The Next Digital Frontier: Biology
- By Winston Thomas
- May 15, 2023
Biology has been at the receiving end of the digital evolution for decades. New advancements help us to understand our physiology, tailor our drug development, and detect signs of debilitating diseases earlier. Dr. Hon Weng Chong from Cortical Labs is doing the reverse.
His company is creating a hybrid chip by fusing living neurons onto computing devices and plugging them into data centers. CDOTrends caught up with Chong to find out how biocomputing is evolving.
Let’s get a bit of a definition. From your perspective, what is a biocomputer?
Chong: Your traditional computer comprises two elements: your hardware and software. A biological computer has a third trifecta: wetware. In the case of what we're doing, the wetware component comprises biological neurons that we derived from two sources: mouse cortical cells and human cortical cells grown from adult pluripotent stem cells.
Why mice?
Chong: We started with mouse neurons because it’s accessible. The brain is also comprised of lots of different types of cells. You have neurons, but you also have cells like glia that support the operation of the neurons. The benefit of going with new mouse cells, especially embryonic ones, is that you have a ready-made mix of neurons.
So, why move to human neurons?
Chong: By sheer foresight and luck. My chief scientific officer, Brett [Kagan], had massive allergies to mice. So he suggested trying human pluripotent stem cells. And, to our surprise, the human stem cells outperformed the mouse cells. It's surprising because we didn't know the ratio of excitatory to inhibitory neurons and the amount of glial cells to neurons.
Is your goal to be part of the orgonite intelligence movement?
Chong: I think we're aligned with the OI group. We're both looking to see if we can use these biological neurons as a computing substrate. The slight difference is that you don't need to go to Orgonite levels to get intelligence. And we've proven it with a flat plane of neurons to control a paddle to hit the ball. Then we've proven you don't need an organizer, although we don’t dispute that going to larger structures with an organizer will give better results.
One of the things that stood out in the articles about your hybrid chip is the low use of energy. Can you put this into perspective in terms of data centers?
Chong: The energy usage is almost similar to a pocket calculator with a solar panel. Now, for perspective, the human brain, which is pretty powerful, only consumes 20 watts of energy. This is at least 10 to 15 times less than a single GPU. There is a secondary benefit as well. These neurons don't produce heat or only a minimal amount of it, saving on the energy bill because we don't have to spend so much on cooling and air conditioning.
Sounds like a perfect fit for hyperscalers. Are they buying into your idea?
Chong: I can't say who, but we are working with the cloud providers to potentially plug neurons into this service so that people worldwide can access the technology without needing a wet laboratory, which is the biggest hurdle to adoption and further research.
Is having no wet laboratories a significant hurdle?
Chong: Yes. And if we can take care of that bit, and abstract away the wetware components and make it accessible to developers and researchers in interfaces that are familiar with, say, JavaScript or Python, we could potentially get the same amount of or maybe a fraction of the interest that had been generated from applications like ChatGPT.
You started with mouse and human neurons. But do you see yourself creating new types of cells other than what is available in nature but is more optimized to run a hybrid chip? Maybe synthetic cells with bigger mitochondria, for example.
Chong: We're trying to shortcut the whole process by taking what nature has essentially evolved through billions of years of evolution. But it doesn't mean we stop there because there are many things we don't need, and it's somewhat irrelevant to this engineering process. For instance, we could give the cells the ability to respond to light or emit light, which is called optogenetics. Optogenetics allows for greater resolution and precision in stimulating and reading the responses.
We know data centers are a significant focus. But what other industries do you think would benefit most?
Chong: Health Sciences. The fact that we are using human neurons opens up an exciting avenue for drug discovery and development. So right now, there's a lot of buzz about generative AI and quantum computing discovering molecules. And it's all great that they are discovering all these molecules. You must still test them against an in vivo or in vitro subject. Our system could theoretically fill that void at much lower costs, allowing for much higher throughput, where a generative AI source could discover a compound and then get fast-tracked on our system. Essentially, we may have developed the world's first in vitro corporate testing system. So that means we can theoretically grow cells, test drugs and see the effects and outcomes. And I think this will be very exciting because we could start making medications faster and cheaper.
So that's one aspect. But the other is also that we take the neurons from individuals and grow them in these tissues that are typically identical to the ones with the donor. So, therefore we should also have the same disease and drug profiles. And this is important because one of the downsides of clinical practice is that if you had epilepsy, you would be tried on one of the seven drugs. You would be told to take that and come back six weeks later and see whether it worked and whether you had any side effects. For many people, this can be a very debilitating process. The hope is that with technology like ours, we bypass that whole process and say don't bother with drugs A to D, go with E because we've shown that the patient responds best.
What about other industrial uses? Robotics comes to mind.
Chong: We showed that our system outperforms reinforcement learning when given a task setting. So, when we simply constrained the reinforcement learning agents only to see the same amount of information that our biological systems did, the biological systems massively outperformed them. The implication is that if anything needs to operate in real-time, like a robot, this could potentially be a better alternative to building deep neural networks for reinforcement learning. The other (application) is cybersecurity. Here, you need to respond to situations you've never encountered before with minimal data.
I don't want to dismiss other uses, like trading Bitcoin, which somebody had. We don't particularly have the internal bandwidth to do these projects. But we are building APIs to allow people to explore this idea in the near future.
What's your stance when installing wetware in humans and augmenting them? Obviously, there is military interest.
Chong: I think anything that improves the human condition is ethically good and sound. You could also argue that an augmentation that improves reflexes and ability could potentially be helpful if somebody has cerebral palsy. It's a measure of what is the intent and how does it improve the condition. You live in Asia, and you know what it's like with the hyper-competitive tutoring and pressure that kids go through. If there was a way to augment the ability for somebody to study harder and longer and get better results, is that ethical use of it? After all, it can help the students achieve more and have greater prospects in their future careers. So really, it's a question I can't particularly answer. I have my personal views that I've expressed, but it needs to be a conversation that is to be had with society as a whole.
Society doesn't want red lines to be crossed. But having said that, all these values change over time. And so we need to reassess continuously. It was the case of in vitro fertilization. When the first IVF baby was born, there were a lot of questions about the ethics of that. Now, we fast forward to today, and it's very normal. People are even getting government subsidies for IVF treatment. So I think human augmentation is something that we cannot just look at in isolation, but we need actually to have a conversation with everyone.
One last question: What if you have sentience? Will that be a challenge or opportunity?
Chong: I think that these neurons are sentient. But it depends on the definition of sentience because I also believe that amoebas are also sentient. If you put noxious stimuli on an amoeba, they'll move away. And so it's the ability to respond to environmental changes, which I think is the definition we put for sentience. The issue has been people conflate the word sentience with consciousness because you can become sentient without becoming conscious.
Ok, let’s say your hybrid chip reaches consciousness, especially when we build more complex biocomputers. What then?
Chong: This is why we need to engage with the bioethics, the philosophy component of it. We also need to question why we need to grow such large systems. So, for instance, are we growing a very large system with more capabilities to find a cure for dementia? We need to ask society if this is a worthy trade-off. This current debate is happening with things like ChatGPT and generative AI systems.
Winston Thomas is the editor-in-chief of CDOTrends and DigitalWorkforceTrends. He’s a singularity believer, a blockchain enthusiast, and believes we already live in a metaverse. You can reach him at [email protected].
Image credit: iStockphoto/solarseven
Winston Thomas
Winston Thomas is the editor-in-chief of CDOTrends. He likes to piece together the weird and wondering tech puzzle for readers and identify groundbreaking business models led by tech while waiting for the singularity.