Book Review: Nexus by Ramez Naam

Originally submitted to Scott Alexander's book review contest

What kind of science fiction are we living in today?

Ramez Naam’s 2012 novel Nexus is not the answer, but there are some parallels:

• A breakthrough technology in the 2020-30s has the potential to accelerate scientific progress and improve the human condition, but also to cause tremendous harm.

• Institutions are increasingly defined by how they balance technological progress with safety.

• A plucky group of developers in San Francisco recognize the risks but build a vastly more powerful version, believing that the benefits outweigh the harms because the good guys (themselves) will find the right safeguards.

• A leading developer of the technology named Ilya inexplicably disappears from the action.

• It is uncertain whether any of this is expected to be a net positive for humanity.

Naam believed that this change was coming. He said that the technological wonders of the next few decades would rival “even the greatest science fiction” and wrote Nexus to prepare readers for the moral and political questions ahead.

Nexus’ achievement is to do this without being reductive. It converts the “safety vs progress” debate into a deeply human narrative that steelmans both sides. I almost cried sympathizing with the pain technology unleashes on Nexus’ protagonists, but I also felt angry when ignorant fears destroyed their chance at a liberating future. Obviously these scenarios are fictional, but feelings are useful information to work through normative issues.

If you spend a lot of time thinking about how society will respond to radical technological change in the near term, you should read this book for its ability to bring these questions to life.  It looks at the fog-of-war of the future and, from a set of assumptions, charts how one possible pathway might feel. Novels can force authors to specify their assumptions and build a “verbal model” of their predictions. While Naam has written some futurist nonfiction, Nexus shines an emotional light on technology in a way that nonfiction just can’t. And while Naam eventually comes around to soft techno-optimism, the book leaves the net harms and benefits as an open question for the reader.

Despite this, Nexus is not our world. Naam was wrong. Wrong about which technologies would take off, who would develop them, how governments would react to these changes, and more. Nexus is a transhumanist story about brain-computer interfaces and biotech, not AI. The technology is developed by grad students in underground hacking communities, with corporations completely absent. The central villain is a supercharged version of the US Government, capable of not only passing constitutional-amendment-level tech regulations, but waging a War on Terror/War on Drugs inspired campaign to enforce them.

It’s fascinating to see an author take such a big swing— pouring his heart into a work of fiction to safeguard the future he thinks is possible— just to have his assumptions fall flat in hindsight. Naam was a futurist who missed AI. Nexus was published two months after AlexNet, but Naam was consistently bearish on AGI, stating that “AI progress has been 20 years away for 60 years,” or that the developers working on AI were “not the ones excited about it.” His own vision of exponential progress in brain-computer interfaces failed to materialize. Neuralink is doing exciting work today, but paralyzed patients first moved a cursor with their minds in 2006. It’s telling that Naam now works primarily on solar energy, where his predictions were more prescient, rather than human enhancement.

Science fiction does not have to predict the future to be valuable. But Naam really thought that Nexus would. The book has an afterword called “The Science of Nexus” making this argument. When asked if science fiction can predict the future in an interview, Naam cited Arthur C. Clarke predicting communication satellites and Jules Verne predicting and even inspiring the submarine. Then, when given a chance to waffle about his own prescience, Naam said that he hopes neuroscientists read his novel to get inspired about avenues of research.

Nexus shows some of the best and worst qualities of fiction as a medium to write about the near future. Charting a path through the fog-of-war is captivating (I couldn’t put the book down) and important for society, but because of this, compelling stories can become blinders. If you think a lot about AGI today, you could also read Nexus as a cautionary tale of getting too emotionally invested in one model of the future. It may be wise to emotionally hedge on a range of possible stories instead.

Lastly, Nexus is just a fun read. Noah Smith called it a “fast-paced thriller with plenty of gun battles and international plots and evil techno-gods vs. peaceful techno-hippie communes and other fun stuff.” This review will argue Nexus still has something to say about our own sci-fi timeline, but I hope it also convinces you to read it for its own sake.

Connecting Minds

In 2040, an illegal drug called Nexus 4 allows users to share information and experiences between minds. In underground raves and secret meditation groups, connection-seekers drink a silvery liquid filled with nanobots that bind to neurons in the brain, wirelessly sending and receiving signals. In response to high-profile terrorist bio-attacks and mind control cults, Congress has empowered an ‘Emerging Risks Directorate’ to wage a full-scale War on Nexus against even these benign use cases. This review will avoid spoilers past the first ~third of the book, though it’s clear throughout that Nexus’ potential is enormous.

The novel’s two protagonists meet at one of the secret “combining consciousnesses” rave/meditation parties. Kaden Lane, a neuroscience PhD student, hosts the party to test his friend group’s new invention, Nexus 5. This upgraded version allows users to run an operating system and write software on their brain to regulate emotions, download information, and more. Meanwhile, Samantha Cataranes, an undercover ERD agent, infiltrates the party to shut it down and bring Kade into custody. Sam embodies the “harms from technology” perspective, with a horrific backstory and fed mentality to contrast against Kade’s optimism.

Kade is an idealistic genius in over his head. Nexus 5 has the potential to change the world, but the book opens with him testing a module to flirt without being awkward and a group psychedelic meditation trip. Still, Kade and his friend group all see Nexus as a way to liberate humanity. One example is Watson Cole, a former US soldier who committed atrocities until he was held in a prisoner-of-war camp and forced to consume Nexus. It connected him with his victims and made him relive their traumas over and over from their perspective. This changed Wats forever, who converted to Buddhism and pledged to spread Nexus for its potential for empathy and peace. Because the technology is illegal, its developers are heroic underground hackers like this. They can’t take the first job offer from Google and corporations can’t shape the story.

In contrast to Kade, Sam is an extremely competent spy with transhumanist body modifications and a cynical worldview. She warns Kade of the unintended consequences and abuse that releasing Nexus 5 would cause. She’s reluctant to even try it, but as the agent with the most reason to distrust Nexus, the ERD chooses her to infiltrate Kade’s orbit. Although she telepathically reveals her identity by accident, Sam and the army of uber-DEA agents successfully capture and blackmail Kade into working for them.

The real drama of the story unfolds when Kade reluctantly agrees and travels to the sexiest sci-fi thriller setting: a machine learning and neuroscience conference. The US government is in a quasi-Cold War with China, who they believe is developing dangerous upgrades to Nexus, human cloning, and intelligence enhancement technologies. Kade and Sam’s mission is to infiltrate a top Chinese scientist’s research team to find out.

AI as we think of it today is basically absent from the story. Software-based AI is confined to narrow task-specific intelligences, like when Sam instructs a “daemon” to alert her if certain conditions in the hotel security system change, or Kade setting up essentially a Google Alert to monitor the web for keywords. Without spoiling specifics, the book is very interested in whole-brain emulation as a path to AGI, but this assumption makes even AI systems extensions of human agency.

Because of this, Nexus has very little to say about technological misalignment rather than misuse. Nexus has no will of its own and always works as a human intends, either its user or a hacker. The book’s catastrophic situations result from people using it badly, not from inherent technical flaws that make it inconsistent with human values. If you are only interested in this distinction as it relates to AI, skip to the section “Is Nexus Really ‘Post-Human’ Tech?”

Nexus’ moral arc is left with:

1. Technology empowers people to achieve their desires in ways that were not possible before.

2. Some people use it for selfish, evil, or ignorant purposes, which the government tries to stop.

3. Others want to use it for noble, altruistic, or harmless ends, but are barred from access.

4. It’s not clear how to enforce 2 without causing 3.

While our timeline debates these questions in ‘AI Governance’ papers and Marc Andreesen screeds rather than with guns, I think Nexus has a lot to say about this. Naam is at his best when he makes you sympathize with the father of a severely autistic child using Nexus to connect with his son, or the horror of a civilian having their mind/body hijacked in a terror attack. If you think AI has the potential to empower a lot of good and bad actors in the near future, Nexus may resonate with you.

Technological Freedom and Safety

The following argument comes from an early chapter of Nexus, not from an EA forum comment thread with six upvotes:

“How can you be so naive, Kade? You’re a good guy. I've felt that. But what about other people who get their hands on this?” [...]

“This is stupid. You can hurt people with guns. You can get them to do awful things with words. We need this. ‘Our current problems can’t be solved by the level of thinking that created them.’ Einstein said that. This can take us to a new level of thinking.”

“Kade, it’s going too fast. You’re talking about changing everything about people, the way we’ve been for a hundred thousand years, in a heartbeat. You can’t know the consequences, you can’t understand how people will abuse this, you can’t know that humanity will survive this.”

Nexus traces a standard technological “caution” vs “progress” debate. When characters argue for less regulation, they say things like “we don’t have to hand control of our lives to faceless bureaucrats and secret police,” or “broad dissemination and individual choice turn most technologies into a plus. If only the elites have access, it’s a dystopia.”

The book is weakest when it tries to directly reason with you like this. But fiction can do more— it’s one of the reasons I’m glad Scott prioritized non-nonfiction in the book review this year. When it works, a novel or even poetry can let readers defamiliarize settled points of view. Fiction might be the closest thing we have to ‘Nexus’ in real life for feeling the experiences of others.

Nexus is fundamentally about empathy. It’s full of examples of what living with emerging technology might feel like. I was moved when Sam experienced the mind of a child staring at the world with awe and wonder. Her own backstory gives her tremendous reasons to care for children, and re-experiencing these wholesome moments was healing for her. Similarly, Wats’ backstory of reliving his victims’ trauma makes the “empathy” angle quite literal. He calls himself the dark mirror of a bodhisattva, the opposite of an enlightened teacher, as his negative karma and risk of being reborn in hell allow him to liberate others.

As they bond throughout the story, Sam and Kade share memories and hold each other's minds in compassion. They relive each other’s formative moments to see their worldviews from the inside out. I wish I could experience this with my loved ones— to live the memories that made my family who they are, to see my partner’s internal world, to fully know and be known. Nexus’ protagonists imagine a world without miscommunication or hatred, where anyone can be understood if they want to be. 

The positive uses of technology in Nexus are inspiring. Researchers can collaborate telepathically while working on life saving drugs, climate tech, and more, working as one mind with enhanced memory and creativity. Parents can understand their mentally impaired children. Millions could be saved from neurological and mental health challenges. The technology is not widely circulated yet in Nexus, so its protagonists mostly dream about this potential.

When Kade meditates with a group of experienced monks using Nexus, he feels a deep equanimity and unity that could have taken years of practice to reach without it. His experience is liberating, with all attachments dissolving away into a single “now.” Naam writes:

“The minds in the room were a web, a tapestry, an orchestra of thought without thought. The room breathed in. The room breathed out. A thought occurred in the mind of a novice. It rippled across the mind of the room. All observed it. All brought attention back to the breath.”

Afterwards, a monk compares the scale of learning meditation to learning to read— a transformation of his relationship to the world as impactful as literacy. I think some serious meditators would agree with this characterization. The monk asks Kade what life would be like if learning to read took decades of practice, such that only a few experts could do it. He asks, if you could show people a faster way, would you? Even though it would be used throughout history for vile speech, to create tyrannical governments and wars, to empower the slave trade, and worse? Kade (and Ramez Naam) seem to say yes. The book argues that human empowerment results in more good than harm, but only tentatively, with serious trade-offs along the way.

Nexus’ moral arc is that voluntary human connection is intrinsically good. I say voluntary because, when it works, Nexus allows users to turn “off” their connection to others, not sending or receiving any signals that they want to avoid. Yet nefarious hackers find a way around these firewalls. Nexus becomes the highest possible stakes version of cybersecurity, with even a single vulnerability turning you into a puppet or worse. I suspect cybersecurity experts would never use Nexus for this reason. Every system has vulnerabilities, but the book’s techno-optimists seem to think they will figure out how to make Nexus secure… eventually?

They fail in all sorts of disastrous ways. State actors force-feed Nexus to turn people into suicide bombers against the communities they love. People experience deep privacy violations from malicious hackers and the government. The book has some pretty horrific sexual assault references of people losing mental autonomy through technology. Naam will introduce a charming side character that you start to care for, only to show how a bad guy with the right tech can destroy them:

“There were more screams coming from the living room. *** was still alive. He was on fire, burning to death. She felt every moment of it in her mind, felt *** collapse to her knees from smoke inhalation, her lungs burning.”

It’s hard to understand why anyone would risk using Nexus. But if the potential benefits are good enough, Naam thinks people would try despite the legal or technical risks. He argues that making it illegal just moves tech underground and removes the ability to ensure safety standards— a direct anti-War-on-Drugs mentality that may or may not have parallels to open-source AI. Naam argues that once the technology is widely released, there would be no going back to fight it on the supply side.

Some characters in Nexus go so far as to dream of becoming “posthuman,” transcending the limitations of human intelligence and mortality. The word comes up dozens of times in the novel, from regulators outlawing “posthuman” technology to scientists dreaming of a “posthuman” successor to the human race. Still, compared to the possibility of AGI, Nexus is surprisingly prosaic on this in hindsight. There is no non-human agency in the novel.

Is Nexus Really “Post-Human” Tech?

At one point, Naam gestures to the difficulty regular scientists would have understanding the work of posthuman minds. If Nexus users have enhanced memory, concentration, pattern recognition, creativity, etc, they may create updated versions of Nexus beyond normal human comprehension. Regulators could be unable to keep up. The US Government’s leading scientist even warns that “Nexus may defy human understanding because it is not the product of normal human thought.” 

Then this idea is basically… dropped? For the rest of the book? Despite their scientists refusing to use Nexus themselves, the government does learn how to enhance their agents with Nexus 5, including Sam. They use Kade’s tech against one of his friends in a way that he had failed to discover. No post-humans were required for this innovation. And while the developers using Nexus are more prolific, the book doesn’t seem at all interested in whether this could lead to exponential improvements or an intelligence explosion. The story centers around distribution and incremental improvements to Nexus 5, a potential blindspot in Naam’s worldview.

To be clear, not every book needs to be a “line go up” AI analogy, especially not in 2012, but recursive improvement is a weird omission in a story about enhancing intelligence. It fits with the novel’s broader disregard for technological risks beyond just a bad guy with access.

Nexus is complementary to human agency in a way that artificial intelligence might not be. The tech has no will of its own, cannot replicate or act without a human brain, and rarely even malfunctions. There is no pre-deployment safety standard— Kade sometimes pushes updates that have not been tested straight to prod on his own brain. Nexus is all about misuse, not misalignment. How alignment plays out in real life with AI will be an empirical question, but as a verbal model of one possible future, Nexus totally misses the possibility. Nexus may be disanalogous to AGI because it does not need to be carefully aligned to its users values.

Nexus is also not interested in narrative factors beyond human agency. The story is driven by its characters’ heroic choices rather than structural forces. There are no economic incentives pushing development in a certain way. PhD students just tinker out of passion, not even complaining about UCSF granting decisions. Nexus portrays geopolitical incentives between China and the United States naively, with the US disregarding economic growth and China disregarding social disruption from tech. Eventually, every major character in Nexus turns into a countercultural renegade against their institutions, rather than pulled into their orbit. Even the US government is portrayed as a cartoonish array of selfish generals and ambitious leaders, with a few edgy scientists resisting from within.

Don’t get me wrong, this makes for a gripping narrative! Human agency is really exciting. Naam has described Nexus as an “X-Men like story” of empowered individuals against a distrustful society, and notes that he chose to make the technology illegal in part because “it’s more fun.” I stayed up late reading Nexus to see what would happen. But all of this reveals why inherently misaligned technology impacting the world is hard to write about. 

AGI x-risk people are constantly critiqued for relying too much on science fiction, but I wish more authors would concretize their predictions in fiction. Even Naam, an explicitly futurist author who worked for “Singularity University” and wrote fiction about radical new technology coming after 2012, missed AI. Verbal models of the future can reveal an author’s underlying assumptions. I would like to read a near-future “man vs. shoggoth” story on Nexus’ emotional level. We don’t have too much “posthuman” near-future science fiction, we have too little.

Does Fiction Help Us Think About the Future?

It’s easy to forget what a strange and courageous thing it was to be a “futurist” in the early 2000s. Serious Journalists would scoff at claims that radical change was coming. The idea that a machine could “imitate intelligent human behavior” lived on websites that looked like this. Fiction was a way to talk about this stuff without getting weird looks. “It’s just a story.”

Now, it feels like everyone is a futurist. New York Times columnists have takes on scaling laws. My barber asked what my AI timeline was. Any moment, my grandma will be asking how tensor factorization impacts the robustness of episodic memory.

Naam was working through a set of “if…then” assumptions about the near future and taking their conclusions seriously at a time when this was hard. He took a big swing and even got the scale of the transformation right. Fiction allowed him to reach a wider audience, to bring an emotional depth that nonfiction couldn’t, and to concretize his assumptions. Did he get some things wrong? Absolutely. But Naam is humble about his mistakes. In a 2022 interview with Noah Smith, he said:

“As forecasters, it’s absolutely essential that we actually look at what we predicted and see what we got right and got wrong. [...] On human biology? On life extension? On brain augmentation? I was wrong.”

I was harsh about this, and about AI’s omission from Nexus, but science fiction is worth defending. It has taken a weird beating from the left for supposedly empowering right-wing billionaires and from the right for supposedly empowering decel fears. I’m not sure either of these things is really true, and I’d much rather have a series of weird possible futures to think about than to demand that it will be prosaic. More writers should build verbal models of the AGI near-future for their readers, ideally multiple.

Nexus is also still worth reading within the overton window of AI debates today. Most of the legislation proposed in Congress and the EU centers around harms from AI caused by humans. Even the “Frontier AI Regulation” paper written by everyone mostly focuses on risks to public safety via misuse, such as disinformation, cyberattacks, and biological weapons. The way Nexus makes you feel about these issues in general could be useful information for taking normative positions, or for thinking about how to persuade others.

I hope readers of this review give Nexus a try. I also hope it spurs a conversation about what other near-future fiction is worth reading. If disruptive change is likely to happen in my lifetime, what should I read to emotionally hedge on a bunch of possibilities? The Lifecycle of Software Objects? It Looks Like You’re Trying to Take Over the World? I’d love to hear some suggestions.