Science Corp. Aims to Plant Ideas in Brains with New Device
Monkeys with GPUs coming soon to a casino near you
One of my favorite books as a kid was Mrs. Frisby and the Rats of NIMH. It tells the story of a group of rodents who were experimented on in a lab and gained profound intelligence and physical gifts. These rats can read and build machines, and they live much longer than the rest of their kind. They, of course, escape the lab that created them and go on to build a flourishing, human-like society underground in the countryside.
If Max Hodak and the researchers at Science Corp. have their way, these types of clever rats – or perhaps something along the lines of monkeys playing poker - might be brought into existence. Oh, and also, Hodak thinks humans could be heading toward an Inception-like future. Let’s get real weird.
Near the end of last year, Science revealed its work on technology that makes it possible to fuse large quantities of lab-grown neurons with an animal’s brain. To do this, Science has built a device that preserves the manufactured neurons in a gel. It then takes out part of an animal’s skull and places its device atop the animal’s brain. In the days that follow, the neurons in the device begin to develop wiring that stretches out from Science’s hardware and into the brain, giving the animal access to extra stores of mental horsepower.
Hodak bills the technology as a new kind of brain interface that builds on what the public has been reading about for the past few years from the likes of Neuralink, Synchron, Precision Neuroscience, and Science itself. Unlike many of these existing brain interfaces, though, Science’s latest technology would not require an implant being placed deep into brain tissue or genetic modifications to a brain to work. To that end, Hodak sees these “biohybrid neural interfaces” as less damaging to brains than current technology. He also contends that Science’s technology will be more general-purpose and reach many areas of the brain, as opposed to current implants that target specific regions like the motor cortex or visual cortex to augment specific mental functions and combat injuries and illness.
“You could place electrodes into the brain and get more bandwidth,” Hodak says. “You could place electrodes in the blood vessels and get less bandwidth. You could do things in between. You could use ultrasound, but that requires genetic modifications to get very high functional resolution. All of these things have complex trade-offs.
“You're going to come out with a motor prosthesis, or a visual prosthesis, or an auditory prosthesis. Or you're going to do modulation for epilepsy, or depression. And obviously, there's a lot of progress that's been made on these things, which I'm all for. But I contend that this is not the promise of BCIs (brain-computer interfaces).”
What Hodak proposes is radical and will require much work to prove out. He’s talking about creating an external store of neurons affixed to someone’s head and those neurons having the ability to offset some of the effects of serious illnesses. The idea is that, if your brain has been damaged, the manufactured neurons head out from the interface and into your tissue and make new connections that reroute around compromised tissue. If you’ve had a stroke, for example, these additional connections could find a path to existing information in the brain relating to speech or muscle movements. This would help convey the data back to the brain interface hardware, which could translate the information into actions on a prosthesis, computer, or other device.
Current brain-computer interface systems place electrodes in small, specific parts of the brain. If, say, someone is paralyzed, you put the implant near their motor cortex and then have them think about moving. The implant captures that desire to move and sends the signal out to a computer. This is how paralyzed people with implants can push cursors across their computer screens with their minds.
Today’s most advanced implants can have more than 1,000 electrodes gathering data from hundreds or maybe thousands of neurons in one part of the brain. Science’s approach, by contrast, may allow its implant to tap into millions of neurons throughout the mind, Hodak says. More connections mean more information and that should lead to better performance for people trying to accomplish tasks.
Hodak has been dreaming about making these biohybrid interfaces since his days as an undergrad at Duke University, where he used to perform experiments with cultured neurons in his dorm room. And he hasn’t been alone. The biohybrid idea has been around for years, and other researchers have taken stabs at the technology. Still, it’s an unproven area filled with daunting technical challenges, including how you keep the manufactured neurons alive and how you get enough of them to interact with the brain.
Science has invested a large amount of money, time, and effort into shifting the biohybrid concept from the lab and toward something resembling a product. The company has, for example, spent years taking a line of stem cells and engineering them into neurons imbued with certain properties that smooth the grafting process.1 The interface hardware is also a work of nerd art with its gel for keeping the neurons alive and semiconductor- and electrode-lined channels meant to both read neuronal activity and write information into a brain.
The interface device is 9 millimeters by 6 millimeters and 4 millimeters tall. Science has placed its hardware atop a mouse brain for up to a month at a time and run many experiments with its manufactured neurons over the course of years. In its paper published at the end of 2024, the company presented some of its earlier work that centers on how mice with the implant respond to light. For Hodak and his team, the experiments proved that the grafted neurons were working and gave the company confidence to keep pursuing its technology.
The company’s co-founders Alan Mardinly and Yifan Kong have led much of the biohybrid effort and, along with Hodak, say they have no idea exactly where the technology is heading or what will end up being feasible with these extra stashes of neurons. Their thought experiments, however, are the stuff of sci-fi.
“Initially, it'll be used for things like motor decoding, or speech decoding, but where we want to go is to something very different from that,” Hodak says. “In your brain, you’ve got what we call phenomenal objects. You could also call them cognitive objects or thoughts. These are not consensus terms in the neuroscience field. This is getting into new neuroscience. But these are information objects that are defined by the total amount of information that you can say about them.”
Hodak goes on to explain that the verbal concept of ‘the sky is blue’ might take up 105 bits of information, while visualizing an image of a blue sky might take up a kilobit of data in your brain. Science would like to be able to read out these cognitive objects as data blocks.
“The really interesting stuff is how do we go from, like, closing your eyes and imagining a scene to copying that to a computer,” Hodak says. “You can just imagine a sound and then download it. We think that may be possible.”
As Hodak tells me this, I expect Mardinly and Kong, who are also in the room at Science’s headquarters in Alameda, California, to soften the stance and pull us back to reality. Rather matter-of-factly, though, they leaned right in and confirmed that, yes, this is very much where Science hopes to go.
“We’ve got a concept where a rat will navigate in an open environment and dig around for rewards,” Mardinly says. “The idea is that we could generate a cognitive map of the area and incept this map into a rat’s brain. It would essentially already know the rules of the game and where to go.”
This type of work would be made possible, the Science team says, because it can manipulate and condition its manufactured neurons in ways that are too difficult in a brain attached to a living creature. “I don’t think this is as much of a science fiction leap as maybe it sounds,” Mardinly says.
Mardinly notes that probing what’s possible has been limited by the small size of rodents’ brains and that the technology’s real scope should become more apparent as Science moves into its work with primates, which will begin this year.
Of all the things Science does, Hodak seems most excited by these biohybrid interfaces. He, however, expects it will take years and hundreds of millions of dollars to perfect the technology. In the meantime, Science will focus on making money to pay for the work via its devices now in clinical trials that help restore vision and on its lines of brain-computer interface hardware and software.
“At the very beginning of the company, we did a proof-of-concept to convince ourselves that this was feasible, which turned into the paper that we just published recently, and which is actually a different architecture than we're doing now,” Hodak says.
“Once we had that, we're like, ‘Okay, we need to get a business that can fund this thing.’ The thing that I was taught was, ‘If it's going to cost you a couple of billion dollars to make this work, it is not enough to just have the idea of the end state, you have to figure out how you get there.’”
For those of you familiar with the book - hang tight, Mrs. Frisby. Help might be on the way.
These include immune system evasion, an off switch and sensitivity to light for the company’s current optogenetic experiments.
Incredible technology, also hilariously bad name for a company. Great write up; I signed up hoping to get informed on these types of technology and you’ve been delivering in spades so far.