk to each other in their own undecipherable language and refused to translate their conversation. All around the world, robots have outsmarted their makers, exploiting loopholes in wildly imaginative ways. Other programs have even demonstrated human intuition.”
“Human intuition?”
“There was a lot of fanfare a couple of years ago when AlphaGo—Google’s DeepMind AI player—beat the world’s champion at the ancient Chinese game of Go. By some calculations, Go is trillions upon trillions of times more complex than chess. No one expected any computer to beat a human at Go for at least another decade.”
“Impressive.”
“That’s nothing. It took the company months to train AlphaGo for this competition. After this, Google took a new approach, letting its newest version—AlphaGoZero—teach itself, playing the game over and over again all by itself. After only three days of training, it grew so skilled that it beat Google’s original program in a hundred out of a hundred games. How? AlphaGoZero had intuitively developed strategies that no human had come up with during the thousands of years we’ve been playing the game. It literally transcended humankind.”
Gray swallowed, feeling a hollowness in his gut.
Jason wasn’t done. “So, when it comes to developing the first AGI, we are at that threshold right now.” He stared hard at Gray. “So maybe before we land, we need to refine our mission parameters. Not only do we need to stop Mara’s program from falling into the wrong hands—we need that program for the very survival of our species.”
On Jason’s iPad, a small text message box popped up.
They both glanced to it and read what was written there. It came from Lisa Cummings, the content curt and blunt.
Kat doing worse.
Must proceed stat with the test
No choice
Jason cast a worried look at Gray.
Gray knew how much the young Sigma analyst admired Kat. “That’s also a mission imperative,” he reminded Jason. “To find out what any of this has to do with what happened to Kat.”
And the kidnapping of Seichan and the two girls.
He tried his best not to let his fear for Seichan and his unborn child overwhelm him. He stared out the window, willing the jet to go faster. Beyond the future ramifications of this operation, there was a more immediate need, one close to his heart.
And not just his.
He imagined that hospital room in Princeton.
Hang in there, Monk.
9
December 25, 9:14 A.M. EST
Plainsboro, New Jersey
Buried in a subbasement of the Princeton Medical Center, Monk paced the control room of the MRI suite. A technician sat at a computer, calibrating the giant magnetic ring in the next room. Another two worked at flanking stations. The group whispered in their arcane language: Any ghosting or blooming? Looks good. STIR and FLAIR all set.
The space—with its dimmed lights, the bustle of activity, the urgent murmurs—reminded him of a submarine’s conn, aglow with sonar and tactical displays. But here the officer of the deck was Dr. Julian Grant, a Harvard-educated neurologist who specialized in altered states of consciousness, from comatose patients to the various spectrums of vegetative states.
The researcher wore a knee-length lab coat over blue scrubs. His shock of white hair belied his age—just fifty-four—suggesting he had gone prematurely gray. Maybe due to some side effect of the massive magnetic energies generated by his custom-built MRI.
Dr. Grant stood with his hands clasped behind his back before a wall of OLED screens. The neurologist studied the baseline images of Kat’s brain. Lisa stood with her colleague, their heads bent together, conferring in low tones.
Monk’s anxiety increased with each pass as he paced across the room. He kept one eye on a station monitoring Kat’s vitals. The team had transferred Kat from D.C. to Plainsboro, New Jersey, via a medevac helicopter. Still, the flight had taken nearly ninety agonizing minutes. Every bit of turbulence spiked Monk’s blood pressure.
While Kat had handled the flight like a trooper, she had destabilized shortly after landing. A petit mal seizure shook her body, testing the restraints of her cervical collar. The doctor traveling with them had wanted to shoot Valium into her drip to calm the event, but Lisa had urged restraint.
Valium could further depress her state of consciousness, Lisa had warned. Making any chance of communicating with her all the harder.
She had looked to Monk for guidance, offering him the option to call off this entire attempt. In the end, he had trusted Lisa and knew Kat would not want him to stop.
So, here they were.
Lisa crossed over to him, while Dr. Grant joined the techs. “We’re ready to go,” she said, eyeing him. “How are you holding up?”
“Let’s just get this done.” He nodded over to the neurologist. “What were you two talking about?”
She sighed. “Julian is concerned about Kat’s cerebral blood flow. Her systolic pressure is erratic.”
Monk knew the functional MRI test measured oxygenated blood flow into the brain. Any loss of pressure could cloud the results or cause the test to fail.
Lisa tried to reassure him. “But the MRI in the next room is one of the newest, the most advanced, with resolution down to a tenth of a millimeter. That’s ten times better than a hospital’s typical machine.”
And why they needed to come all the way to New Jersey.
Monk prayed it wasn’t all for nothing.
“Still—” Lisa began.
Monk noted the worrisome tone in her voice. “What? Tell me.”
“From the baseline scans—comparing to what Dr. Edmonds trans mitted earlier—the size of her brain contusion has increased. Only incrementally, but it’s still larger. Indicative that the lesion has begun to bleed again. Maybe due to the air pressure changes during the flight. Maybe from the small seizure.”
“Meaning she’s getting worse.”
Monk took a deep breath and held it.
Have I doomed Kat?
Lisa took his arm. “You know this is what she’d want.”
He tried to find solace in her words but failed. Still, he exhaled, saying, “What’s done is done.”
They stepped over to the control console. Through a window above the curve of monitors, a nurse stood beside the gantry bed that cradled Kat’s gowned body. He wished he could be in there, holding her hand. But due to the incredibly powerful magnetic field generated by the device, nothing metallic could be near it when it was activated. That included his prosthesis and the microelectrode arrays wired into Monk’s cortex.
“We’re all set,” one of the techs said.
Dr. Grant nodded. “Let’s begin.”
As the operators engaged the MRI, a heavy clanking of giant magnets echoed from the neighboring room. Dr. Grant leaned over one monitor as a grayscale image of Kat’s brain filled its screen.
The neurologist spoke without looking over. “There are three critical questions from here. Is the patient truly awake in there? Will she be able to hear us? And can she respond with enough vigor for the machine to register?”
Monk swallowed, praying all three answers were yes.
Or I put Kat in jeopardy for no reason.