At a Rise of AI Virtual Chat in 2020, host Fabian Westerheide talks with Claudia Pohlink of Deutsche Telekom's innovation labs about quantum machine learning. Pohlink, who leads the unit working on AI, blockchain and quantum computing at Telekom, explains her role as identifying trends and building a business around them rather than being a technical expert. She walks through the basics of qubits, superposition and entanglement, contrasts classical gate quantum computers such as IBM's with quantum-inspired hardware from Fujitsu and D-Wave, and argues the quantum business will be a cloud business. She cites Germany's roughly two-billion-euro investment in quantum hardware, Google's quantum-supremacy benchmark solving a task in 200 seconds versus 10,000 years for a classical supercomputer, and use cases in routing optimization, cryptography and quantum chemistry. Audience questions cover quantum entanglement, whether quantum machine learning deepens AI's black-box problem, and where to follow the field. Pohlink argues quantum computing may eventually help explain deep learning rather than obscure it further, while stressing that most current knowledge still comes from academic papers rather than mature products.
Key takeaways
- Pohlink describes her role at Deutsche Telekom as identifying trends and building a business around AI, blockchain and quantum computing, not as a technical expert herself.
- Germany has committed roughly two billion euros to develop quantum computing hardware domestically, aiming to catch up with US and Asian players.
- Google's quantum processor solved a benchmark task in about 200 seconds that Pohlink says would take a classical supercomputer roughly 10,000 years.
- Pohlink expects the near-term value of quantum computing to lie in routing optimization, cryptography and quantum chemistry rather than general-purpose computing.
- Pohlink argues quantum machine learning could eventually help explain deep learning's black box rather than make it harder to understand.
Detailed notes
Claudia Pohlink's path into quantum computing at Deutsche Telekom
Fabian Westerheide introduces Claudia Pohlink as head of the unit at Deutsche Telekom's innovation labs working on AI, blockchain and quantum computing, describing it as the German equivalent of Google X. Pohlink says she is not a technical expert but manages a team of experts, and her job is to identify trends, turn them into a business, manage and promote them. Her background is in business and marketing; she says she moved toward data-driven fields and studied data science. She learns mainly by doing, by exchanging with experts inside and outside the company, and by visiting IBM's quantum computer lab in the US, which she found impressive though visitors are not allowed inside the lab itself.
Quantum hardware and the case for a cloud business
Pohlink distinguishes classical gate quantum computers such as IBM's Q system from quantum-inspired hardware such as Fujitsu's Digital Annealing Unit, which simulates quantum computing on a chip, and from D-Wave's early quantum annealing processors, alongside Google's processor. She notes that building quantum computers is still very expensive, citing a pitch she saw as an investor for a startup building qubit-based hardware, and states her opinion that users will not need to buy quantum hardware themselves because the quantum business will be a cloud business. She also mentions Germany's roughly two-billion-euro government investment aimed at developing quantum hardware domestically, since the leading players are currently in the US and Asia.
Qubits, superposition and entanglement
Pohlink explains that unlike a classical bit, which stores either zero or one, a qubit can hold both states at the same time through superposition, while entanglement describes the relation between qubits so that acting on one affects another, multiplying computational power. She compares this to a spinning coin showing both faces at once until it is measured and stops. Asked directly by an audience member what quantum entanglement is, she says she cannot fully explain it, describing qubits as connected by something like a cable so that they do not work independently, and notes that a quantum system's state is only fixed once it is measured.
Why classical computers hit their limits
Pohlink argues that Moore's law is ending because transistors can no longer be shrunk indefinitely without heat and cooling problems, and that some computations, such as combinatorial optimization, grow too complex for classical computers. She illustrates this with the legend of doubling grains of rice on a chessboard, reaching quintillions of grains, and with the example that seating ten people at a table already has 3.6 million possible arrangements. She connects this to cryptography: prime factorization, the basis of modern encryption, is hard for classical computers but could in principle be broken quickly with a sufficiently powerful quantum computer, a method she attributes to Peter Shor's 1994 work.
Quantum supremacy and use cases
Pohlink cites Google's 2019 quantum-supremacy result, where a specific benchmark task took a quantum processor about 200 seconds and roughly half a euro of electricity, against about 10,000 years for a classical supercomputer. She groups quantum applications into three areas, quantum communication, quantum sensing and quantum computing, and highlights routing optimization, drawing on Telekom's own network problems and traffic-routing examples, as well as quantum chemistry, where she says the chemical industry is investing heavily to simulate molecules and design new materials and drugs.
Q&A highlights
Asked whether quantum machine learning will make AI even more of a black box, Pohlink says deep learning is already hard to explain, and argues quantum computing might instead help explain AI, since researchers are exploring quantum versions of deep-learning models. She adds that familiarity with Python and machine learning is a good starting point for working with quantum computers. On where to learn more, she says most substantive material still comes from papers and conversations with vendors such as IBM, D-Wave and Fujitsu, since quantum computing remains largely a scientific topic. Westerheide closes by previewing the next Virtual Chat, on face recognition and surveillance with the head of TrueFace (the transcript gives the speaker's name inconsistently as both "Sean Moore" and "Sean Morris").
Quotes
If you measure them, they exist. If you don't measure them, you don't know if they exist in the state or not.
Claudia Pohlink
I would say it will be helpful to understand all the stuff behind, which is not really explainable right now.
Claudia Pohlink
Questions answered in this talk
What is quantum entanglement?
According to Claudia Pohlink at a Rise of AI Virtual Chat in 2020, entanglement means that qubits become connected so that acting on one affects another, which multiplies computational power; a system's exact state is only fixed once it is measured.
Why can quantum computers solve problems classical computers cannot?
According to Claudia Pohlink at a Rise of AI Virtual Chat in 2020, Moore's law is reaching physical limits and some problems, such as combinatorial optimization or prime factorization used in encryption, grow too complex for classical computers, while quantum hardware can compute many states simultaneously through superposition and entanglement.
What are current use cases for quantum computing?
According to Claudia Pohlink at a Rise of AI Virtual Chat in 2020, the main use-case areas are routing and logistics optimization, cryptography, and quantum chemistry, including simulating molecules for new materials and drug discovery.