Brain machine interfacing, as someone who does research in the field and is getting a PhD in a brain machine interface lab, I think I’m qualified to comment on the progress neuralink. There’s a lot of hype out there, curing disease, ending paralysis, a world where we are part of the machine and the machine is part of us. Is it science fiction, or is there more to it?
Nuralink is brain machine interface research at the silicon valley pace. Promise the world and even if you fall short, well there’s always next time. It’s interesting to see the contrast between research we do in a lab and a company that uses the move fast and break things approach. Which… is kind of scary when you’re working with the brain, but maybe that’s just me.
In the lab if we make a claim we need data to back it up. Not just data, data that shows something to be true and is replicable. That last part, replicability, is the foundation of science. If I do something and it works, you better be able to do the same. If not, then there is a problem somewhere.
If my brain machine interface (BMI) were to cure paralysis for example, I better have demonstrated it at least in several animal studies at LEAST. A BMI for depression, same thing, where’s the proof? What evidence do I have that I have a technique that allows me to do this. As someone who is trying to pioneer a new technique I need a LOT of evidence to support my claim. It’s been a year long battle to get enough data to satisfy my peers and myself so that I can make any claim at all, much less something so grand as the ability to cure something.
Here’s the cold hard fact. The brain is hard. It’s the most complex biological system we know of today by far. There’s no comparison. We can barely understand brains of worms (feel free to play with the model yourself!), much less a dog, cat, or human.
In my lab we struggle to understand the language of the brain and I’ve written in the past about how hard that is to do. The brain is a beautiful neural symphony and while we can appreciate the beauty of its song, the notes and more importantly the meaning is (almost) completely lost on us.
BMI’s are still very rudimentary, crude, we may even say ugly things. They don’t always work correctly and even when they do they are painfully slow. Right now, in my lab I can control a exoskeleton using my brain, but it takes seconds of deeply concentrating for the BMI to recognize that I want to move… and I can only get it to start and stop. That’s the state of the art, cutting edge stuff.
Now the lab I’m in is non-invasive, so we really can move fast and break things without hurting anyone. They do have more invasive BMI that let you do things like type on a computer (we can do this non-invasively too, but invasive is somewhat faster and more accurate), it’s still far slower than you or I typing by hand and painfully slow for speech.
Now why I am explaining all of this? Well decoding is theoretically the easy part, we ask you to do something then we collect data on the changes in the brain before, during, and after you do the task. Collect enough of this data and eventually we can find a pattern, but that’s decoding.
Now say I have a magic machine that will end paralysis, it will read from the motor cortex of the brain your intent to move, then translate that information past the portion of the damaged spinal cord, deliver signals to the area after the damage and cause the spinal circuitry to do the thing we want it to do.
In order to do that, we would need to first have an accurate estimation of your intent to move and how (imagine how frustrating it would be if you wanted to step forward and your body took a left turn every once and awhile). Next we would need to translate that information to something the spinal cord can understand.
As I’ve mentioned in the past, the spinal cord speaks the language of the brain and the muscles, but the brain and muscles speak two very different languages. So now we have a whole new problem, we need to know the language of the spinal cord. Next we need to accurately simulate with the correct voltage and frequency the nerves that activate the circuitry in the spinal cord that creates movement.
Technically we could bypass the spinal cord all together and go straight to the muscles, but why? You would need tons of surgery to attach the electrodes to the individual nerves. That’s one way, now we want to send all the fun information that tells you where your body is in space, if it is hurt, if it is hot, if it is cold, etc. back to the brain. Then and only then do we have a fully realized way to end paralysis.
Going one way would be a disaster, you wouldn’t be able to tell without looking where your body is in space, you wouldn’t be able to balance, and most importantly you wouldn’t be able to tell if you were injured.
Where are we on the scale of start to end in this hypothetical? Well we’re at the starting line. We’re doing some of this (not very well mind you) in animals, but humans are a whole other level of complexity. I also just happen to research in a spinal cord injury lab and we can crudely activate the spinal circuitry right now if we wanted to. The problem is we can’t control it very well this is open-loop (as in we control it not your brain or any sort of feedback), much less have two way communication between the brain and spinal cord.
That isn’t to say that neuralink is a waste. I’m happy people are getting interested in neuroscience, neuroengineering, and BCI. We need and want people to be excited about all this. The thing that is the most interesting from my perspective is the thing getting minimal attention though, the sewing machine like tool that was designed to place the electrodes.
When we place microwires in the brain they tend to bend and kink. Think of trying to drive a overcooked piece of spaghetti into a meatloaf. It is difficult, but believe it or not, not impossible to do (think of hurricanes sending straws through trees) it’s a matter of speed really and they have achieved something very cool with this piece of tech, now the question is, can we use it on humans and how accurate is it?
There’s still a long ways away before we can have a BCI you or I would be interested in for fun, much less one that would be used to treat illness or paralysis. That isn’t to say that neuralink isn’t going to do something great for the field, that isn’t to say they haven’t done great things already. It’s just a friendly reminder that expectations should be kept at a minimum.
The brain is hard because it’s complicated. That’s not a bad thing though, if it wasn’t we wouldn’t be smart enough to try and figure it out. While a fully realized BCI isn’t going to happen tomorrow, there’s no harm in dreaming about what the future would be like as long as we remember it’s still the future and nowhere near the present.
What happens in the Lab doesn't have to stay in the Lab!