Neuroscientist Brett Kagan details how Cortical Labs built the CL1, blending human neurons with hardware for adaptive computing.
Brett Kagan, a neuroscientist by training, is spearheading the efforts of building biological computers of the future.What if your computer could think more like a brain, not with transistors, but with neurons? Earlier this year, Melbourne-based Cortical Labs unveiled the CL1 : the world’s first commercial biological computer.
It uses 800,000 lab-grown human neurons reprogrammed from skin or blood samples to process information. These cells don’t just run code, they learn, adapt, and even outperform machine learning systems in certain scenarios. The CL1, which began shipping this summer at $35,000 per unit, includes a custom life-support system for the neurons and consumes a fraction of the energy needed by today’s data centers. Early adopters range from pharma researchers to finance professionals, game developers to AI scientists.spoke to Brett Kagan, Chief Scientific Officer of Cortical Labs, about the journey from DishBrain to CL1, the promise of synthetic biological intelligence, and why this might be the start of a new computing paradigm.Let’s start with the origin story. You’re a neuroscientist by training — how did you go from the academic study of neurons to building a biological-silicon interface that could play Pong? Brett Kagan: I’ve always been interested in neuroscience and the brain for my adult life. It started with psychology and getting an idea of how people worked. I became fascinated by the mechanism that actually allowed them to work – neural systems. , which we call the brain, I was fascinated by that. I went down the traditional academic route, did a PhD in neuroscience. I was starting my post-doctoral work in regenerative medicine when I found myself quite frustrated with the way academic science progresses – it is very hard to pursue new ideas. When Hon approached me and said that they were starting this company that would use brain cells as a computational device, that aligned with what I’d been interested in my entire life. We set out to see this proof of concept of not just can you grow cells in a dish, but could you actually use them for anything of interest. Since Pong is one of the first computer games and one of the first examples for machine learning, we thought it would be a fitting case for the first synthetic biological intelligence learning platform to be used for that. The CL1 is the first biological computer enabling medical and research labs to test how real neurons process information. Credit:DishBrain was a scientific provocation. CL1 is a product. What key engineering steps had to happen between those two moments? It’s been roughly a two and a half to three year journey to go from proof of concept scientific validation to a commercial product . In terms of the key engineering steps, the short answer is everything. Basically, the process that allows something to work in the lab as a proof of concept under very tightly controlled conditions is incredibly different than building something scalable and highly reproducible. Reproducibility is hard enough in science. It’s even harder at a commercial level. We had to go down to the ground truth. We’re the fullest stack company . You hear people talk about full-stack companies because they have front-end and back-end. But we’re the fuller stack company because we do not just have user interface code and website, we developed low-level code, kernel-level code, Field Programmable Gate Array , fabricated our own printed circuit boards , and the full hardware stack for the CL1 system. The development of the world’s first code deployable biological computer from proof of concept to CL1 prototype stage through the years. Credit: Brett Kagan/ Cortical Labsto stem cells to differentiating those into complex neural structures, including organoids and printed neural circuits. For the proof of concept, we leveraged the existing architectures, used off-the-shelf components and protocols that had been established by other lab groups.. There was a lot of original work that went into but the foundation we were building upon was from others. With the CL1, we built the foundation. These things are always tricky. Doing multi-disciplinary work, getting biology to work well with software, and hardware to work with software. Fortunately, we have an amazing team that works well together, and this amazing culture at Cortical Labs makes it possible. Of course, our timelines have been adjusted, and the quality of the product has been improved. There were things that we have outperformed, and in some area,s we have taken a little longer to get there, but the end result has been something more robust and more capable than what we originally set out to build.So the the really obvious ones and I think it’s good for people to know that there is time scales to this. The difference between us and quantum computing is that some of our time scales start immediately. Something like disease modeling and drug discovery. It’s not as exotic as talking about general intelligence, but it can help everyone. Nearly every person will know somebody who’s affected by a mental health, psychiatric, or neurological disease. Building a platform that can help you better understand disease, understand the brain, and understand what drugs have impacted is impactful immediately. In the long term, some of the things that neurons do are better than what we’ve been able to do with silicon intelligence. One of them is general intelligence. Whether it’s a bee or a cat or you and I, we all show different levels of general intelligence. We don’t have any ability to do this in silicon. We’re not saying we’ll build a device with these capabilities yet, but we start from this point of knowing that biology can do it. The only question is how. That’s a very powerful place to be. Some other, more concrete things along the way are fuzzy data, minimal data. So we have a paper where we’ve demonstrated that biology learns faster than machine learning in terms of the number of samples. This is exciting of course and we know this to be true. We know that we as humans need much less data than what machine learning needs to learn. We can deal with fuzzy data. We can deal with changing data. We can also do some interesting things that are harder to describe, like intuition. It’s very hard for me to explain to you why I know a thing through intuition.yet it can be quite accurate with minimal data in certain cases. Using the systems we built, we can figure out what that is. So whether we figure out the algorithm that drives biology and then implement it in silicon or build an entirely new type of information processor, where the biological neurons are the basis, there are some very powerful opportunities.The CL1 consists of a life support system that allows you to supply the cells with what they need to function. It consists of a micro electrode array , allowing us to read the shared language of neurons and silicon, which is electricity. The MEA acts as a bridge between biology and silicon and allows interactions between them with submillisecond latency. This means you can read the information from the cells, apply that information to some environment or problem set, then update the cells with more information through these small electrical pulses in under a millisecond if you want. The interface for the life support system for the CL1 helps monitor environmental conditions inside the biological computer. Credit: Brett Kagan/ Cortical Labs You can do this all using a very simple Python API and very low-level hardware that has been built to actually interpret this Python code. This is still in the early stages, but given the parallels with how our brain works, we’ve generated promising evidence, such as a pong game, which suggests this is a meaningful way to share a language between biology and silicon.If you’re using what we call Cortical Cloud, then you just need to be able to do a bit of basic Python code. If you want to buy the CL1, there are more requirements because it requires an operating wet lab. You’re welcome to purchase units provided you meet these minimum criteria, can do the work safely and ethically, and so on. Both options are available from Cortical Labs. We are here to help people access a new technology to try and grow this industry as a whole. We view the success of the company like Nvidia as a nice road map. Nvidia doesn’t write computer games or do crypto, but they provided the infrastructure for other people to advance the fields. And we’re very happy to be the ‘Nvidia of biological computation’. We’ve been blown away by the interest; just about every continent except Antarctica is interested in using this. We have people from academic labs to high schools to finance traders to cryptocurrency to gaming to music, and people with amazing ideas, so we’re excited to support them. We don’t know the best way to work with cells, we may have a pretty good idea based on current knowledge, but that’s pretty insufficient. Our hope is that a lot of different people approaching this problem from a lot of different angles, actually will be able to come up with a solution. IE:The exciting thing is I couldn’t guess. There are so many opportunities happening. We’re pursuing some very exciting work ourselves, but I’m excited to see where we’re changing the way we interact with the cells and the new algorithms we’re developing. Other people are doing exciting stuff now with our technology, too, from drug testing and disease modeling and basic neuroscience research, which as a scientist, I find very exciting, all the way up to more speculative uses in the realms of intelligence and AI and robotics as well. So there are a lot of things, and I don’t know for sure the timeline, but what’s exciting is to see what will arise.What’s one lesson or mindset shift you think more engineers should adopt when approaching intelligent systems built on biology? There are two aspects. One is the nature of the dynamic, what’s called critical bordering between chaos and order. Dynamic systems moving from something like a transistor, which can be a zero and a one, to a multi-state chaotic or the border of chaos system that is ultimately best modeled probabilistically, is quite a mind shift. Then the other thing is this idea that we call environmental programming. We can, to a degree, rewrite the genetics of the system, but it’s not going to help us so much because that’s a lot there. But the way we interact with these neurons isn’t by programming the neurons, it’s by programming the environment. And so we need to control the environment and the information structure of the environment. So, these are two big interrelated ideas that people need to get.
Biological Computation CL1 Cortical Labs Inventions And Machines Machine Learning Neurons Neuroscience Silicon
United States Latest News, United States Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
Figment Acquires Rated Labs to Bolster Staking Data for Institutional ClientsThe acquisition, whose value wasn't revealed, will allow Figment to provide its clients with stronger data tools, including Rated's Explorer and data APIs.
Read more »
Grado Labs Expands Award-Winning Signature Collection With The New S750 HeadphonesThese new premium headphones from Grado Labs feature a brand-new dynamic driver design and new ear cushions. The S750 are the third model in Grado's Signature Collection.
Read more »
Grado Labs Expands Award-Winning Signature Collection With The New S750 HeadphonesThese new premium headphones from Grado Labs feature a brand-new dynamic driver design and new ear cushions. The S750 are the third model in Grado's Signature Collection.
Read more »
Jennifer Lopez Reveals Brett Goldstein as Best On-Screen KisserThis is additional taxonomy that helps us with analytics
Read more »
Jennifer Lopez Claims Brett Goldstein Is the Best KisserThe woman has kissed George Clooney, Matthew McConaughey, Ralph Fiennes, Richard Gere, Owen Wilson, Josh Lucas...
Read more »
Ripple Labs Reportedly Leading $1B Fundraise for XRP TreasuryThe fintech firm wants to start a digital asset treasury for its own native cross-border payments token, XRP.
Read more »