Brain Chips in Humans
Katrina Borthwick - 6th June 2023
On the 26th of May Elon Musk’s brain chip firm, Neuralink, announced that they had received FDA approval to launch their first in-human clinical study of a brain implanted device.
This is just after it was announced that Swiss researchers have already successfully used a brain implant to treat a 40 year old Dutch man who was paralysed in a cycling accident. The man can now walk just by thinking about it. The video posted online shows the patient, Gert-Jan Oskam, wearing a personalised headset and using a walking frame to assist him. He was previously confined to a wheelchair. The outcomes were published in the journal nature on 24 May:
Walking naturally after spinal cord injury using a brain–spine interface | Nature
The hope is that such devices can be used to overcome disabilities such as paralysis and blindness, and to help people with disabilities to use computers and technology. There is also talk of using it for non-degenerative conditions, as well as for mental illness.
However, on deeper reading, it appears Musk’s main aim is to help prevent humans being displaced by artificial intelligence (AI). And that is where it gets a bit weird. More on that later…
How does the Swiss experiment work?
For this experiment, electrodes in the brain monitor electrocorticographic (ECoG) signals in the sensorimotor cortex area of the brain, the part that controls motion. The implant itself is 4.5mm by 4.5mm, consisting of an 8x8 grid of 64 electrodes that pick up different directional activity. This hardware sits a circular titanium case with a similar thickness to the skull, and it is implanted in the patient’s skull. Anyone with a cochlear or BAHA implant will have a similar circle in their skull, so there are established protocols for doing this.
The patient then wears a headset – which looks a bit like ‘big’ headphones, but perched higher up on the skull. The headset has two antennas, one for power and one for picking up the ECoG signals. This all goes to a processing unit that decodes and tailors the signals into commands. This is experimental tech, so it’s currently a bit clunky looking, but should be able to be reduced in size in future.
The commands are then sent by wifi to an ACTIVA RC implantable pulse generator in the spine, which again is established tech, and is often used to deliver deep brain stimulation in patients with Parkinson’s disease. This device delivers targeted electrical currents to specific areas of the spine using special implantable paddles. From there the nerves stimulate the legs to move, as they normally would in someone without paralysis.
Obviously there is a lot of careful calibration required, and many muscles that need to be controlled, in order to make this possible - so the setup would need to be customised to each patient via software.
The Neuralink device operates on the same principles as this Swiss device, utilising wireless signals to replace the lost brain-spine connection.
The great merger – saving the human race
Musk has been trying to get his brain chips off the ground for a while. It started with him first setting up his brain chip firm Neuralink in March 2017, but since then it has been beset by delays.
The company tripped over some animal rights violations in his earlier testing, and Neuralink’s current FDA approval follows an earlier rejection on safety grounds in 2022, primarily related to the lithium battery, the potential for the tiny wires to migrate to other parts of the brain, and questions around the ability to remove the device without causing brain damage.
Before his initial launch of Neuralink in 2017, Musk said it would be ‘explained’ on the “Wait but Why” website. The article is live now, and it is a truly unusual read. It was written by Tim Urban (an internet writer and TED talker) and follows at great length the evolution of humans, and what he calls the ‘Human Colossus’ which is now teaching the ‘Computer Colossus’ to think. In other words, people are creating super intelligent AI.
The theory is that the Computer Colossus will either go bad, or simply take over. He says:
the Human Colossus is not going to quit until the Computer Colossus, one day, wakes up. This is happening….. This is what keeps Elon up at night. He sees it as only a matter of time before super intelligent AI rises up on this planet—and when that happens, he believes that it’s critical that we don’t end up as part of “everyone else.” That’s why, in a future world made up of AI and everyone else, he thinks we have only one good option:
To be AI.
Musk’s vision is one of having a digital layer that is like having a device in our head. For example, imagine having all the capabilities of an iPhone in your brain – augmented reality, memory access, searches, maps, smart home controls – plus, while we’re at it, the capabilities from other implants such as enhanced audio, muscle control, controlling Parkinsons or epilepsy, and health monitoring. Your brain is the computer, and the digital layer piggybacks off it, but it feels like part of you.
This may not feel as weird as we might expect. It isn’t that different to how the brain already operates. Conscious thought is confined to the prefrontal cortex, with the rest of the brain performing supporting activities that we don’t have to think about. We also know that humans have good neuroplasticity, in that with a bit of training we can often repurpose alternative signals to replace lost functions. An example I’m personally familiar with would be cochlear implants, where hearing is replaced by electrodes emitting electrical pulses directly into the cochlear to restore hearing. It takes a while for the brain to make sense of the new signals, but after a year a respectable level of hearing can usually be regained.
The challenge to this technology is ensuring that the speed of communication with the AI enhancements is not lagging too far behind the speed that our brain operates at. So long as there is lag, it will always be felt as separate to us – for example, typing into our phone’s search engine is a pretty slow form of communication compared to the signals that run back and forth in our brain. We can’t have symbiosis until we can get good bandwidth and low latency.
So, if that can be achieved, we may see the ‘great merge’ between computers and the brain, the ultimate counter to AI superintelligence:
“People will be a lot more powerful, which is scary, but like Elon said, if everyone is Superman, it’s harder for any one Superman to cause harm on a mass scale—there are lots of checks and balances. And we’re less likely to lose control of AI in general because the AI on the planet will be so widely distributed and varied in its goals.”
Neuralink and the Brain’s Magical Future (G-Rated Version) — Wait But Why
So are we going to become AI soon? And will this save the human race?
Humans’ abilities are more expansive than an AI’s, where an AI can only respond to the data that is available to it. Humans do not require a steady flow of externally provided data and can imagine, anticipate, feel, and judge changing situations, which allows them to shift from short-term to long-term concerns and navigate ambiguity. This is ‘authentic intelligence’. AI ‘artificial intelligence’ may be able to simulate ‘authentic intelligence’, for example a chatbot providing sympathetic responses, but that is not the same as having a true understanding or experience.
A third type of intelligence is ‘augmented intelligence’. That is what Musk is driving at, and is something that has been evolving in practice since 1998, when chess masters Kasparov and Veselin partnered with computers in their match against each other in Spain. The match was a draw. The theory behind this result is that the computers nullified the calculative and strategic advantages Kasparov normally held over Veselin. Kasparov then started approaching his matches differently, focusing on the strategic elements and offloading calculations to the computer. We do this every day with the devices we already have. When was the last time you actually memorised all your work appointments for the day, or did long division? Can you remember how to do long division? In some ways we are almost cyborgs, with our phones on us all day.
I’m not convinced that Musk will be leading the charge here though. He catches a lot of headlines, but as you can see from the Swiss study, he isn’t leading the charge on all things, and he often seems to underestimate the time and expertise it can take to move things into reality.
Musk’s approach seems to be a bit all over the place. He is trying to develop an alternative to compete with Open AI’s ChatGPT, but at the same time he’s calling out the potential harms of artificial intelligence. I’m sensing he doesn’t trust anyone with AI but himself. He also seems to have a ‘headline first’ approach, which contrasts with the scientific method and what we have seen from the Swiss research. They are the ones who actually managed to get someone to walk, using the same technology he is proposing to start testing on. But he has a bunch of money, so maybe he will hire them, or get there some other way.
My main concern is that people might be underestimating the amount of time and hard work needed to get to the kind of technology Musk describes. To get from measuring sensorimotor effects to capturing complex thoughts is a big leap. We can try to hijack the sensorimotor channel to do things connected to our thoughts (e.g. thinking about our hand movement to move a cursor), but where we still have actual hands to move it gets confused, and we aren’t achieving much.
Moving from motor function to pure thought is huge. We don’t have a good understanding of how to map out these complex signals, and I don’t believe we’re going to be in the space where a neural implant can pick up on complex thought and integrate it with the digital for decades. Will that be in time to beat the feared AI apocalypse? Who knows.