Neuralink and the Waking Dream

It was discovered by accident. 

While colleagues at the Soviet newspaper at which he worked took notes, Solomon Shereshevsky listened to his boss's assignments impassively. It didn't take long for his superior to chastise him. 

But Shereshevsky had a straightforward response: he didn't need to take notes. He could simply remember. We can imagine the interaction, in all its layered incredulity.

Newsman; round-collared shirt; munching cigar: Shereshevsky, my friend. Look around you. Do you see? The only one not to take notes. Get a pen.

Shereshevsky; heavy, foxed jacket; porthole glasses; electrocutee's startled coiffure*: A pen? Why bother?

Newsman: Well, so you might review your notes, of course. 

Shereshevsky: The hassle. With respect, I'll remember it. 

Newsman: Likely. You'll remember a thirty-minute speech? How about this? 

*holds newspaper aloft*

If I read you this article, front to back, I suppose you'll remember that too?

Shereshevsky: Yes. 

Yes. He could. 

While Shereshevsky's boss was shocked by his employee's gift for memory, in stupefaction as he recited the article verbatim, Shereshevsky was equally surprised that others did not share his ability. Couldn't everyone remember...everything? 

Not quite. He was sent straight from the office to Moscow's Academy of Communist Education. It was the start of Shereshevsky's second life, cataloged by neuropsychologist Alexander Luria. Starting on that April day in 1929, Luria would study the man he called only "S" for the next thirty years.

What he discovered challenged our notions of forgetting and remembering, of memory and imagination, of personality and selfhood. Nearly a century later, brain computer interface (BCI) companies like Elon Musk's Neuralink may be on the cusp of bringing the singular abilities and frailties of Shereshevsky to the rest of the world. 

The consequences may redefine what it means to be human.

The mind of S

Luria did his level-best to stump S. He asked him to remember a string of numbers and letters, phrases and sentences, passages in languages that he didn't speak. 

It didn't seem to matter. Whatever unfathomably long and mind-numbing concatenation Luria volleyed his direction, S would remember it. And not only would he remember it, he would hang on to the details for years afterward. In their very first meeting, Luria reported asking S to memorize a list of letters and numbers; fifteen years later, S could repeat them at the drop of a hat. 

"I simply had to admit that the capacity of his memory had no distinct limits," Luria concluded in his work The Mind of a Mnemonist.

If this sounds like a superpower — the storyline of Limitless set in Stalinist Russia — it wasn't. Luria deduced that much of S's ability derived from a particularly convoluted case of synesthesia. The stimulation of one sensory pathway — sound, for example — involuntarily aroused reactions from others. The ring of a bell not only had a particular tone but a taste, a smell, a feel, a color. Each sense was imbricated, entangled with others. 

That phenomenon, synesthesia, has influenced great art. Nabokov saw hues in words, perhaps explaining his technicolor prose; Kandinsky found color, hearing, touch, and smell to be intertwined. Even S's five-fold synesthesia occasionally had a rather charming bounce to it. To hear S describe how he perceived numbers reads like a story in and of itself: 

Take the number 1. This is a proud, well-built man; 2 is a high-spirited woman; 3 a gloomy person; 6 a man with a swollen foot; 7 a man with a moustache; 8 a very stout woman—a sack within a sack. As for the number 87, what I see is a fat woman and a man twirling his moustache.

That was the challenge of S's mind, though. For every word, every number, every sound he heeded was ballasted with dozens of inferences, sensations, feelings. Reading the newspaper while eating breakfast was impossible, with the flavor of words clashing against the taste of food. Abstract concepts like "nothingness" or "infinity" often seemed to be beyond him: what sense does "nothingness" have to someone who sees and feels it so concretely? Metaphors were a strange, tri-dimensional puzzle. 

There was the matter of forgetting, too, which S couldn't manage. Or, at least, not without real trouble. He wrote down things he wanted to forget and set them alight, hoping they would recede from memory as ink crumbled into ash. In later years, he was said to have turned to alcohol as a reliable inducement of oblivion.

Lastly, there was the problem of imagination. Usually considered a talent, it's hard not to read in S's case as a flaw. Studies by researchers like Daniel Schacter, Endel Tulving, and Elizabeth Loftus have suggested a link between memory and imagination; in simple terms, the former's fidelity influences the latter's richness. Since nearly all his experience was cataloged in ultra-high-dimensional detail, S's imagination competed with the vividness of real-life.

He is said to have been able to raise his heart rate just by pretending he was on a speeding train or elevate his temperature by visualizing placing his hand on a hot stove. As he told Luria, he often found himself lost in worlds of his making, missing a day of school because he'd pretended the clock had stopped and become absorbed in that new reality. To get through the tedium of workdays, he reported "splitting" himself, with one version of S going to work while the other dallied in a dream world. 

In the end, S did not seem to live a particularly happy life. Blessed as he was, in some respects, S's mental ability proved an impediment, making it almost impossible to read, to understand abstraction, to move through life. 

But what if we could smooth out the kinks, providing the luminescence of S's memory without the heavy freight of synesthesia? What if we could remember everything and still function? 

For some in the BCI space, that future is not just possible but inevitable. 

Building in a Jungle by the Sea

There has always been a harmonious meta-poetry in the field of neurology: the brain studies itself. But it was not until the mid-twentieth century that we would implant a chip into the chicanes of the cerebral cortex. 

"Madman" Jose Manuel Rodriguez Delgado was at the heart of this movement, taking advantage of the breezy regulatory climate to experiment on dogs, cats, chimpanzees, and humans. His experiments were as groundbreaking as they were theatrical: in 1965, he stopped a charging bull in its tracks by sending a radio signal to a device embedded in its brain. 

Though the starkness of Delgado's treatments resulted in a clouding of his reputation, he was, at heart, a technologist looking for better solutions to the lobotomies favored by his predecessors. Indeed, Delgado pioneered the use of "stimoceivers," small coin-sized devices that could stimulate electrical activity. This was a significant improvement from the wires that ran from brain to an external box, a system prone to infection. He would later pioneer work in non-invasive BCI, monitoring electrical signals from outside of the skull. 

Interspersed with press-friendly events like the Halting of the Bull were sleepier, clinically successful experiments. Deep brain stimulation (DBS) proved useful for sufferers of chronic pain and epilepsy, though Delgado did note side-effects in some instances. Patients' personalities would change after treatment in some cases, animated by sexual arousal, new energy, a happier disposition. That influenced Delgado's belief that character was "dynamic," not static.  

Alongside and after Delgado, others joined the field, bringing innovations in data collection, analysis, and experimentation. But progress has been slow, particularly compared to the pace of traditional computing. That is due in no small part to the more (reasonably) cautious regulations in the US and the hazards of working in the brain. One researcher described it as "a jungle by the sea"; muggy, salty, unseasonably warm. That has made it a harsh climate in which to implement technology. 

Extracting reliable data safely has traditionally been challenging and improved at a comparatively modest rate. If computing is governed by Moore's Law — which observes that the density of an integrated circuit doubles every two years — BCI is beholden to Stevenson's. 

Ian Stevenson and Konrad Kording proposed that the amount of trackable neurons in the brain doubles every 7.4 years. At that rate, the duo predicted it would take until the year 2231 before the brain's 100 billion neurons could be monitored simultaneously. That discrepancy in development speed goes some way to explaining why BCI has lagged other technological fields. 

Yet, the last five years have seen a slow flourishing with low fidelity products chasing consumer markets and more robust solutions attempting to solve complex health problems from paralysis to Parkinson's disease. 

We may have monopolies to thank for it.

Good monopolies and BCI's slow flourishing

In one of his irritatingly clever reversals that define Zero to One, Peter Thiel describes monopolies' benefit. In particular, he distinguishes between Monopoly-as-Rent-Seeker and Monopoly-as-Creator. While the former uses its entrenchment to feast on consumers' wallets until inevitable disruption, the latter deploys its formidable resources to unlock new areas of abundance, especially those tricky, capital-intensive innovations they are uniquely positioned to capitalize upon. 

Creative monopolists give customers more choices by adding entirely new categories of abundance to the world. Creative monopolies aren't just good for the rest of society; they're powerful engines for making it better...Monopolies drive progress because the promise of years or even decades of monopoly profits provides a powerful incentive to innovate. Then monopolies can keep innovating because profits enable them to make the long-term plans and to finance the ambitious research projects that firms locked in competition can't dream of. 

Perhaps BCI would have remained stuck in the domain of scholastic institutions, research organizations, and shadowy governmental groups were it not for the rise of the modern, mega-corporation. The last half-decade or so has seen Microsoft, Facebook, and "human monopoly" Elon Musk take an interest in the field, pursuing solutions. 

The growth to the venture capital asset class has also helped; funds are searching farther afield and betting on wilder technologies as they look to find an edge. Investments in BCI are still modest — roughly $100 million a year — but multiplying. (The government has also contributed hundreds of billions to its US BRAIN initiative). That's resulted in a gently bubbling ecosystem of players running the gamut from gimmicky consumer products to critical medical interventions. 

Though there's considerable complexity in almost all of these products, the starkest demarcation severing the industry is the matter of invasiveness. BCI is either invasive, meaning monitoring is done literally inside the brain, or non-invasive, meaning it is done outside. 

The difference is one of signal quality. 

Imagine a radio sitting in the other room. A song comes on that you have to listen to. Ideally, you'd like to hear this absolute jam with as much clarity as possible. But, because your roommate is a nefarious super-villain, you know there's a vanishingly small (but non-zero) chance the radio is equipped with C4 and might detonate at any time. So you have two choices, stay safely behind your thick bedroom door and listen to Adele's latest banger, "The Architecture of Regret," through the muffle of drywall, or risk a maiming and get right up close to hear it properly. 

This is more or less the distinction between invasive and non-invasive BCI. By entering the skull and cerebrum, you improve accuracy (you can hear the song clearly) but increase risk (you might blow something up). 

Beyond this rubicon, BCIs seek to achieve one or more of three primary objectives: augmentation, analysis, and treatment. There is plenty of overlap between these categories and reasonable muddiness. (How does one categorize an intervention aimed at mood disorders? Is that treatment or augmentation? How about energy levels or stress-reduction?) It's worth noting that tools in the first and third categories are usually "read/write" — capturing the brain's signals and influencing neural activity through external impulses — while "analysis" is mostly "read-only." 

As you might expect, those focused on pure augmentation take a non-invasive approach. Companies like Thync sell patches to provide electrical neurostimulation meant to increase productivity and energy. Halo Neuroscience also operated in a similar way before shutting down and selling its assets. Neurable, which pivoted from an initial focus on gaming, seems to be playing in this space, too, messaging an upcoming solution as a way to better manipulate virtual objects in a work environment. 

Dreem sits in between augmentation and analytics, leaning toward the latter camp. Via Professor Xavier-style headgear, Dreem analyzes its wearer's sleep, making recommendations off the back of its observations. It does not seem to improve sleep through any stimulus of its own, though. Emotiv positions itself as more of a true analytical tool, selling to individuals but targeting researchers. Like OpenBCI, Emotiv offers headsets and software to track and better understand neural data. 

Non-invasive treatments include those from companies like Flow (depression), Q30 (traumatic brain injury), BIOS (cardiac disease), and BrainCo (stress). Kernel, created by the founder of payment system Braintree, aspires to treat chronic pain, mental health issues, and stroke survivors, as well as alluding to plans aimed at improving focus, attention, and "cognitive load." 

Because of the inherent risks of operating in the brain, invasive BCIs are treatment-focused. Paradromic's device is first targeted at paralysis sufferers with speech and motor deficits. Per the company, this will be the "first high-bandwidth data interface between brains and computers," potentially providing a meaningful uplift along Stevenson's meridian. Neuropace treats seizures with its product. 

Finally, there is Neuralink. Founded by Elon Musk in 2016 and backed with $100 million of his own capital, Neuralink has ambitious plans to upend the BCI market, starting with acute treatment before moving mainstream. In time, Musk believes that invasive BCI will be sufficiently safe and desirable to attract a mass-market user. 

Already, Neuralink has made impressive strides (interspersed with bombastic announcements that have puzzled industry stalwarts), showcasing working tech in pigs. Earlier this month, during a Clubhouse conversation, Musk revealed the company's tiny filaments had been implanted into a chimp's brain with the animal able to mentally manipulate an external video game: 

We have a monkey with a wireless implant in their skull with tiny wires who can play video games with his mind. You can't see where the implant is and he's a happy monkey. We have the nicest monkey facilities in the world. We want them to play mind-Pong with each other.

In time, Musk believes Neuralink will enable superhuman powers. New skills could be downloaded straight to your brain, just like The Matrix movies, cognitive capacity could grow tenfold, telepathic communication would be trivial, memory would be infinite and infallible. Ultimately, the goal is to merge humans and AI, creating a new type of being and avoiding species extinction. 

"This is going to sound pretty weird, but ultimately, we will achieve symbiosis with artificial intelligence," he said. 

It may happen sooner than we think.

Inflections

In their 2011 paper, Stevenson and Kording suggested that if their law held, we would be able to simultaneously monitor 1,000 neurons by 2025. 

But Musk is good at contriving scale breaks; in Neuralink, he may have another. The company's current design involves 1,024 electrodes. While reports of the number of neurons this allows Neuralnk to monitor vary, the minimum amount mentioned is 1,000. Others suggest the number is much higher, potentially in the several thousand, representing a sharp uptick in monitoring potential. As one source noted, "it is roughly 15 times better than current systems embedded in humans."

With designs on expanding its BCI array to upwards of 3,000 electrodes, Neuralink may find itself monitoring 10,000 or more neurons in comparatively little time. Given what is possible with the manipulation of a few hundred — restoring limb movement, thwarting seizures, treating Parkinson's — the result might be profound. 

Encouraging, Musk is not alone in pushing data collection capabilities. Neuropixels, the product of Belgian company Imec, apparently monitors thousands of neurons, too. 

There are other reasons for optimism that I am not qualified to properly assess (I am already out on a limb). Late last year, Synchron received "breakthrough device designation" from the FDA for its "Stentrode" device. The Stentrode is a neurovascular product that is inserted into blood vessels. From that position, it can pick up brain impulses and allow sufferers of paralysis to manipulate external objects. The minimal invasiveness of the procedure and its fast-tracking put it in a promising position. 

Optogenetics, or the use of light to control neurons, is early but mind-bending. Dr. Yuste at Columbia has shown the efficacy of the technique on mice, infecting them with a virus that created a new "read/write" modality. One gene in the virus illuminates neurons when active, allowing Yuste to "read", while the other controls neuronal activity when stimulated by infrared light. With just a laser, Yuste is able to "write," inducing "hallucinations" that influence mouse behavior. "We manipulated the mouse like a puppet," Yuste said. 

Particulate selves

In my favorite movie of all time, The Great Beauty (La Grande Bellezza), the protagonist is the author of a book entitled The Human Apparatus. (I remember because it is a title of which I am intensely jealous.) 

It is in the world of BCI that we see that notion come to life: man not as machine, or machine alone, but mechanized, nonetheless; a fleshy, human apparatus; the brain a series of tethers and pulleys and links of energy and light. 

What are we when we become an apparatus? It would be tempting to say soulless; equipment is the class of the unfeeling and insentient. But I don't think that's right. If we can remember everything, if we can learn infinitely, if we can talk wordlessly, at which juncture is humanity lost? It seems likelier we will move in the other direction: becoming pan-sentient, all-feeling, effulgent, positively phosphorescent with emotion and sensation. We may, in short, become a lot more like S. Interpreters of a numinous experience, but somehow split or fractured, humans made fractal. 

Who will be in charge of that synthetic divinity, and what does it say about who we are? 

Dr. Yeste, the Man who Stares at Mice, believes we need a robust corpus of "neurorights" to protect against misuse. If a mouse can be played like a puppet, could a human? Yeste argues that neural data should be protected like other medical records and that we should not be able to self-incriminate with information gleaned from our brains. 

Another one of Yeste's concerns is BCI's effect on our personality. Jose Manuel Rodriguez Delgado suggested that "no human being has an inherent right to his own personality." The inference was that problematic personalities could be (and perhaps should be) altered. Who would make such a call? 

Already, some recipients of BCIs report an estrangement from themselves. Alongside a group of other researchers, Yeste shares a story: 

In a 2016 study, a man who had used a brain stimulator to treat his depression for seven years reported in a focus group that he began to wonder whether the way he was interacting with others — for example, saying something that, in retrospect, he thought was inappropriate — was due to the device, his depression or whether it reflected something deeper about himself. He said: "It blurs to the point where I'm not sure ... frankly, who I am."

For centuries, Buddhists have exhorted the absence of a self, some finite object carried reliably from one moment to the next. Is BCI the death of our delusion? If you and I can talk instantaneously without moving our lips, if we can subscribe to shared hallucinations, common personality modifications, equivalent skills, where do I end and you begin? It is easy to think of ourselves as fundamentally solid. Adaptable but defined, described, discrete; a tree, growing outwards, edging towards sunlight. It is soothing to feel this way, but it feels likely to become less and less true. We are not definite but particulate, a cloud of traits and characteristics, each malleable, the boundaries of your cumulus intersecting with my cirrus, selves in cross-pollination. 

Perhaps this, not monitoring neurons or treating patients, will be the most significant long-term challenge of BCI. Helping us live with what we are becoming.

______

*Since there are no surviving photos of Shereshevsky, I always imagine him as Luria, depicted here.