Guillaume Verdon: Beff Jezos, E/acc Movement, Physics, Computation & AGI | Lex Fridman Podcast #407

Added: Dec 30, 2023

Verdon's Background and Philosophical Views

Verdon's journey began with a desire to understand the universe and develop a theory of everything. This led him to study theoretical physics, information theory, and quantum computing. He realized that nature operates on different scales, from quantum mechanics to thermodynamics, and he sought to understand and simulate these processes.

He also discusses the relationship between speech and thought and the importance of freedom of speech for fostering freedom of thought. He believes that anonymity and pseudonyms are crucial for maintaining an efficient marketplace of ideas.

The Merger of Identities

Verdon shares his experience of being doxed, where his anonymous identity as Beff Jezos was linked to his real identity. He expresses concern about the potential suppression of speech and ideas if individuals who challenge the status quo are exposed. He also discusses the impact of technology on anonymity and the potential for the proliferation of bots and fake accounts.

Effective Accelerationism and AI Safety

Verdon explains the concept of effective accelerationism and its focus on maintaining variance and adaptability in the development of AI. He argues that the market is more efficient at achieving AI safety through reliability engineering than heavy-handed regulations. He also emphasizes the importance of open-source AI and the equitable distribution of AI capabilities to prevent monopolies and maintain a near equilibrium of capabilities.

He also addresses the potential risks of rapid AI development, such as the centralization of supply chains for hardware and the need to consider the geopolitical implications of AI development. He advocates for a balance between technological progress and safety measures, emphasizing the need for diverse voices and contributions in the AI research and development space.

Exploring Quantum Intelligence

Verdon delves into the concept of quantum intelligence, which involves embedding AI algorithms into the physical world and understanding the quantum mechanical aspects of nature. He highlights the potential of quantum machine learning to capture quantum information from the world in superposition without destroying it. This allows for the acquisition of quantum information and the development of compressed representations with useful information in their latent form.

Universal Training Algorithm for Quantum Deep Learning

Verdon co-authored a paper on a universal training algorithm for quantum deep learning, which involves backpropagation with a quantum twist. This algorithm aims to adjust parameters in quantum machine learning by inducing phase kicks and forces to emulate the dynamics of a quantum particle moving in N dimensions. The advantage of this approach is its potential to tunnel through the landscape and find new optima that may be difficult for stochastic optimizers. However, he acknowledges that the current architectures for quantum computers make such algorithms extremely expensive to run.

Challenges in Quantum Computing

Verdon discusses the challenges of building a quantum computer, highlighting the need for interdisciplinary teams to tackle the complex problems in quantum computing. He explains that the engineering limits in quantum computing make it extremely difficult to scale up each component to a sufficient quality for error correction. The process of encoding information within a code within a code within a code within a code requires a lot of redundancy and an algorithmic refrigerator to pump out entropy from the subsystem representing the logical qubits.

The Future of Quantum Computing

Despite the challenges, Verdon remains optimistic about the potential of quantum computing and quantum machine learning. He emphasizes the need for cross-disciplinary collaboration and communication to push the boundaries of quantum intelligence. He also hints at the broader physics-based AI that his company, Extropic, is exploring as an alternative to quantum computing.

Implications for AI and Humanity

Verdon reflects on the implications of AI and quantum intelligence for humanity, emphasizing the need to explore a wider space of intelligences inspired by physics rather than the human brain. He challenges the anthropocentric view of artificial general intelligence and advocates for a broader understanding of intelligence that goes beyond human-like capabilities. He also discusses the potential risks and benefits of AI, highlighting the importance of aligning superintelligences with human values and the need for responsible development and deployment of AI.

Quantum Computing and Research

Verdon explains the significance of having sufficient quality and control over quantum bits in order to make scaling up quantum computing worthwhile. He discusses the long engineering process and the recent milestones achieved in quantum computing. He also shares insights from his research on quantum energy teleportation and the challenges of understanding the energy and entanglement of the vacuum.

Effective Accelerationism Movement

Verdon describes the Effective Accelerationism movement as a metahuristic framework that aims to promote adaptability and growth in the human techn capital mimetic machine. He discusses the overarching goals of the movement, which include increasing energy harnessing, promoting human flourishing, developing artificial general intelligence, and advancing interplanetary and interstellar transport. Verdon emphasizes the importance of embracing malleability and adaptability in culture and encourages diverse thought and independent agency.

Cult and Humor in the Movement

Verdon addresses the perception of the Effective Accelerationism movement as a cult and the role of humor and memes within the movement. He explains that the use of humor and memes was initially a strategy to grow the movement under the radar. However, he acknowledges the potential tension between embracing absurdity and maintaining the rigor of discourse. Verdon also discusses the concept of "mimetic optimism virus" and the decentralized nature of the movement, allowing for diverse interpretations and forks.

Elon Musk and Jeff Bezos

Verdon expresses admiration for billionaire entrepreneurs like Elon Musk and Jeff Bezos, highlighting their contributions to technological innovation and space exploration. He acknowledges Musk's concerns about AI safety and the need for regulations in the AI industry. Verdon also mentions the potential role of Neuralink in addressing AI safety concerns and the importance of maintaining adversarial equilibria in the AI landscape.

AI Safety and Centralized Control

Verdon emphasizes the importance of AI safety and the urgency of making progress in the field. However, he expresses concerns about the potential misuse of AI safety as a means to gain centralized control. He advocates for maintaining the ubiquity of opportunity to contribute to AI and the importance of balancing progress and safety in the AI industry.

Physics-based Computing and AI

Verdon explains that Extropic was born out of his dissatisfaction with the quantum computing roadmap and the pervasive problem of noise in quantum computing. He discusses the challenges of maintaining a pristine environment for quantum mechanics to take effect and the high cost associated with this constraint. He emphasizes the importance of using physics to engineer the ultimate physical substrate for generative AI from first principles of physics, information theory, computation, and thermodynamics. Verdon describes Extropic's focus on building a physics-based computing system and physics-based AI algorithms that harness out-of-equilibrium thermodynamics to perform machine learning as a physical process.

Tensorflow Quantum and Differentiable Programming

Verdon discusses his experience with Tensorflow Quantum, an attempt to combine deep learning and differentiable classical programming with quantum computing. He explains the challenges of differentiating and backpropagating through hybrid graphs and the interdisciplinary effort required to build such a system. Verdon emphasizes the importance of teaching and learning from one another in interdisciplinary teams and the value of collaboration in pushing the boundaries of knowledge.

Advice for Building a Team

Verdon shares his insights on building a successful team, emphasizing the importance of hiring individuals with high fluid intelligence and a broad expertise. He highlights the value of interdisciplinary collaboration and the need for team members to be open-minded and hungry to learn. Verdon encourages individuals to follow their curiosity, explore ideas, and not be constrained by traditional boundaries in their fields of work.

Effective Altruism and Effective Accelerationism

Verdon discusses effective altruism and effective accelerationism, highlighting the differences in their approaches to doing good and optimizing outcomes. He acknowledges the pros of effective altruism in trying to do good optimally but raises concerns about the potential for power and capital acquisition leading to corruption. Verdon encourages individuals to have an experimental mindset and explore different identities through alt accounts to engage in low-stakes experimentation with ideas and beliefs.

Productive Work Day and Personal Identity

Verdon shares his daily regimen, including his preferred work hours, diet, and cognitive performance enhancers. He emphasizes the benefits of ketones and caffeine in enhancing clarity of thought and focus. Verdon also discusses his experience with powerlifting and its impact on neural adaptation and cognitive performance.

Meaning of Life and Mortality

Verdon delves into the meaning of life, attributing it to thermodynamics and the evolutionary algorithm that drives adaptation and optimization. He discusses the importance of death in maintaining adaptation, novelty, and malleability within systems. Verdon expresses his belief in the value of mortality and the potential for new and improved versions of oneself in the future. He also shares his perspective on the quest for quantum gravity and its role in understanding the world, predicting outcomes, and steering through the challenges of a changing landscape.