It is not often that an SBIR-involved start-up (founded in 2015) and with a single (NIH) Award (Phase I in 2016- Phase !! in 2017) and a handful of employees lands an $18.3 Million DARPA Award only weeks. The fact that the Award is one of only SIX made by DARPA with Five of those to major universities only underscores the unusual nature of the Paradromic Inc. achievement. Well done guys!!!
Date: Sep 10, 2017 Author: Zara Stone Source: Fortune (
click here to go to the source)
If at all possible, Matt Angle would like to shake Elon Musk's hand and give him a hearty thank you. The 33-year-old biomedical startup founder says Musk's recent launch of Nueralink (a company that is developing brain implants to blend human with A.I.) has done him a favor. Angle's eight-person startup, Paradromics, is working on brain implants, and Musk's initiative adds extra validation to their work. Well, that and the fact they've just been announced as one of the (sic), and will receive $18.3 million. It's a big deal for the startup, as the other five awards all went to big universities. Angle's task? To create a brain implant that works as a speech prosthesis, helping the many people who have lost contact with their loved ones due to speech disabilities.
The solution to this problem: Mind reading. Well, utilizing neuroscience advances to create a nickel sized chip to be implanted into the brain. The idea is that when the person wants to vocalize something they think the words and the chip translates those brainwaves into words and feeds them to an external voice box that acts as a speaker. "When we started in 2015 and talked about a data link for the brain, no one realized how close we were to commercialization," says Angle. The technical terms for this: creating high-data-rate cortical interface with penetrating microwire electrodes for high-resolution recording and stimulation of neurons.
This may sound like science fiction, but it's actually got a lot of hard science to it. Since 2006, researcher Jack Gallant from Berkeley University and his team have been using functional magnetic resonance imaging of brains to help them decode thoughts -- the idea being that different types of brain activity can be correlated to specific words. Someone sees a cat, for example, and a region of the brain that 'sees' cat then lights up. These results get fed into a pattern classifier and can then be used to translate thoughts -- he's now created word maps of the brain. If this center can be wired it could mean real-time translation -- and a huge quality of life improvement for people with speech problems. Other research demonstrates how advanced this type of brainwave-reading is becoming. In Japan in 2013, scientists used fMRI scans to identify what people dreamed about -- and generally got it right. To be sure, it wasn't completely precise -- they could tell if it was a car, but not the model of a car, but this remains a strong basis for development.
The potential for Paradromics to create a working speech prosthesis is huge -- but they're not the only ones in the game. Los Angeles stealth startup Kernel is also working on similar technology, and they have a far bigger budget to play with as CEO Bryan Johnson has invested $100 million so far. His remit is to create communication between brain cells, focusing on memory and attention -- which theoretically could treat Alzheimer's and the like. "The relationship between human intelligence and artificial intelligence (HI + AI) will necessarily be one of symbiosis. The challenge and potential of exploring this co-evolutionary future is the biggest story of the next century," he writes in a Medium post.
The Paradromics team recognize the challenges ahead but feel positive about the future. They describe their workflow process as similar to computer vision projects -- "but instead of 30 frames per second we have 30,000!" For now, Angle's goal is to have a workable investigational device by the end of DARPA's allocated four years, ready to start a round of clinical trials. Currently, their system has only been tested on mice in the labs -- the chip's currently the size of a remote control, attached via electrodes to the mouse's brain via a hole in its skull (standard lab procedure for this, before you get grossed out).
The key to getting the neuro-implant right is partly about the materials, and the team is building their own custom chips, bonding CMOS sensor technology with microwire bundles, says 29-year-old Yifan Kong, the team's Hardware and Fabrication Lead. The usual suspects for chips, such as silicon, can only handle a limited number of electrodes, and they needed more bandwidth.
"We're building a general purpose modem for the brain," he says. "The hardware doesn't cost that much, the cost is all the research time and the surgeon's time."
So what possibilities does the future hold for brain implants? It will be a while before they become commonplace as there's still a number of problems to solve -- if irritated, the brain can reject the implant, the transfer of data might not be as seamless as hoped, and the language translation is still in its infancy. But if this gets better, it's not hard to envisage extrapolating the software and using it for business and recreational purposes -- a cool usage would be an implant that lets you speak other languages, letting you be seamlessly bilingual, for example. Angle says this is physically possible, with one caveat. "It's difficult to see people getting elective brain surgery," he says. "For quality of life, that's different."
Zara Stone reports on the intersection of technology and culture. You can find more of her work at www.zarastone.net. Follow her on Twitter and on Facebook.
