Episode Summary
Executive Summary: Jack Hidary argues that AI and quantum physics are converging into “large quantitative models” that can model the physical world—especially molecules, materials, and biology—far beyond language models. Sandbox AQ uses quantum equations on GPUs to generate data for AI, enabling faster and more precise drug discovery, battery chemistry, materials design, and eventually better clinical prediction, while quantum hardware matures in the background.
Main Topics: AI vs. quantum: shared information-compression principles: The discussion frames both neural networks and quantum physics as systems that compress vast complexity into useful representations. AI compresses text/images into model weights; quantum equations compress physical dynamics into solvable mathematics. Limits of large language models: LLMs are powerful but constrained by their training corpora, making them probabilistic, prone to hallucination, and fundamentally limited to remixing existing internet data rather than discovering physical truths. Large Quantitative Models (LQMs) for the physical world: Sandbox AQ’s core idea is to use quantum physics equations to generate accurate simulation data, then train AI on that data to explore molecular and materials spaces that ordinary data-driven models cannot access. Drug discovery and precision medicine: Quantum-enabled modeling can help design molecules that bind specific biological targets, potentially reducing drug development time, cost, and failure rates, while enabling more personalized therapies. Materials, batteries, and energy applications: The same methods can be applied to battery chemistry, solar materials, and other engineered substances, allowing exploration of much larger chemical design spaces and better optimization of energy technologies. Quantum computing and hardware timeline: Hidary distinguishes Sandbox AQ from quantum-hardware companies: the company runs on classical GPUs today but is designed to benefit from future fault-tolerant quantum computers, likely emerging in modular form later this decade. Quantum sensing and near-term utility: Beyond computing, quantum sensors are already practical today for navigation, medicine, and environments where GPS is denied or degraded, representing an immediate commercial and societal impact.
Key Arguments: AI and quantum are similar because both model the world through compression and abstraction, turning massive complexity into manageable predictive structures. LLMs are limited because they depend on text corpora; they cannot reliably discover new molecules, battery chemistries, or physical phenomena absent from their training data. Quantum equations provide the most accurate way to generate data for physical systems, making them the best foundation for training models in chemistry, biology, and materials science. Sandbox AQ’s value proposition is not building drugs or hardware itself, but providing a software platform that computes physical systems at scale across many industries. Drug development is inefficient: it takes years, costs billions, and has very high failure rates; better simulation can improve hit rates before clinical trials. The near-term breakthrough is not necessarily a full quantum computer, but the ability to use classical GPU compute to solve useful quantum-physics problems today. When scalable, error-corrected quantum computing arrives, it will amplify existing methods, but meaningful progress is already happening now with classical computation plus quantum equations. Quantum sensing offers immediate benefits, unlike quantum computing which still needs major hardware scaling and error correction. The future of R&D will shift from trial-and-error exploration to simulation-driven design, with more targeted interventions and broader search spaces. Information theory is the unifying principle across AI, physics, and quantum science: the ability to compress, represent, and generalize complex reality.
Data Points: Alzheimer’s research progress: 40 years with nothing to show - Used to illustrate the scale of unmet medical need and the inadequacy of current discovery methods. Sandbox AQ funding: $500 million - Described as the company’s major financing round after spinning out of Alphabet. LLM parameter scale: 400-500 billion parameters; up to 1-2.5 trillion in newer models - Used to show how far language models have scaled. Brain neuron count: 86 to 100 billion neurons - Referenced while explaining artificial neural networks as brain-inspired systems. Drug development time: 7-10 years preclinical; 4-5 years minimum in clinic - Outlined as the current timeline for bringing a drug to market. Clinical trial failure rate: 90% - The proportion of drug candidates that fail after entering clinical trials. Drug effectiveness in patients: ~20% - Estimate given for how often an approved drug works for an individual patient. Battery chemistry options: ~19 elements - Mentioned as the design space for battery materials and electrolytes. Physical-to-logical qubit ratio: 1,000:1 - Used as a rough estimate for error correction overhead in quantum computers. Current quantum hardware scale: A few hundred physical qubits - Described as today’s rough state of the art across leading platforms. Needed logical qubits for cryptography cracking: ~5,000 logical qubits - Estimate for breaking RSA/ECC-class encryption. Physical qubits needed for that task: ~5 million physical qubits - Derived from the 1,000:1 error-correction ratio. Projected near-term quantum hardware milestone: 2029 - Predicted timeframe for modular blocks of 1,000 to 5,000 physical qubits. Projected practical quantum-computing era: 2031-2032 - Estimated time for engineering, error correction, and integration of modular systems. Fountain Life diagnostic data: 150 gigabytes - Used in sponsor message to describe the data volume processed by its health screening platform.
Pivotal Quotes: "Instead of the world of large language models, we've now entered the world, Peter, of large quantitative models, LQMs." — Jack Hidary: Core thesis of the episode: physics-based modeling plus AI. "The equations are the bedrock of the universe. They are the fundamentals. They're upon which everything is built and created." — Jack Hidary: Explaining why quantum equations are the right starting point for simulation-driven discovery. "We're actually exploring a bigger landscape." — Jack Hidary: Describing how quantum-enabled AI expands the search space for drugs and materials beyond brute-force experimentation.
Implications: If quantum-aware AI scales, R&D in drugs, batteries, materials, and sensing could become far faster, cheaper, and more precise. The winners may be software platforms that generate reliable simulation data, not only hardware builders or biotech labs.