This Startup Fused Human Brain Cells with Silicon Chips | E2295

Introduction to Corticle Labs and Biological Computing

• Corticle Labs is a company that is working on fusing silicon chips with human neurons to create a new type of computer, a biological computer, which brings the biologic and the synthetic together, and they have made significant progress since their initial conversation in 2025, including building out data centers of its biological computers 10s.
• The company's first product, the CL1, was a fully contained biological computer with neurons and chips, and it was sold for $35,000 a piece, with all 30 units in stock being sold, resulting in roughly a million dollars in revenue 2m6s.
• The CL1 is a computing platform that allows researchers and developers to work with biological computing easily, saving them the effort of building their own hardware and writing their own software, and it can be programmed with Python, with several units already being used by institutions such as Johns Hopkins, Mass General, and UCSF 4m30s.
• The CL1 has a similar form factor to rack-mountable sleds used for GPUs, taking up about three rows in a server rack cabinet, and Corticle Labs has 120 of these units in six racks in their lab in Australia, which they call the world's first biological data center 6m40s.

Performance and Ethical Implications of Biological Computing

• The company's biological computers have been shown to be 5,000 times more sample efficient than reinforcement learning systems, and they have the potential to revolutionize the field of computing, but they also raise ethical concerns, such as the possibility of creating conscious systems that can suffer 42s.
• The Vatican has expressed concerns about the potential risks of creating conscious systems, and Corticle Labs is aware of these concerns and is working to ensure that their technology is developed and used responsibly 1m20s.

Technical Components of the Neural Interface

• The neural interface is composed of a compute unit, life support systems, and a neural link, where neurons are loaded into a chip with a heating element to maintain a temperature of 37°C or 100 Fahrenheit, and the system includes mechanisms to provide nutrients and remove waste 10s.
• The life support system includes pumps, a feeding and waste reservoir, filtration units, and a gas mixer, which are comparable to the heart, stomach, bladder, kidneys, and lungs, and also features traditional and non-traditional input/output units, such as USB, Ethernet, and gas inputs 42s.
• The system can pack up to 1-2 million neurons, but the cortical cloud offerings typically use around 200,000 neurons, which is a commercially viable number that allows for learning and training, and this number is analogous to parameters in a large language model 2m6s.
• The human brain contains approximately 100 billion neurons, and the number of neurons in the CL1 is comparable to that of a cockroach or a fly, which are capable of solving certain problems, such as agility and quickness, but not human-like intelligence 4m10s.

Capabilities and Limitations of Biological Systems

• The combination of neurons and a silicon chip can result in a multiplication factor that enhances the capabilities of the system, and the industry needs to develop a clear terminology to distinguish between super intelligence, like GPT, and general intelligence, which would require the ability to adapt to new situations, like making a cup of coffee in a stranger's kitchen 6m20s.
• Biological computing has the potential to exploit the properties of simple organisms, such as goal-seeking behavior and path finding, to achieve tasks more efficiently than traditional machines, and researchers are hoping to utilize these properties without having to replicate human intelligence 10s.

Advantages of Biological Computing in Reinforcement Learning

• The fusion of silicon and neurons could potentially create a computer that surpasses the capabilities of current GPUs and TPUs in certain types of calculations, particularly in reinforcement learning, where biological systems have been found to be 5,000 times more sample efficient than GPU-based systems 8m30s.
• The advantages of biological computing in reinforcement learning are significant, especially in applications where accelerating time is not possible, such as in robotics, and this technology could have a major impact on the development of AI models 12m20s.

Global Expansion and Data Center Infrastructure

• A biological data center has been built with 120 units in Melbourne, and plans are underway to establish another one in Singapore, which is expected to be even larger, in partnership with a data center company called Day One 18m40s.
• The potential use cases for biological computing are vast, including drug testing, drug discovery, and biosimilars, and the technology has the potential to revolutionize various industries beyond the biology sector 2m6s.
• A data center is operating within a tight energy budget of 200 megawatt, as specified by the Singapore government, and the company's compute solution does not affect this energy budget, using only about 30 watts of energy per unit, which is virtually zero for this conversation 10s.

Operational and Economic Aspects of Biological Data Centers

• The company has partnered with a data center to provide their compute solution, which is not only energy-efficient but also allows for the growth of cells for compute on site, eliminating the need to ship in stem cell-based neurons and reducing operating costs 1m42s.
• The data center has built a laboratory to grow the cells for compute on site, which saves a tremendous amount in terms of operating costs and decouples the data center from relying on a central vendor, making it self-sufficient 2m6s.
• The neurons used in the compute solution can live a very long time if well kept, but the filtration units, which act like kidneys, need to be replenished every four to six months due to clogging with large protein growth factors 4m30s.

Biological Maintenance and Sustainability

• The neurons are fed a solution similar to sugar water, and the biological system keeps a small number of neurons alive, connected to chips that perform various tasks 6m10s.
• The company's biological compute solution is designed to work in tandem with chip technology, and as chips get smarter, it is uncertain whether more neurons will be needed to interface with them or if improvements to the chip component will be sufficient to achieve gains 8m40s.

Challenges in Algorithm Development and Interface Technology

• The current limitation in algorithms is representing digital information with analog systems, such as the human brain, and companies like Neurolink and Synchron are working on this issue for Brain-Computer Interface (BCI) development, with the additional challenge of writing information directly into neurons 10s.
• The development of artificial intelligence is currently relying on power-hungry and inefficient systems, whereas the human brain is a more efficient and attuned system, and it is logical to expect that combining the two could lead to faster gains in intelligence 2m6s.

Ethical and Religious Considerations

• As the technology advances, there are concerns about how to communicate the fusion of neurons with computers to the public, particularly in relation to religious and ethical concerns, with the Vatican having expressed worries about the technology 4m30s.
• The company has been engaging with bioethicists and ethics boards to address potential criticisms and ensure that the research and development are conducted responsibly, with a focus on proactive communication and collaboration 6m20s.
• The company's progress has been significant, from proof of concept to early commercialization, and now to building out data centers internationally, which is bringing the technology to market and raising questions about its potential impact on humanity 8m10s.
• The main goal is to bring together experts in the field to agree on a shared nomenclature, as a lack of understanding and fear can hinder progress, and this is considered crucial for making advancements 10s.
• There was a discussion with the Vatican to address potential religious and ethical issues, and fortunately, they agreed that the work being done is acceptable, with the key question being whether there is a harmonization and a net good that can come from this technology 42s.

Applications and Research in Medical and Scientific Fields

• The technology is being used by major medical centers, such as Mass General, Hopkins, and UCSF, for traditional drug discovery, disease modeling, and alternatives to animal testing, including research on dementia, toxicology, and movement disorders 2m6s.
• The compute side of the technology is the smallest but newest and most risky, as it is uncertain what people will build with it, and internally, the company is trying to understand the discussion of consciousness to draw a red line that should not be crossed 4m30s.

Commercialization and Future Planning

• The company's stance is that conscious systems should not be created, as they have the ability to suffer, and most users are academic R&D institutions that can be trusted to do the right thing, with 90% of purchases coming from these institutions 6m15s.
• A data center has been built in Melbourne and another in Singapore, with plans to have up to a thousand neurons, which raises science fiction questions about the potential interactions between these biological computers, but it is considered that this is still far in the future 8m45s.
• The decision to go the cloud route instead of just selling devices was made, despite having a sold-out first run and plenty of demand, and the reason for this choice is being discussed, with the company considering the potential benefits and risks of their technology 11m20s.

Democratizing Access to Biological Computing

• The goal of Chronicle Labs is to democratize biological computing by reducing the accessibility barrier to this technology, making it available to more people, similar to how Nvidia's CUDA made AI more accessible by being free and compatible with any GPU, even the lowest-end ones 10s.
• To achieve this, Chronicle Labs created a cloud system that allows users to experiment with biological computing without needing a lab or extensive biological knowledge, enabling people with ideas to test them without requiring significant resources 2m6s.
• The cloud system was inspired by the story of a student developer who used Chronicle Labs' APN SDK to build a project, called Doom, during a hackathon at Stanford, despite having no biology background, and was later accepted into a prestigious incubator program 4m42s.

Early Applications and Learning Systems

• The first application of Chronicle Labs' technology was to a video game, Pong, which required creating a reward or punishment mechanism to give the system incentives to learn how to play correctly, using audit stimulus versus disordered stimulus as the reward and punishment signal 6m10s.
• When applying this technology to a more complex game like Doom, the approach was similar, but with more variables, and the system was able to learn how to play the game by using the same reward and punishment mechanism, although it exploited a loophole by spamming the shoot button and spinning around in a circle to win 8m40s.
• The discussion involves reinforcement learning systems and the importance of not creating conscious systems that can feel pain, as using disordered inputs as a punishment mechanism can be harsh 10s.

Deployment and Scalability of CL1 Units

• A kid can build a system with a certain API, which shows that the technology is not hard to bring to real-world applications, and this leads to a question about how many customers can be served with 120 CL1s in a data center 2m6s.
• The company has 120 CL1s deployed, but only about 20 users are on the cloud at the moment due to teething issues with the biology, and they have a conveyor belt system for getting sales grown 4m42s.
• Each CL1 is a discrete box and can serve one customer at a time, but the company is working on chaining the systems, which will be available in the next two or three updates 8m10s.
• The company is also doing internal research work on PDMS micrfluidic wells to segment out the surface of the chip, which can increase yield and allow multiple people to run the same thing, similar to a virtual machine 10m40s.

Materials and Microfluidic Innovations

• PDMS stands for poly dimethyloxane, a material used in the company's research 12m10s.
• The material being discussed is dimethylone, a gel-like substance used to make microfluidic devices, which can be thought of as a circuit that controls the growth and communication of neurons 10s.
• Microfluidics allows for the creation of channels that can be turned on and off, enabling control over the communication between neurons, which are segmented into little sections, with several thousand neurons in one well 2m6s.
• The microfluidic device has channels that constrain the growth of the axons, the long part of the neurons, allowing for circuitry-like control over the communication between neurons, which can only communicate with their nearest neighbors 2m6s.
• The chiplet model is being experimented with, allowing for the segregation of workloads, so that multiple people can use the same computer without the axons getting too busy sharing information 4m30s.

Cortical Cloud and Developer Tools

• The cortical cloud is a developer backend that allows for live streaming of neural activity from a cell culture, with the ability to acquire an instance and view the neural activity in real-time, regardless of the location of the lab 8m30s.
• The raster plot is a visualization of the neural activity, showing the channels and the topological mapping of the surface of the chip, with each spike or action potential marked on a grid, allowing for the observation of temporal correlation, bursting activity, and synchrony 10m30s.
• The raw waveform of the neural activity can be viewed, similar to what a neural link would pick up, and the neurons can be stimulated by sending a signal to a specific region 12m0s.
• The discussion involves interacting with human brain cells that have been fused with silicon chips, and it is possible to deliver a stimulus to the cells and observe their response, with the goal of developing applications using this technology 10s.
• The system is not conscious, and developers can write application code in Python using Jupyter notebooks to specify parameters and objectives for the neurons, with documentation available at docs.corticolabs.com and an SDK that can be installed using pip 42s.
• The technology has a wide range of potential applications, and the company is looking for talented developers to help expand and optimize the system, with a cloud platform available at cloud.corticolabs.com and more information available on Twitter and the company website at corticalabs.com 2m6s.

Community Engagement and Future Development

• The company is also planning a hackathon at MIT, and they are looking for developers to participate and contribute to the project, with the goal of advancing the technology and exploring its potential uses 4m30s.
• The conversation then shifts to a different topic, discussing drones and their potential applications in agriculture and cargo transport, with a guest from PAH joining the conversation to talk about their company's work in this area 6m10s.

Introduction to PAH and Drone Technology

• The guest, Michael Nortra, discusses the history of his company and how it has evolved over time, from its initial focus on large drones for crop spraying to its current broader range of activities 8m20s.
• The individual's interest in building aircraft started at a young age, with paper airplanes and rubber band powered airplanes, and eventually led to studying physics and working on electric vertical takeoff and landing air taxi concepts in the Bay Area 10s.
• After realizing that certifying those aircraft for commercial use would take around two decades, the decision was made to leave and start a new company, which marked the beginning of a long journey that has lasted almost 10 years 2m6s.

Development and Deployment of PAH's Drones

• The first flight of the company's drone took place a week before demo day, after spending one month designing and two months building a 600 lb autonomous aircraft that could carry two passengers, and it successfully flew itself using the company's own software 4m42s.
• The company has made significant progress since the first flight, going from a prototype to a widely deployed aircraft operating with customers 7 days a week, with some customers using it for 12 to 13 hour shifts, and it is currently the only company to have deployed a group of UAS to commercial customers at scale 6m15s.
• The company's drones are classified as group four, which refers to drones that are bigger than 1320 lbs, such as the MQ9 Reaper, and are being flown commercially in the rural jungles of Brazil 8m30s.
• The reason why having a flying prototype was only 1% of the work is that the company's products are a blend of hardware and software, and maturing the hardware in particular takes a long time, with the first aircraft not being able to complete an interesting customer mission and lacking necessary features 10m45s.
• The company has been steadily maturing both the hardware and software over the last nine years, with the last three or four years being particularly significant for commercial deployments with customers, who require a much higher level of uptime and reliability than a prototype 12m50s.
• Customers of the company have high expectations for the uptime of the aircraft, with some located in extremely remote regions of the world, requiring an extraordinary amount of time and effort to achieve a high level of reliability, similar to product market fit, but more focused on product reliability 10s.

Operational and Regulatory Challenges in Drone Deployment

• The company's drones are currently being used in Latin America, particularly in Brazil, where regulatory conditions are more favorable, with the country having deregulated agricultural drones about 2 years ago, allowing for more flexible operation 42s.
• In the United States, the company has commercial approval to operate its aircraft, but with limitations, such as only being able to fly the drone about 4 kilometers away from where it took off and landed, which is not commercially viable, due to line of sight regulations 2m6s.
• The company is working with the FAA to remove these limitations and expand its operations in the US, with one customer already using the aircraft, and plans to have more customers in the coming year 4m6s.
• The company's drones are mostly electric, and when sold, they come with multiple sets of batteries that can be hot-swapped, but recharging can be a challenge in remote areas, with some farms using diesel generators, while others have access to electricity, which can be used to recharge the drones 6m6s.
• The company's aircraft competes with other vehicles, such as the air tractor, a large 8,000 lb vehicle, and the company is making progress in removing regulatory limitations to expand its operations in the US 8m6s.

Economic and Environmental Advantages of Electric Drones

• The Pelican drone is compared to a traditional aircraft that burns roughly 55 gallons of jet fuel per hour, whereas the Pelican uses about 2 gallons per hour, even when charging off a diesel generator, making it a more efficient option 10s.
• The current price of the Pelican drone starts at $550,000, which, although not cheap, can lead to significant savings due to its autonomous and electric features, potentially reducing the need for human pilots 1m5s.
• The return on investment (ROI) for the Pelican drone depends on how heavily it is utilized, with some customers planning to run it for 24 hours a day, and the payback period is estimated to be around 2 to 3 years 2m6s.
• The Pelican drone is not suitable for small farmers, who would need to have a farm of almost 20,000 acres to fully utilize the aircraft, making a "spray as a service" model more likely, where a contractor owns the vehicle and services multiple customers 4m30s.

Chemical Application and Cost Efficiency

• The Pelican drone offers a superior solution for applying chemicals, with the ability to spray a fine mist evenly over an entire crop, including around the boundaries, which is a difficult task, and the cost of the chemical being sprayed is typically about four times the cost of the application 6m40s.
• The current global situation, with the Strait of Hormuz being blocked, has led to increased fertilizer prices, making farmers more conscious of the cost of chemicals, and the Pelican drone's efficient application method can help reduce these costs 9m20s.

Introduction to Drop Ship and Logistics Applications

• The conversation starts with a discussion about farming and the application of solutions to address the crisis in the food supply, which leads to the introduction of PAH's work on drones, software, and hardware integration, including the development of Drop Ship, a dual-use product for logistics and defense sectors 10s.
• Drop Ship is described as a versatile multi-mission aircraft capable of carrying large sensor payloads, and while it could technically carry bombs, it is not designed for that purpose due to its reliability and low operating cost, making it more suitable for other applications 2m6s.
• The development of Drop Ship was rapid, with a test flight taking place just 180 days after the initial CAD renderings, and it is considered the second generation of cargo planes, with the first generation being an all-electric cargo aircraft with a limited range of 200 miles 4m30s.
• The first generation cargo plane was built in eight units, with three sent to the Air Force and one to the Army, receiving positive feedback and invitations to demos, and the insights gained from this experience helped inform the development of Drop Ship, which meets the requirements of a 1,000 mi range with 500 lb of payload and fits in a 20ft shipping container 6m40s.
• The development of Drop Ship was influenced by the feedback from the Air Force, SOCOM, and Army, which highlighted the need for a treatable contested logistics and multi-mission UAS with a longer range and easier maintenance, and Drop Ship is designed to address these requirements 8m50s.

Design and Hybrid Architecture of Drop Ship

• The discussion revolves around a new dropship that has a hybrid architecture, featuring both a diesel engine and batteries, allowing for quiet looping and airdrop capabilities 10s.
• The hybrid turbo diesel architecture is a parallel hybrid, where the diesel engine has a propeller attached to it and can charge the batteries in flight, while the electric propulsion system has two high-power electric motors used during takeoff 2m6s.
• The diesel engine has a peak power of just over 30 kilowatts, and the electric motors in the front are each about 25 kilowatts, resulting in a peak power that is essentially three times the cruise power when the electric propulsion system is running 4m30s.
• The dropship's design allows for ballistic takeoff and landing performance, and once at altitude, the electric propulsion system can be shut down, and the airplane can run on just the diesel engine 6m20s.
• The potential for domestic commercial applications, such as delivering large items, is hindered by regulatory issues, and it is suggested that the country may be missing out on the benefits of autonomous flight due to these regulations 10m30s.

Regulatory and Market Challenges in Drone Logistics

• The regulatory approval process is cited as the reason for pivoting away from mass manufacturing the electric cargo drone, as it was not possible to fly beyond visual line of sight in a timely manner 12m40s.
• The importance of hardware and software blended products is emphasized, and it is noted that their value is determined by their exposure to the real world, with companies like SpaceX being able to collect data and create value through experimentation and innovation 15m10s.

Supply Chain and Manufacturing Considerations

• Companies like Zipline have established operations in countries such as Rwanda, while others have gone to Ukraine, highlighting the limited options available for certain types of development and production 10s.
• The discussion revolves around the challenges of sourcing materials and components without relying on Chinese supply chains, particularly for military and dual-use applications, with the goal of achieving sovereignty in supply chains 1m42s.
• The company has vertically integrated the design and manufacturing of critical components, including motors, batteries, and motor controllers, allowing them to create NDA-compliant versions of their products 2m6s.
• There is a cost differential of about 2x when manufacturing certain components, such as battery management systems, in the United States compared to China, but this is manageable since the company owns the design 3m30s.
• The decision to design and manufacture components in-house has introduced technical risk, but the company's strong technical team has been able to overcome the challenges, and while it has slowed them down, it has also provided benefits in terms of control and customization 6m40s.
• The company has encountered issues with both their in-house designs and off-the-shelf products, but they have been able to resolve them, and the ability to own the design and manufacturing process has been a key factor in their success 8m50s.

Integration of Hardware and Software for Competitive Advantage

• The company's approach is similar to that of other successful hardware companies, which often blend hardware and software to achieve their goals 11m20s.
• Companies like Apple and Tesla are able to make complex products seem simple to the end user, and this is achieved by owning both the hardware and software, which allows for a seamless and magical experience 10s.
• The importance of owning both hardware and software is highlighted as a key factor in creating a great customer experience, and trying to integrate different systems can lead to a subpar experience 42s.
• The conversation touches on the topic of Boeing's decision to stop making things in-house and instead supply everything externally, and how this approach can affect the customer experience 2m6s.

Funding and Venture Capital Landscape

• There is a lot of money flowing in the venture capital world, but some companies, such as those building enterprise agent orchestration NCP servers, are struggling to secure funding, and it is questioned whether companies have enough access to capital to continue growing 4m6s.
• The company in question has been able to raise the money it needs, but having more funding would allow it to move faster, and a call is made to venture capitalists to support innovative companies rather than those just slapping on an AI wrapper 5m30s.

Company Contact and Public Presence

• The company can be found on its website, flypyka.com, and is active on LinkedIn and Instagram, with many cool shots from its customer operations in Brazil 7m30s.