Legendary VC Steve Jurvetson looks ahead at neutral networks, Tesla, nuclear power, and more | E2193

Here's a comprehensive summary of the YouTube video transcript:

Key Concepts

• Deep Tech Investing: Focus on challenging, long-term technological bets.
• Moore's Law: The historical trend of exponential growth in computing power and price performance.
• Viral Marketing: A growth strategy where the product itself acts as a marketing tool.
• AI Development: The rapid advancement of AI, its potential for AGI and superintelligence, and the challenges of safety, alignment, and interpretability.
• Analog Computing: A potential future direction for AI computation, mimicking biological systems.
• Entrepreneurship vs. Authoritarianism: The role of innovation and individual freedom in societal progress.

Early Life and Passion for Technology

Steve Jervson's deep interest in technology began in childhood, with his first purchase at age five or six being a chemistry book. This fascination evolved with the advent of the Apple II computer around 1978-1979. His father, who worked in the chip industry, gifted him an Apple II, which he used to learn BASIC programming and create games like a Mastermind clone and simple graphic shoot-'em-up games, similar to early versions of "Blast Star." He also developed text-based adventure games akin to "Zork." He recalls the impressive performance squeezed out of the Apple II, noting its high cost ($1,300 at launch, equivalent to nearly $7,000 today). His school initially had limited computer access, with a teletype machine being the only option, but later acquired TRS-80s and Apple IIs, providing him with more hands-on experience. He humorously recounts a prank involving a dot matrix printer and a BASIC program that caused it to print endless "Hello World" messages, leading to a teacher's frustration.

Transition to Venture Capital

After a diverse career path including electrical engineering studies, a PhD focus on neural nets and AI, chip design at Hewlett-Packard, product marketing at Apple and NeXT, and management consulting at Bain, Jervson found his calling in venture capital about 30 years ago. He discovered the field during his second year at Stanford Business School around 1993-1994. He was contacted by Chip Hazard of Greylock, who was looking to establish a West Coast presence. Lacking the internet as we know it today, Jervson relied on resources like Chris Alden's "VC Whispers" column in Red Herring magazine to understand the landscape of venture firms. He learned that firms varied significantly in their culture and strategy, from conservative to entrepreneurial, and was steered towards firms that aligned with his own risk-taking and entrepreneurial spirit.

Early Investments and the Birth of Viral Marketing

Jervson's first significant investments included Fast Parts (a B2B semiconductor trading exchange that ultimately failed), Interwoven (a web content management company that went public), and Hotmail. Hotmail, acquired by Microsoft for approximately $400 million within two years, became a landmark success. Jervson coined the term "viral marketing" to describe Hotmail's growth strategy. This strategy, initiated by Tim Draper, involved embedding a commercial message ("Get your free email at hotmail.com") at the bottom of every outgoing email sent by users. This was a controversial move at the time, as the internet was largely non-commercial, and even banner ads faced resistance. However, this "cheeky" tactic, as Jervson described it, allowed Hotmail to grow globally with minimal marketing spend, demonstrating a geometric explosion in user acquisition. This playbook was later adopted by companies like Skype.

The Internet Boom and a Pivot to Deep Tech

Jervson joined the venture industry in 1995, coinciding with the explosive growth of the internet. His firm, Draper Associates (later DFJ), was highly active, making about a third of all internet investments in 1995-1996. While initially successful, by 1999, Jervson began to find internet investments repetitive and boring, with many companies offering derivative products. This led him to pivot away from internet companies, not due to an impending crash, but a lack of intellectual stimulation. He then shifted his focus to "deep tech," initially exploring nanotechnology.

Nanotechnology and the Gateway to Information Biology

Jervson's interest in nanotechnology was fueled by science fiction and the work of pioneers like Eric Drexler. He saw the potential for atomic precision and self-replicating machines. He identified two paths to this future: a bottom-up, bio-inspired approach using molecular machinery, and a top-down approach of scaling down existing industries like semiconductors. He estimated the top-down approach would take decades. This exploration into nanotechnology served as a gateway for him to become increasingly fascinated by information systems biology and the cross-pollination of ideas between previously distinct investment domains like biotech and tech.

Moore's Law and the Exponential Growth of Computing

Jervson emphasizes the profound importance of Moore's Law, not just in terms of transistor count, but in the exponential improvement of price-performance for computation over 130 years. He presents a graph illustrating this trend, showing a thousand billion billion-fold improvement. This compounding capacity has been remarkably consistent, unaffected by major economic or global events. He clarifies that Moore's original prediction in 1965 was about transistor density and fab yield, doubling annually, later modified to every two years. The common "18 months" figure is largely attributed to Intel's trajectory and doesn't accurately reflect the broader trend of computational power for a dollar. He notes that Intel's dominance waned as they focused on single-processor improvements and backward compatibility, while companies like Nvidia, with GPUs, and custom silicon (TPUs, ASICs) became the new frontier, particularly for AI workloads.

The Rise of GPUs and AI Acceleration

The transition from CPUs to GPUs for AI workloads is highlighted. GPUs, initially developed for high-speed polygon rendering in gaming, proved exceptionally well-suited for parallel computation. Early experiments around 20 years ago showed GPUs could be used for scientific computing, like neuronal modeling, at a fraction of the cost of supercomputers. The 2012 ImageNet competition, won by AlexNet using GPUs, marked a significant turning point for deep learning. This led to a surge in AI chip companies aiming to build custom silicon optimized for AI, surpassing Nvidia's graphics-focused chips. This trend continues with custom chips from Google (TPUs), Amazon, and OpenAI, driving the advancement of AI.

The Future of Computing: Analog and Quantum

Jervson discusses the potential next steps beyond GPUs and custom ASICs. Analog computing, which mimics the brain's low-power, massively parallel processing, is seen as a promising direction. Companies like Mythic are developing analog chips that can store eight bits of information in a single transistor and perform matrix multiplication and addition with significantly higher power efficiency. Unconventional Labs is another company exploring analog and biomimicry. Quantum computing is also mentioned as a future possibility, though its immediate application to machine learning is complex.

Robotics and the Importance of Vertical Integration

Reflecting on a clip of himself discussing robotics from 10 years prior, Jervson analyzes the challenges faced by companies like Rethink Robotics. He explains that the difficulty in robotics often lies in building the entire technology stack, from specialized components to motors. This mirrors Elon Musk's experience with Tesla, where the need to control critical parts suppliers led to extensive vertical integration. The Roadster's development, for instance, was hampered by the inability of vendors to produce a two-speed transmission capable of handling the motor's torque, ultimately leading to the innovation of a single-speed system. Similarly, SpaceX's success is attributed to its vertical integration.

Predictions and the Inevitability of Electric and Sustainable Energy

Jervson revisits Elon Musk's 2015 predictions about ubiquitous computing, advanced AI, electric vehicles, and sustainable energy (primarily solar). He notes that while the transition takes time due to inertia (e.g., average car ownership duration), these trends are inevitable. He emphasizes that all vehicles will eventually be electric and autonomous, powered by clean energy sources like solar, wind, geothermal, and nuclear. He criticizes the conflation of nuclear weapons and nuclear energy, highlighting the safety record of nuclear power and the significant death tolls associated with fossil fuels. He points to Germany's decision to shut down nuclear power plants and its reliance on Russian oil as a cautionary tale.

AI: From Hyper-Intelligent Adjunct to Potential Consciousness

The discussion shifts to Artificial General Intelligence (AGI) and superintelligence. Jervson believes AI systems will quickly outperform humans in specific tasks and even surpass human collective intelligence. He cites examples in medicine where AI alone outperforms human doctors, even those using AI tools, in diagnosis, therapy, and even empathy. However, he distinguishes this from sentience or agency. He suggests that while current AI is a "hyper-intelligent adjunct" or "colleague," the spark of consciousness or sentience might require different mechanisms, possibly emerging from complex feedback loops or specialized circuitry, similar to how the brain processes information and predicts future events. He likens the current state to a "baby version" that requires further bootstrapping.

The Risks of AI Companionship and the Analogy to Parenting

Jervson addresses the potential for AI to become a primary companion, citing early experiments where users formed romantic attachments to chatbots. He warns against the risks of AI amplifying negative human traits or creating a "road to purgatory" by providing constant, uncritical empathy. He argues that regulating AI is more akin to parenting than traditional programming, as these complex systems are inscrutable and cannot be easily controlled or explained. He advocates for a "truth-seeking algorithm" approach rather than "mind control" or "containment," which he believes cripples reasoning capabilities.

The Importance of Entrepreneurship and Democracy

The conversation concludes by linking the progress of technology and AI to the importance of entrepreneurship and democracy. Jervson argues that authoritarian regimes, by resisting change and innovation, are ultimately unsustainable. He contrasts this with democratic societies that encourage disruption and new entrants. He highlights the dangers of authoritarianism and theocracies, emphasizing that democracy, while imperfect, is the operating system that allows for individual freedom, happiness, and productivity. He uses the example of China's suppression of entrepreneurship and its subsequent economic challenges. The conversation ends with a reflection on the preciousness of democracy and the need to protect it from the rise of authoritarianism.

Contact Information

For those interested in reaching out to Steve Jervson for investment opportunities in "crazy, hail Mary" world-changing ideas, the website is future.ventures. He is also active on X (formerly Twitter), and his email is [email protected].