by Chip Walter
Already the impossible had happened in 2018, when the scientific world learned that Chinese scientist He Jiankui had used Crispr to alter embryos that were then implanted in a woman who later bore twin girls. He did this entirely on his own. His goal, he said, was to eliminate genes that could cause the newborns to get HIV, because the children’s father is HIV-positive. But the fact that the scientist had independently and forever changed the DNA of living humans appalled scientists and bioethicists.18 If one gene like this could be edited, what would stop rogue scientists from creating someone’s idea of a superbaby? Or unwittingly developing a baby with a new and unknown disease?
The fear was that this kind of gene editing was moving too fast. Scientists were still trying to suss out the ways stem cells reacted to the different bone or liver or heart cells they might replace. How long were they effective? Which cells triggered what signals? These explorations took time, and they were complicated. It wasn’t like changing a carburetor. And of course, there was the problem of the Big Black Box, the brain, which, for reasons only evolution could fathom, did not generate large reservoirs of stem cells like many other parts of the anatomy.
And what if older neurons were replaced wholesale with new stem cells? They might scramble different sectors of the brain by destroying the new connections between the originals. Fiddle with those, and who knew what mayhem might follow? Memories, learning, and other cerebral functions that the brain had grown accustomed to might simply vanish. On the other hand, in the case of a disease like Alzheimer’s, maybe new memories would be better than no memories at all.
One age-related affliction particularly intrigued Hariri, and he suspected curing it could slow death’s march pretty quickly: sarcopenia. Sarcopenia was a disease that accelerated the loss of muscle in some people before their time. Since the FDA defined it as a disease, clinical trials were possible. Furthermore, anyone who was aging—and that was everyone—suffered from a form of sarcopenia. It was simply what happened as time passed. Hariri had found that by the time the average Homo sapiens turns 25, muscle mass begins to drop off at the rate of 1 to 2 percent a year. By age 60, half of it is gone.
One might think that this loss would only affect a body’s strength and steadiness, and that would be true. But there are other factors as well. It turns out that most of the blood in the body resides in muscle. Its high venous capacitance is essential to long-term health, because venous capacitance is what delivers the goods to your immune system. So now a 60-year-old human hasn’t only lost half of his muscle mass, but also half of his venous capacity and immune system. This means that anyone over 25 has begun suffering from a chronic form of sarcopenia—and though you can slow the process with proper diet and exercise, it inevitably continues. There’s no way around it. In Hariri’s mind, placental stem cells offered a perfect cure, because the cells could regenerate muscle. That would, in turn, reduce fat, renovate the vascular and immune systems, and generally reverse aging. One 2009 study even disclosed that increasing muscle lowered the likelihood of cancer.19
Would a world brimming with baby boomers find a treatment for sarcopenia appealing? Forget worrying about ageism or flabby arms, compromised immune systems, or the sad smiles of people as they gazed at the elderly couple over there with their walkers, thinking, Aw, isn’t that sweet? Because that couple, even with lots of years in the bank, would not need walkers anymore. They would be vigorous, healthy, upright, and perfectly capable of pulling their weight in the real world. One hundred could become the new 60! That, at least, was how Hariri and Diamandis saw it playing out.
But right now, all of that was still science fiction. No one was going to be biologically “topped off” with stem cells and rejuvenated just yet. Nevertheless, Hariri and Diamandis—the two Celularity founders—had high hopes. The company was already developing several clinical trials to move as quickly as possible toward FDA approval. Maybe stem cells would become the holy grail of radical longevity, and thousand-year life spans would abide as the human race broke its evolutionary bonds. Who could say? All you could do was try. As one scientist put it, “The science will go where the science will go.”
26 | THE SEED OF THE SINGULARITY
If the science necessary to solve aging was going to go anywhere, one last remarkable and ironic piece of the longevity puzzle would have to fall into place. Smart machines would need to arise in defense of the human race. Already machine learning was embedding itself in the medical arts, and digital technology had long ago become science’s handmaiden. Venter’s work with the Human Genome Project had marked a milestone. But now, as the search for immortality deepened, much more digital muscle would be necessary.
Art Levinson himself had put the facts concisely: When it came to flipping the genetic switches needed to evade aging, there was no way any human at a lab bench —no matter how gifted, how insightful, or how hardworking—could possibly locate and comprehend their magical pathways. And without that, curing the Ultimate Problem was simply not going to happen. Homo sapiens required a tool that was faster, smarter, and more tireless than humans themselves. The kind that Riccardo Sabatini had used in the Face Project made a good example. “Machine learning” was one term that Sabatini and other computer scientists used to describe this brand of work, but there was another more common name that nearly everyone had heard of—artificial intelligence (AI).
AI is different from other forms of computer code. It consists of legions of algorithms that are eerily similar to the human mind itself: AI can learn to solve problems without being explicitly told what to do ahead of time. It can, in some ways, think for itself, at high speed. It’s the stuff of Terminator-style invasions and countless dystopian futures. Ironic, then, that such capabilities should now emerge as our saviors. It’s doubly ironic that those same tools have been the source of so much of Silicon Valley’s wealth. It’s almost as if evolution itself had somehow anointed the Valley—with all its computing power and money—as the chosen instrument for immunizing death. This creates, in the machines and their algorithms, a kind of symbiosis: digits and molecules, biology and technology coming together in a strange and unexpected harmony.
* * *
—
RAY KURZWEIL COULD HAVE told you this was going to happen. More than 50 years ago, when he was 14 years old, he wrote a paper that outlined how a machine might somehow become as intelligent as a human. That was even before he had landed a spot in the Westinghouse Science Talent Search and shaken the hand of President Johnson. He hadn’t yet divined a direct connection between artificial intelligence and longevity just yet, but he always fervently believed that truly intelligent machines could solve nearly any problem.
The essentials of that thinking hadn’t really changed since Kurzweil’s paper. In fact, he used much of it as the basis for his best-selling 2012 book, How to Create a Mind. The book argued that human-level intelligence could be created in computers by reverse engineering the human brain. Figure out how the neocortex worked, employ pioneering software and hardware to do the same in a computer, and voilà! A fully humanlike, but entirely artificial, machine.
Just after the book came out, Larry Page suggested that Kurzweil join Google to “bring natural language understanding” to the company—figure out, in other words, how computers might someday talk and communicate just like humans. Initially, Kurzweil had only planned to ask Page if Google, or Bill Maris’s Google Ventures, might like to invest in the business he wanted to create based on the book. But, Page said, just come into the Google fold. This way, Kurzweil could work with the canny computer scientists at Google and tap into its bountiful digital resources—not to mention free office space and all the hardware and software cycles a big thinker could ask for.
So in December 2012, Kurzweil, for the first time in his life, joined a company that didn’t have his own name on the corporate logo. But that was okay. The dream of creating something as remarkable as a virtual mind—the holy grail of artificial intelligence—was deep in the
man’s DNA. If he had to become an employee to solve the world’s problems, including death, he could live with that.
The team’s first endeavor under Kurzweil’s tutelage as a director of engineering was to create machine-learning algorithms that could understand users’ emails and then provide short but sensible answers, all on their own. This turned out to be tougher than almost anyone thought.
It took nearly five years of work with a group of 35 scientists before Team Kurzweil created its first Google product: Smart Reply, a Gmail mobile app. The initial version, launched in May 2017, listened to the email you received, and then Smart Reply provided three possible answers, short responses like “Let’s do Monday,” or “Yay! Awesome!”
Smart Reply wasn’t anything that was going to give Skynet or the Terminator a run for its money, nor was it going to solve aging—not immediately. But in Kurzweil’s view, it made a good first example of artificially intelligent software comprehending a human thought and then providing a response that made sense. On the surface, it might appear trivial, but it really wasn’t. And in the end it would lead to life everlasting. How?
Building on Smart Reply, Kurzweil planned to ratchet up his project to the point where machines could, on the fly and in context, speak as fluently in any language as he, or anyone else, could. The new version would be able to pull all the right words in all the right order out of thin air, and carry on an entirely sensible, humanlike conversation. Once that was possible, he figured the machines would be pretty much as smart as we were, which also made them the seed of the Singularity he felt would arrive in the mid-21st century.
The seminal concept behind Kurzweil’s work was something he called intelligent pattern recognizers—layers and layers of them that reside in the brain. In his view, these modules were what had made the Homo sapiens neocortex—the most recently evolved sector of the human brain—such a ringing success. Kurzweil estimated the cerebral cortex houses about 300 million of them, each consisting of clusters of neurons. Placed in context, he held that these modules rapidly bootstrap simple concepts in an increasingly complex human hierarchy that, layer by layer, delivers remarkable insights like art, mathematics, and language. The modules manage this by quickly identifying a few low-level cues, then sensibly pull in more modules to generate still more bootstrapped knowledge.
For example, a module that sees the visual image of a horizontal bar and then sees two sides of a pyramid would, in the general context of a sentence, immediately recognize it as an A. Other related modules would see additional letters related to A, to piece together the word “Apple,” rather than, say, “Pear.” More modules would attach additional words and then tastes: maybe the smell of pies, memories, a location that the modules then figure out is a kitchen—until the next thing you know, you’re craving a piece of your grandmother’s apple pie, just out of the oven. This, in turn might trigger all sorts of other thoughts, feelings, memories, and insights. All of this happens in a blink, powered by the brain’s hundred billion interconnected neurons.
This might seem an exceedingly simple example, and a long way from Shakespeare’s “Tomorrow, and tomorrow, and tomorrow / creeps in this petty pace from day to day…” But for Kurzweil, the point was that this network of interwoven, highly flexible modules supplied the wellspring of human intelligence. And his goal was to develop the artificially intelligent software that could reverse engineer this unique human trait.
If such an advance were possible, it might not be immediately obvious how artificial intelligence would lead to immortality. But to Kurzweil, it was all of a piece. With the advent of AI, he foresaw the evolution of a newer, far more powerful version of the human body and mind: one that wasn’t strictly biological, but instead employed nanotechnology, cell-size nanobots, that could clean out arteries, strengthen muscle, and boost organs while simultaneously allowing the brains of mere mortals to access the vast cerebral spaces of the Cloud. But not in the way we do now, with clunky phones and iPads, but with invisible, cell-size machines injected like serums into the cerebral cortex, essentially becoming enhanced, artificial brain cells, something I found myself calling neurobots.
Within decades, Kurzweil predicts, millions of people will be physically invincible, supplied with trillions of neurobots capable of linking directly to the ubiquitous Cloud. Anyone thus augmented will not require stem cell rejuvenation, or even revamped genetics. They wouldn’t need to ask Google which Michael Keaton movie won an Academy Award nomination in 2014; the answer would simply be there, available like every other memory. The average human would not watch a movie; she would be immersed, imagining it more completely than our own recollections currently do. One would not hum a song; the music would come into the mind, full-blown, in the highest possible fidelity. In a blink, neurobots will even be able to shift your reality from wherever your body is currently located into any other place you might like: Kathmandu, ancient Rome, or a beach in the Seychelles, complete with warm sun and crystal clear water lapping your toes, every sensation as real as real. It wouldn’t be real, but it would feel that way, thanks to a seamless, sensory melding of the neurobots rearranging the chemistry in your brain.
Best of all, this new human hybrid could be digitally backed up and then downloaded to create a cloned copy containing all of the information in your mind and body, so that even if your “self” suddenly died, you would have a perfect backup available to resume life as if nothing at all had happened—true immortality that would, once and for all, absolutely obliterate Gompertz’s beta.
Kurzweil considered this a fourth and final bridge beyond his previous three, the ones he and Terry Grossman had envisioned together in the early years of the 21st century. With it, his ultimate view of everlasting life would at last emerge at a time and place that didn’t simply upgrade old-fashioned biology—the kind Calico and HLI were working on—but upgraded it with nanotechnology that made you immortal and incredibly intelligent, almost godlike. Kurzweil, of course, would never use the term “godlike.” To him, entwining humans and machines so thoroughly that they became indistinguishable was simply the next natural course of human evolution.
* * *
—
ONE MIGHT FEEL that Kurzweil’s Bridge Four thinking was just a touch outside the views of the average Homo sapiens. Some, however, felt it was a very real threat. Elon Musk and, prior to his death in 2018, Stephen Hawking, had warned that superintelligent AIs could take over the planet—partly thanks to the work Musk’s friend Larry Page was supporting. “I have exposure to very cutting-edge AI,” Musk told attendees at the National Governors Association in July 2017, “and I think people should be really concerned about it.”
Earlier, Hawking had written in an open letter with Musk and a few dozen other artificial intelligence experts that the emergence of AI would lead to creatures so smart and swift they would leave us looking like the cerebral equivalent of an amoeba. It could, he said, become the “worst event in the history of our civilization.”
Remarks like these aggravated the Kurzweilian brain. More and more, he grew peevish with media cries that repeatedly told the world that in no time, we’d all be living in a dystopian future where our overlords transformed Siri into some menacing version of George Orwell’s 1984. But look how technology had advanced the human race! Despite the horrors of the last century, the rate of death caused by war over the past 600 years had dropped several hundred fold. Murder rates were rapidly declining. FBI statistics showed that between 1993 and 2015, the U.S. murder rate had plummeted 50 percent. The same was true of property crime. Despite media reports of our collective demise, Kurzweil believed, the world was a better, safer, happier, and smarter place, mostly thanks to the advances that the keepers of science and innovation made possible.
For Kurzweil, the smart thing was to let technology march ahead, Tom Swift-style, because that was where we were headed. It was all right there in LOAR (the Law of Accelerating Returns). Yes, one had to be vigilant and control the power of smar
t machines. He had been saying that for years. But no need to hit the panic button. Machines wouldn’t match human intelligence for another 10 years, and the Singularity itself wouldn’t arrive until 2045, a date with destiny that he planned to keep, when he celebrated his 97th birthday. The best approach would be to put safety measures into place along the lines of Isaac Asimov’s “Three Laws of Robotics.”20 Like the first stone knives, created over two million years ago, all technologies could be used for good or ill. But if properly managed, artificial intelligence would surely be our saviors, not our terminators—our partners, not our competitors. Just watch: AI was going to save our skins. Kurzweil could see it. Levinson and Venter saw it too, each in their own way. There could be no doubt: Smart machines was where the end of The End lay.
27 | THE NEW ORACLES
In early 2016, when she became Calico Labs’ chief computing officer, the views and aspirations of Raymond Kurzweil did not inhabit the mind of Daphne Koller. This wasn’t because she shunned Kurzweil, or digits, algorithms, or computer code; far from it. She loved them all. It was just that Kurzweil wasn’t her favorite when it came to the stratagems of machine learning and artificial intelligence. Bridge Four wasn’t her thing. Her inspiration was Thomas Bayes, an 18th-century mathematician and Presbyterian minister. She had explored his work and theories in the 1990s, and then used them to develop some of the most advanced artificial intelligence algorithms in the world.