a16z Podcast | A New Lab Rises

Lab Structure and Evolution

• Berkeley labs operate on a five-year timeline with specific visions, transitioning organically from the AMP Lab ("Algorithms, Machines, People," focused on big data analytics) to the RISE Lab ("Real-time, Intelligent, Secure Execution," focused on real-time decision-making).
• The RISE Lab was established to address the transition from generating insights to making intelligent, real-time, and secure decisions on live data.
• Key success factors for Berkeley labs include physical proximity to Silicon Valley, a culture of open-source innovation, and the absence of patents.

AMP Lab Achievements

• Spark: Developed as a big data execution engine that later became the foundation for the commercial company Databricks.
• Mesos: Designed as a resource management system to allow multiple cluster computing frameworks to share hardware; it evolved from Hadoop support to handling long-running services based on Twitter feedback.
• Tachyon: An in-memory storage engine that resulted in the founding of Eluxio.

Academic-Industry Collaboration Dynamics

• Projects are led by PhD students who often intern at companies (e.g., Facebook, Twitter) to ensure research addresses real-world production scale and problems.
• Industry partners, such as Twitter and Facebook, provided critical feedback that shaped the evolution of open-source projects like Mesos.
• The "tipping point" for open source was the ability to build commercial businesses on top of open-source software (SaaS/hosted offerings), moving it from a fringe academic activity to a mainstream industry engine.
• Large corporations (e.g., Microsoft, Google) have shifted from closed-source models to heavy open-source participation, driven by the need to crowdsource R&D.

RISE Lab Vision and Strategic Direction

• Core Goals: Enable real-time, personalized, secure, robust, and explainable decisions (the "holy grail" of AI).
• Robustness Requirements: Systems must handle noisy inputs, component failures, and unforeseen inputs (e.g., an AI recognizing an elephant as "unknown" rather than misclassifying it).
• Explainability: Critical for trust, particularly in high-stakes fields like medical diagnosis where users need to understand the "why" behind an algorithm's decision.
• Cloud-Edge Continuum: Developing systems that span both cloud and edge computing, allowing functionality to migrate bidirectionally between centralized data centers and edge devices (e.g., self-driving cars).
• Future of Computing: Moving beyond traditional "if/then" programming toward data-driven synthesis where AI agents interact with the real world to learn (e.g., reinforcement learning).

Specific RISE Lab Projects

• Ray: A cluster computing framework designed to simplify the building of next-generation AI applications, specifically for reinforcement learning and multi-agent environments.
• Clipper: A model serving platform addressing the challenges of lifecycle management, scaling, and updating AI models as data environments evolve.
• Opaque: A security project utilizing hardware enclaves (e.g., Intel SGX, ARM Trust Zone) to ensure data and computation remain secure even if the OS or hypervisor is compromised.
• Ground: A data provenance service designed to track data semantics, lineage, and ownership across multiple distributed storage systems.

Demographics and Market Context

• Over 60% of PhD applicants to Berkeley now apply for AI programs, reflecting a massive industry shift toward intelligence and decision-making.
• The lab collaborates with major partners including Capital One, Ericsson, Huawei, and Ant Financial (Alibaba) to solve cross-industry problems that single companies cannot address alone due to siloed structures.
• The naming of "RISE" was a deliberate branding exercise adhering to David Patterson's rule of creating four-letter acronyms starting with "R," following the pattern of "RAID" and "Frisco."

Community and Outreach

• The lab hosts bi-annual retreats and "camps" (e.g., AMP Camp) to foster community, offering tutorials on pre-alpha software to both students and industry practitioners.
• The philosophy emphasizes "open innovation," requiring collaboration outside company walls to drive success in the modern software landscape.