Elon Musk, chief executive officer of Tesla Inc., waves while departing court during the SolarCity trial in Wilmington, Delaware, last year. Photo / AP
At first glance, it seems like the right candidate needs only a range of standard office skills: "Effective multitasking"; "great verbal and written communication"; "attention to detail". But then the requirements become rather more demanding: "Understanding of Class III implantable neuromodulation devices." Not everyone, it is clear, will have what it takes to work for Elon Musk.
For this is a job ad just posted by Musk's five year-old company, Neuralink. And not just any job ad. This is not one of the usual tech openings for software engineers, or digital designers dreaming up products of the future. This is a job ad for a clinical trial director – for someone, in other words, to run tests on living humans of a product that's ready for surgical insertion now.
In a way, it should come as no surprise. Other job vacancies at Neuralink – animal care specialist; licensed veterinary technician – testify to the progress of experiments on monkeys. But that should not detract from the import of the moment. For though it is not as well known as Tesla and SpaceX, Musk's more celebrated electric car and rocket ventures, Neuralink is, if anything, even more ambitious, aiming to develop a device, known as a brain machine interface (BMI), that will ultimately plug our grey matter directly into the connected, digital world, allowing us to control devices, tap global knowledge, and even communicate with each other, purely with our thoughts.
No wonder that under the job ad's heading "What we offer", the very first line reads: "An opportunity to change the world."
Of course, playing with the brain – shocking it, switching it, slicing or dicing it – has been entrenched in medical mythology since Mary Shelley dreamed up Dr Frankenstein and his monster almost exactly two centuries ago. Even the idea of permanently inserted devices in the brain is not wholly new. For a quarter of a century, a company called Medtronic has been implanting deep brain stimulators into patients with Parkinson's disease to prevent tremors. But those are like pacemakers for the brain, therapeutic devices keeping its signals steady.
Today, however, a new breed of technologist, including Musk, dreams not of jolting the brain, but of merging it with our computer world, and so potentially allowing not just the sick to be healed, but also driving an astonishing augmentation of humanity's powers of cognition and communication. And if human trials go well, they will be a giant leap closer to making it come true.
Doing so would mark a profound breakthrough, for scientists have been trying to connect brains to computers for more than 50 years. In 1969, year of the Apollo mission, a young German-born neuroscientist called Eberhard Fetz pulled off his own moonshot by wiring a single neuron in a monkey's brain to a machine that fed the animal a treat if the neuron fired. Within two minutes, the monkey had grasped the process and was triggering the machine at will to tuck in, just by thinking. This century, implants have been used to help paralysed patients moved prostheses like robot arms with their thoughts alone.
But the kit can be clunky: the process of 'reading' the electrical signals in our brains inaccurate; 'decoding' them hard; and relaying them to whatever object we want to control, slow.
The 1982 thriller Firefox, starring Clint Eastwood, showed what might be possible. In that film, that object of thought control was a Soviet fighter jet which could be flown with the mind alone. According to leading researchers today, the film fired imaginations not just in the cinema audience, but in the US Army, too.
"It's interesting; that Firefox scenario has driven a lot of [mind control] projects in the military," says Andrew Schwartz, professor of neurobiology at the University of Pittsburgh, whose own lab has received funding from the US Defense Advanced Research Projects Agency (Darpa) in developing brain control of a robotic prosthetic arm. Schwartz is sceptical of whether a Firefox jet really would give pilots an advantage. But that hasn't stopped manufacturers investigating other mind-controlled machines.
In 2018, for example, Nissan announced a prototype with brain-to-vehicle technology, in which a car could tap directly into the driver's brain and act on signals to brake before the neuron message had reached the foot on the pedal. Nor did the military give up on such split-second advantages in communication. One researcher working on a BMI designed to capture words a person was merely thinking has described being approached by an officer from the US Army Research Office interested in creating a helmet which could silently beam unspoken thoughts from soldier to soldier on the battlefield.
Yet as the decades passed, the private sector – turned off by the idea of risky procedures to implant unproven technology in healthy people – remained wary. Funding was limited. Ambitious start-ups repeatedly went bust. But then, in 2016, Musk founded Neuralink, and everything seemed to change.
His company's BMI is far sleeker than the standard existing equipment, known as a Utah array, which consists of spiky silicon pads inserted surgically inside the skull, which read brain signals and feed them into sockets resting on top of the head, from where cables run them to computers that decode them.
Neuralink's device, by contrast, is near enough invisible, with 64 super-thin wires, each with 16 electrodes attached to 1,024 specific neurons inside the head, detecting their activity, then decoding them in a computer less than the size of a 2p coin, hidden under the skin, with the thought commands broadcast wirelessly.
You can watch the result for yourself, in a YouTube video that Neuralink posted last April of a rhesus monkey called Pager drinking banana milkshake while playing the classic arcade game Pong using thought control. That topped a 2020 demo in which a pig called Gertrude had been the star.
By then, Facebook had joined the party, announcing in 2017 that it too was working on a device that could read thoughts and turn them into sentences for "sharing" at the rate of a 100 words a minute. The difference was that it would do so without drilling into anyone's head, instead using a wearable skullcap. The project was led by Regina Dugan, a former boss at Darpa; the military agency also has ties to Neuralink, helping to fund a surgical "sewing machine" device that Musk's company plans to use to stitch its array of wires into the brain.
Increasingly, though, it is venture capitalists, not men and women in uniform, who are providing the financial backing to drive BMIs. "Because Elon Musk was ready to put in $100m of his own money, a lot of people in the venture capital world were willing to follow in his footsteps," says Schwartz. Many bet against Tesla and regret it now, having missed one of the most profitable investments of the century. As a result, says Professor Krishna Shenoy, professor of bioengineering at Stanford University, BMIs now enjoy "nearly infinite resources".
Hundreds of millions have poured into the sector in the last year alone. It seems that a tipping point has been reached.
The majority of BMI projects remain for therapeutic purposes – to restore sight to the blind by feeding signals from a camera directly to the brain, to give speech to those without by reading their thoughts and broadcasting them, or to help the paralysed interact with the world again for themselves, by moving robotic prosthetics or computer cursors with their mind alone. For Nathan Copeland, who was paralysed after a car accident, and who has collaborated with Schwartz's colleagues, the BMI is like a mouse, which allows him to draw using computer software, or play video games. As such games become ever more immersive, the ability to interact with them provides paralysed people with real opportunities to escape their condition, if only virtually.
But Musk has repeatedly claimed an even higher purpose for Neuralink – the aim of elevating man's abilities by fusing with the computer world. In this vision, BMIs will allow us to compete with super-intelligent robots, which in future might otherwise destroy us. The company's mission statement? "If you can't beat them, join them."
Of course there are medical risks, notably the not-negligible risks of infection. But the main concern for many scientists is the ethical risk – what privacy is there in a world in which a Facebook device can read your thoughts? Then there is the potential for hacking or tricking BMIs, to make them perceive objects that are not there, or control devices they are not meant to. Researchers have described exploiting such flaws to "run [mice] like puppets" or get one monkey to control the arm of another.
The biggest risk, however, may be that reality does not meet the hype, at least at the wilder speculation. Facebook ended up shelving its headset project (if only to focus on another, wrist-worn brain-reading device). And while the gadgetry gets cooler and more slimmer and more marketable, some of the fundamental neuroscience remains just really, really hard.
"There's an engineering arrogance, they feel they can do just about anything given the right technology," says Schwartz. He says Neuralink's team is focused on making its device sleek and small. "And I think they can do a beautiful job with that. But that's the easy part. The hard part is understanding the signals in the brain, interpreting what they mean."
Brain signals that control movement are relatively straightforward to decode, which is why so many existing BMI projects focus on moving prosthetics or cursors. "That's why for paralysis, stroke patients, for medical use, it's very reasonable to think [BMI implants] will happen in the next five to 10 years."
But decoding complex thoughts, or merging with the internet, "is way off", says Schwartz. "It remains in the realm of science fiction." For which there will surely be some, recalling the way Shelley's Frankenstein turned out, who will be profoundly grateful.
The S&P/NZX 50 Index closed down 0.10%, falling to 12,627.32.