Podcast Summary
Biology's Digital Revolution: Synthetic Biology and the Physicality of Biological Code: Synthetic biology applies digital concepts to biology, but its physical nature presents unique challenges. AI can help solve some of these challenges, but working with biological code requires a physical lab setting and a different mindset.
Biology is undergoing a digital revolution with the development of tools and infrastructure for synthetic biology, which can be thought of as the manipulation of DNA as code. However, unlike computers, biology is a physical system where the code is made up of actual molecules and proteins. While there are similarities, the way cells process this code is vastly different and presents unique challenges. Ginkgo Bioworks, a leading company in synthetic biology, aims to make cells as easy to work with as computers. The core idea of synthetic biology is that it runs on code, and the question is what can be brought over from programming into this world that actually works. One area of overlap is the use of AI, which can help solve some of the challenges in understanding and manipulating biological data. However, unlike programming, working with biological code requires a physical lab setting and a different mindset. Tom Knight, a computer architect who made the connection between DNA and code in the mid-90s, famously moved from working on computers to working with DNA in a lab, only to discover that the results were not consistent and that some things were just unpredictable. This physicality and unpredictability present unique challenges, but also offer exciting opportunities for innovation in fields like food, medicine, and agriculture.
Exploring Complex Systems in Computer Science and Biology: Computer systems offer predictability and controllability, but biological systems are more unpredictable and complex, presenting unique challenges and rewards, especially in AI and biological engineering
Working with complex systems, whether in computer science or biology, presents unique challenges and rewards. While computer systems offer the advantage of predictability and controllability, allowing for easy debugging and design, biological systems are more unpredictable and complex. However, their inherent power and potential make them worth exploring, despite the challenges. The field of AI, specifically neural networks, is an example of this, as it offers the potential for incredible advancements, but also presents difficulties in understanding and debugging. Similarly, the field of biological engineering holds immense potential, despite its unpredictability and the challenges of working with evolutionary systems. The magic and wonder of discovering and manipulating these complex systems is what drives researchers and innovators in both fields. Ginkgo Bioworks, as a company, aims to apply principles of abstraction from computer science to biology, separating the lab work from the scientific research, allowing for greater efficiency and scale in biological engineering.
Ginkgo Bioworks: Driving Scale and Innovation in Custom Organism Production: Ginkgo Bioworks uses its infrastructure to create custom organisms for clients, improving protein production and enabling advancements in various industries through a unique business model.
Ginkgo Bioworks, a leading organism design company, has been driving enormous scale in lab work to test various genetic designs for their clients, enabling the development of new products through a unique business model. The company's customers, which include pharmaceutical giants and agricultural companies, interface with Ginkgo by agreeing on a specification for a desired cellular function. Ginkgo's scientists then use the company's infrastructure to create the custom organisms, which are licensed back to the clients. The goal is to improve the quality and production of proteins, often used as catalysts in various applications like cold water laundry detergent. The process involves starting with a host strain that is already efficient at producing the protein and then engineering improvements through DNA sequencing. The use of artificial intelligence (AI) is a newer development, and while it's not yet directly integrated into the infrastructure, it does play a role in analyzing and optimizing the data generated from the lab work. The ultimate goal is to enable direct access to Ginkgo's infrastructure for scientists at these companies, but that's still a work in progress. The company's focus on scale and innovation in lab work has led to significant advancements in the production of custom organisms for various industries.
Exploring AI in Biology: Ginkgo Bioworks and Google's New Venture: Ginkgo Bioworks and Google aim to create an AI model to understand and work with proteins, potentially outperforming humans in biology. The technology's significant potential lies in industries where large molecule development costs dominate, like agriculture and industrial processes.
Ginkgo Bioworks is exploring the use of foundation models in biology, aiming to create a model that can understand and work with proteins, going beyond their current focus on catalysis. This new idea, announced in a deal with Google, could potentially allow computers to outperform humans in the field of biology faster than in areas like language processing. The potential market size for this technology is significant, especially in industries like agriculture and industrial processes where a large portion of costs go into the development of the actual molecule. While the cost of developing a drug is high, most of the expenditures are not directly related to the molecule itself, making the pharmaceutical market a smaller potential market for this technology compared to industries where a larger portion of costs are dedicated to the molecule. Google, with its AWS-style business model, provides the necessary compute power for such endeavors. The potential for AI in biology to surpass human capabilities could lead to groundbreaking discoveries and innovations.
Biotech industry's unique challenges in adopting common software platforms: The biotech industry's diverse work and limited data hinder the adoption of common software platforms, but the potential for AI-discovered drugs could be unlocked with tech industry's commoditized infrastructure
The biotech industry, despite its size and importance, operates differently than the tech industry when it comes to software platforms and common infrastructure. The work being done in biotech is too diverse for common platforms due to the unique needs of each company or modality. However, the belief is that there are commonalities in engineering organisms that can be leveraged. The lack of AI-discovered drugs is due to the limited availability of data and the need for a breakthrough in neural networks. Companies like Recursion are trying to address this data gap, but it may still not be enough. The tech industry's commoditization of infrastructure could help overcome these challenges and lead to the discovery of AI-discovered drugs. Overall, the biotech industry's structure and the challenges it faces in adopting common platforms are vastly different from the tech industry.
Preparing for Future Pandemics: Rapid Response and Monitoring: Investing in rapid vaccine response and persistent monitoring can help prevent the spread of infectious diseases, as demonstrated by the success of Operation Warp Speed and the importance of early detection in preventing pandemics.
Ginkgo Bioworks, a leading biotech company, generates its data through its own 300,000 square foot robotic lab to serve customer projects. They have been particularly successful in the area of infectious disease research, emphasizing the importance of being prepared for future pandemics. The COVID-19 pandemic served as a reminder that modern healthcare systems do not make us immune to pandemic-scale infectious diseases. To mitigate this risk, efforts should focus on rapid vaccine response, such as Operation Warp Speed, and monitoring, or "bio radar," to detect and respond to potential outbreaks. The use of persistent monitoring and rapid response, similar to cybersecurity, can help prevent the spread of infectious diseases. The beauty and challenge of these organisms is that they replicate, so catching them early is crucial. The DOD maintains millisecond preparedness, and history shows that rapid response can effectively prevent the spread of diseases. While some diseases may be harder to contain than others, the logic remains the same: catching it early wins.
Investing in tech solutions for biology threats: Despite the low probability of a deadly biology virus being created using AI and LLMs, we should invest in detection and rapid response technologies, similar to our defenses against computer viruses. Global monitoring and rapid vaccine generation are also crucial approaches to pandemic response.
While the risk of a lone actor creating a deadly biology virus using AI and LLMs is currently low due to the difficulty in accumulating necessary data and expertise, we are unacceptably exposed to such threats and should invest more in technological solutions for detection and rapid response, similar to our defenses against computer viruses. The conversation also touched upon the importance of global monitoring and rapid vaccine generation as smart grounded approaches to pandemic response. Additionally, the unique approach of biotech startup Ginkgo, such as super voting shares for employees, was discussed as an example of innovative decision-making. Despite the challenges faced by the startup, Ginkgo's founders' determination and unfunded origins illustrate the importance of perseverance and thinking outside the box in the biotech industry.
Control and governance of powerful tech platforms raise ethical questions: Giving humans, rather than divorced capital, the power to make decisions in tech companies is a good starting point for ethical governance
The control and governance of powerful technological platforms, whether in the realm of biotech or social media, raises significant ethical questions. These platforms have the potential to significantly impact people's lives, and decisions about who should control them are crucial. The speaker from Ginkgo Bioworks shares their experience with this issue, discussing the tension between founders and capital markets, and the decision they made to give employees a greater role in the company's governance through supervoting shares. This approach, inspired by media companies, empowers the human workers who are most connected to the company and its mission. Ultimately, the question of who should control these platforms is a complex one, but the speaker argues that giving humans, rather than divorced capital, the power to make decisions is a good place to start.
Governance and Decision-Making for Hired Gun CEOs at The New York Times: CEOs at The New York Times face unique challenges in decision-making due to human control, potential popularity contests, and alignment with workforce. Share voting is proposed as a solution to create a more aligned workforce.
While the unique model of The New York Times being controlled by humans rather than capital is intriguing, it raises questions about governance and decision-making for hired gun CEOs. Unlike founders who have moral authority, CEOs may face challenges in making unpopular decisions and could potentially fall into popularity contests or trying to appease employees. The proposed solution is share voting, where employees accumulate more shares by working longer and building value, creating a more aligned workforce. However, the conversation also touched upon the efficiency of human evolution compared to advanced technology and the potential for exploring alternative architectures in AI. Ultimately, someone or something needs to govern these systems, and it's essential to consider the implications of various governance structures.
Learning from Evolution for Efficient Designs: Explore adaptive designs inspired by nature's evolution process, focusing on hyperoptimization and efficient exploration of body plans.
The evolutionary process, as seen in nature, holds valuable insights for creating more efficient and adaptable systems, especially when it comes to self-replicating systems. The concept of hyperoptimization for energetics and other factors arises naturally in such systems. This idea moves us beyond handcrafted designs and into an era of explosive growth, similar to the Cambrian explosion. The book "The Plausibility of Life" is a great resource for those interested in this line of thought. Instead of learning from brains, we should learn from evolution itself, as it has evolved things to be more evolvable. An example of this is the skeletal system, where skin and nerves are adaptive to bones, making for a more efficient exploration of body plans. With the vast amount of energy and time evolution has spent, it has discovered numerous solutions worth exploring for those designing alternative architectures. So, the advice from Jason is to wrap the "skin" around the "bones" in your designs, ensuring they are adaptive and efficient. This concept, proven effective over billions of years, is ripe for exploration in the field of artificial intelligence and system design.