Deep dives into deeptech. More PhD than MBA, more hacker culture than compsci grad. Someone sent you this? Sign up here
Good Morning edwardlorilla1986.paxforex@blogger.com,
A neuron is a natural marvel. Grow one in a petri dish and it will intermittently fire, sending an electrical signal along the cell. Introduce a neurotransmitter such as serotonin or dopamine, and it will exhibit different behaviours, delivering electrical current with different strengths and timings. That's the basis of how life responds to external stimuli, the foundation of high-order thinking.
On its own, a single neuron doesn't accomplish much. But join a bunch of them together and a nervous system emerges. When that nervous system is part of a body that can detect its immediate environment, then these neurons are crucial to forming the organism's reactions—fight, flee, seek more input, and pretty much any other response that you can think of.
Networks with more neurons have more connections, and the way thoughts are formed become more complex, more mysterious.
|
So far, it isn't clear how our cognition emerges. The matter has captured our collective imagination over the past months, with various artificial intelligence tools penning poems and generating images in ways that imitate human creativity. But what if, instead of writing code to mimic human thought, biological matter drives the process? This is the main idea behind the brain-on-a-chip (BoC).
One startup based in Singapore, Cortical Labs, is making headway in this space. Its researchers are mounting neurons on high-density, multielectrode arrays so they can do a bit of heavy lifting when solving problems. To find out more about what the company is up to, I spoke to Brett Kagan, the startup's chief scientific officer.
Brain games
Kagan has the credentials that you'd expect from someone operating at the intersection of specialised fields—his work blends psychology, neuroscience, stem cell therapy, regenerative medicine, bioinformatics, ethics, and philosophy. When he joined Cortical Labs in September 2019, the goal was to see if living neurons could be coaxed into performing "intelligent" actions.
Here's how Kagan put it:
|
There's an elegance to the idea of taking what's naturally available and letting it grow into a role. Kagan and his colleagues at Cortical Labs even came up with a catchy way to demonstrate how this can play out. Last October, you might have seen coverage about DishBrain, Cortical Labs' proof of concept, learning how to play Pong, the Atari video game. (The outlets that covered DishBrain include BBC, Science Alert, and NPR).
|
At this point, it's important to note that the cells grown by Cortical Labs on its chips do not form an actual brain, even though the word appears in DishBrain's extremely catchy name. While the BoC can respond to stimuli, there is no emotion, consciousness, or high-level cognition. There is, however, what Kagan and his colleagues describe as sentience—in that it "responds to sensory impressions through internally adaptive processes," but has no awareness.
The making of a brain-on-a-chip
Cortical Labs creates machines with biological intelligence by building upon three pillars—biological advancements, hardware and software development, and neurocomputational understanding.
The neurons used by the company are grown from stem cells, which are generated from human blood samples or skin, meaning no living being is harmed in the process. Kagan likened this tissue to what you might give a doctor for a blood test.
He describes how these brain cells and computers communicate in fundamentally the same way, making it possible to fuse the two:
|
With communication established, coaxing comes next. The idea is to use neurocomputational techniques to convince the neurons to perform useful tasks, specifically based on the concept that a neuron "wants" to control its environment and predict outcomes. The model's design is based on theories that describe how intelligence arises.
Given DishBrain's name, and given that it's created using tissue grown from human cells, it's easy to anthropomorphise Cortical Labs' BoCs. But the fact is the neurons in DishBrain don't experience emotions. Kagan and his colleagues have considered the matter at length. The first paper published by Cortical Labs wasn't a proof of principle, but commentary on the ethics underlining the company's research.
It's difficult to imagine how you or I could use DishBrain—or more precisely, how it might be embedded into products that we use. The company's next step is to define ways for other folks to utilise their creation, in the same way that room-sized mainframes were simplified and redesigned over time so that anyone can now use a personal computer. Cortical Labs is trying to make it possible for anyone with basic coding skills to interact with BoCs, so accessibility and affordability are the goals.
Fresh thoughtforms
Cortical Labs isn't the only company seeking to create true generalised intelligence by utilising biological material, so a future where brain cells are fused into computers feels within reach. That brings about some interesting questions.
While we intuitively understand how humans think, and intellectually grasp how a silicon chip processes information, there isn't an easy way to think about how a BoC reacts to conditions that are presented to it. Even though BoCs are made from materials that exist in living beings and silicon computers, the frameworks that shape our knowledge about them don't quite fit how we might consider the way BoCs function.
That's all to say that beyond the scientific research and development performed by companies like Cortical Labs, new mental models need to be defined so that there can be meaningful applications for BoCs.
Kagan said Cortical Labs' goals are too out there for the company to secure government funding. The resources needed for the R&D into meshing silicon and tissue had to come from private investors.
|
DishBrain appears to be a sufficient demonstration that there is something to Cortical Labs' thesis. In April, the company raised US$10 million in a funding round led by Horizon Ventures, the investment arm of Li Ka-shing, the wealthiest person in Hong Kong.
In the first edition of Present/Future, we touched upon how commercialisation for deep tech companies can be tough to navigate—there needs to be a balance between the end goal and near-term sustainability to convince investors that years of research deserves backing.
Kagan casts Cortical Labs as a company that is developing a platform rather than a biotech product, with the aim of creating something that is along the lines of computer transistors and silicon chips. A fully autonomous brain-on-a-chip is an exciting outcome that sparks the imagination.
A new kind of processor could change how big problems like drug discovery and disease modelling are processed and solved. Aside from the fact that these are multibillion-dollar industries from which Cortical Labs could draw revenue, there are real life-changing solutions that could be unlocked. I think that's what makes this incredibly exciting.
Regards,
Brady Ng
P.S. Be sure to write to brady@the-ken.com with your thoughts and suggestions, and please do share Present/Future with your friends and colleagues.
No comments:
Post a Comment