851: Quantum ML: Real-World Applications Today, with Dr. Florian Neukart

Episode Summary

Executive Summary: Florian Neukart argues that quantum machine learning is already useful when built as a hybrid with classical HPC, not as a replacement. He explains Terra Quantum’s platform approach, real-world optimization and simulation use cases, hardware constraints like coherence and connectivity, the promise of post-quantum security, and future applications ranging from drug design to fusion and space tech.

Main Topics: Hybrid quantum-classical computing: Neukart’s central thesis is that quantum value comes from integrating quantum processors into classical workflows rather than trying to run entire problems natively on quantum hardware. Practical applications today: He describes current value in optimization, machine learning, and simulation, especially for scheduling, logistics, chemistry, and finance. Quantum hardware and chip selection: The discussion covers current hardware ecosystems, including IBM, Google, Rigetti, D-Wave, AWS, and NVIDIA, plus the importance of error correction, coherence, and connectivity. Quantum-safe security: He explains the emerging security stack: post-quantum cryptography, quantum key distribution, and quantum random number generation. Terra Quantum’s platform and go-to-market: Neukart outlines Terra Quantum’s software stack, APIs, TQ42 platform, and strategy of abstracting chip complexity from end users. Future frontier applications: He highlights long-term opportunities in personalized medicine, battery chemistry, fusion, and even space elevators as quantum-enabled breakthroughs. How to get started: He emphasizes that a physics degree is not required for quantum ML; data scientists and ML practitioners can begin with Terra Quantum’s documentation and TQ Academy.

Key Arguments: Hybrid quantum computing is already practical when quantum resources are used only where they add value, while classical HPC handles the rest. Quantum ML should not attempt to convert whole neural networks into quantum circuits; only selected parts should be hybridized. Real-world optimization problems such as scheduling and logistics are already suitable targets because classical methods rely on heuristics and approximations. Quantum simulation is especially promising for chemistry and drug discovery because these problems are fundamentally quantum mechanical. Hardware progress depends on better coherence, fidelity, connectivity, and error correction, not just more qubits. Google’s Willow result matters because scalable error reduction is a key blocker for larger quantum systems. Quantum-safe security will likely combine post-quantum cryptography, quantum key distribution, and quantum random number generation. Enterprises need agnostic cloud-based software because most quantum hardware will remain remote and expensive to own directly. People with ML and engineering backgrounds can begin using quantum ML now; building chips is the part that truly requires physics expertise.

Data Points: Episode number: 851 - Super Data Science Podcast episode identifier Past related episode: 721 - Earlier podcast episode with Amira Abbas on quantum machine learning Quantum circuit runtime on chip: Nanoseconds to microseconds - Neukart describes the short time window available to run circuits before decoherence/collapse Qubit scaling effect: 2 to the power of n - He explains that each added qubit doubles the configuration space in an ideal quantum system Example with 2 bits: 4 possible configurations - Illustration of quantum superposition scaling Example with 3 bits: 8 possible configurations - Illustration of quantum superposition scaling Repetition for probabilistic readout: About 1,000 measurements - He says results are typically sampled many times and majority vote is used Two-qubit fidelity: 99.10.7% - He cites a very high example fidelity value when discussing hardware error rates (as spoken in the transcript) Quantum computer vs classical RSA attack: Less than 2 minutes vs 1 billion years - Nadella comparison cited for RSA 2048 Classical supercomputer comparison in Google marketing example: 10 septillion years - Used as an example of quantum advantage messaging in the transcript Superconducting chip cost: $15–20 million - Approximate cost mentioned for local superconducting quantum systems NIST post-quantum standards: 3 algorithms - Neukart notes NIST recently released three standards/algorithms for post-quantum cryptography TQ Academy: Training program - Terra Quantum’s educational offering for universities and companies TQ42: Platform name - Terra Quantum’s developer platform for getting started

Pivotal Quotes: "We don't take, for example, an entire neural network and try to express it as something that can be run on a quantum chip. We take only parts of it." — Florian Neukart: Explaining the hybrid approach to quantum machine learning "If you have a background already in machine learning, then perfectly equipped already to do quantum computing." — Florian Neukart: On who can get started with quantum ML without needing a physics degree "The barrier is not in the era anymore." — Florian Neukart: Describing Google Willow’s significance for scalable quantum error correction

Implications: Quantum ML is moving from theory toward enterprise use in targeted hybrid workflows. The near-term winners will be teams that combine classical ML/HPC with quantum tools, while preparing for quantum-safe security and new R&D frontiers in chemistry, medicine, and energy.