I’ve seen this movie before…
Neuralink appears to be the first practical step towards a Ghost in the Shell type world where our brains can be plugged in and hacked. Ghost in the Shell is a cyberpunk masterpiece. I lowkey place The Matrix in the same world as Ghost in the Shell, just a couple decades ahead. In The Matrix, machines can plug into our brains and give us a complete VR simulation world. In Ghost in the Shell you can dive into more abstract cyberscapes to do your hacking and social media. Perhaps most impressively, The Matrix postulates that you could learn anything in a matter of seconds.
Elon Musk has been careful not to make any such grandiose promises, except on Joe Rogan. That being said, he has let slip some extraordinary claims. Still, I’m not here to debate the man, I’m simply going to write my own predictions for this technology. Here’s a non-exhaustive list of movies that immediately come to mind:
- The Matrix
- Ghost in the Shell
- Surrogates
Rehabilitation and Diagnostics
Rehab is probably going to be the top use for Neuralink. The human brain is just too big and too complicated for a few thousand electrical probes to get a good picture of what’s going on. We’ve had people controlling computer cursors and other very basic tasks with brain-machine-interfaces (BMIs) for quite a few years now. I think that Neuralink will, at best, make such technology more accessible and sustainable.
Will it help restore sight and sound to the blind and deaf? Certainly not the early versions. Will it help restore motor function to paralysis and ALS victims? That’s probably even farther away.
Sensing and interpreting output is one thing. Every brain is different and so you’ll need a lot of data and computational power outside of the Neuralink implant to even make sense of a lot of the output. It’s simply not possible to jam enough computer power into a quarter-sized implant to do that. I am skeptical if phones even have enough horsepower to do this. I suspect that you’ll need to integrate with cloud services, and more powerful computers, to get many of the benefits Elon has suggested.
However!
There would be a huge value to everyone walking around with high quality sensors in their heads (as well as the rest of their bodies). Stroke, aneurysm, dementia, depression, PTSD - all these things could be detected very early. The impact on the quality of life for people by avoiding suffering would be phenomenal. I suspect that Neuralink will ultimately find a lot of value by integrating with the ENS (Enteric Nervous System) as well. The brain-gut-axis is turning out to be extraordinarily important for mental and physical health, carrying huge implications for chronic illness.
High Bandwidth BMI
I seriously doubt this will come about. Here are some reasons why.
Limits of the human brain
It can take us a while to formualte cogent, cohesive thoughts. We recruit specific regions of the brain when we intend to speak, and when we are listening and reading. Some of these regions are quite large, far too large for Neuralink to adequately survey. Elon says that writing, reading, speaking, and listening are very low-bandwidth. From a strictly numerical standpoint, I agree. But I also think that humans can only get so much faster. We have evolved to ingest general purpose information by listening to other humans. We simply lack the hardware and software to ingest data through other means. Even our reading and writing is just a facsimile of our speech, representing the sounds of speech with visual symbols.
Sure, you can watch a lecture at 2x speed and still get most of it. We even have multimodal learning, where you see, do, hear, and practice all at once. It’s simply a fact that human brains take a while to acquire new skills and knowledge.
Neuralink would have to fundamentally alter the way the human brain works - which I don’t think it will do with its first iterations. I would absolutely LOVE to be wrong. Maybe using Neuralink is a skill, a talent that we will have to develop. Perhaps we will have to learn to communicate with and through the device. Maybe it will ultimately be faster. It would be wonderful if I could type up this blog without a keyboard.
Still, I think there are fundamental limits to how fast the human brain can assimilate and integrate new information. I think those limits are biological and won’t be changed by a tiny implant.
Sorry Neo, no helicopter program for you.
Computer horsepower
The human brain’s processing power is roughly the equivalent of a 1 petaflop computer. It’s a really bold claim to say that a pocket-sized machine could communicate with that faster and more effectively than it already can. The one caveat is if the power of the human brain is actually what the Neuralink device relies on. Still, I’m highly skeptical.
At full power, my Pixel 4 operates at 954 GFLOPS, just shy of 1 TFLOPS, which is 1000x less powerful than my brain.
Why do I use this comparison?
I anticipate that Neuralink will have to build a model unique to everyone to essentially simulate their brains in order to calculate the exact pattern of neural stimulation required to communicate with us at a high bandwidth. In principle, I do believe that Ghost in the Shell and The Matrix technologies are possible. The human brain relies solely on neural impulses for all input and output (IO). We know this for a fact.
When you want a human brain to learn something, you want to change the state of neurons, their individual “memory” in the form of synaptic connections. There’s an initial state and an end state. We will probably need to be able to simulate or model a large portion of your human brain in order to communicate quickly and effectively with just you. If Moore’s law holds, then it will be 20 years before a smartphone reaches the petaflop mark. That’s 2040, which will be approaching Singularity anyways. If the overall goal is to prevent an AI holocaust, that’s likely to be too late.
Direct sensory stimulation
While I do believe the high bandwidth BMI is 20 years away, I think we could see direct audio and visual stimulation much sooner. Paradoxically, the part of the brain that handles optical processing is at the back of the head. This makes it an easy place to drop some electrodes. Perhaps this is why Neo jacks in at the back of his head? In Ghost in the Shell the connectors are on the back of the neck, closer to the brain stem.
This could ultimately allow for hardware sensors being integrated, used to replace defective or missing eyes and ears. Wouldn’t that be cool! You can upgrade your eyes to have telescopic and thermal vision! You could hear better than cats and dogs. Maybe you’ll be able to relay extra sensory information via BlueTooth on your phone.
In this way, I think Neuralink is far more likely to result in some cool augmented reality abilities rather than full VR. At least for the foreseeable future.
Telepresence robotics
Did you ever see the creepy Bruce Willis movie Surrogates? Basically, I think we’re far more likely to see that sort of thing than anything else. Elon Musk already demonstrated the ability to detect motor movements from neural signals. With motor output and sensory input, you could remotely inhabit any machine, not just human-like ones. In that movie, a bunch of soldiers are sitting in a call-center, remotely piloting battle mechs.
The movie was really cool in the premise but the villain was very 1980’s. Spoiler alert: He’s going to take over the world by enslaving everyone with his telepresence robotics technology! Bwahahaha! Yeah, that was dumb.
I recently played through a game called Lone Echo where you occupy the perspective of a robot working on the outside of a space mining station. I think that Neuralink could enable that sort of remote work for humans. Dr. Who even had an episode about this, except the telepresence robots were learning polymer goo that became sentient and rebelled, killing the workers.