Hartmut Neven
German scientist
- 4 min read
- 2 chapters
Related on Lably
Through Computer vision, Robotics and Google
Education
Hartmut Neven studied Physics and Economics in Brazil, Köln, Paris, Tübingen and Jerusalem. He wrote his Master thesis on a neuronal model of object recognition at the Max Planck Institute for Biological Cybernetics under Valentino Braitenberg. In 1996 he received his Ph.D. in Physics from the Institute for Neuroinformatics at the Ruhr University in Bochum, Germany, for a thesis on "Dynamics for vision-guided autonomous mobile robots" written under the tutelage of Christoph von der Malsburg. He received a scholarship from the Studienstiftung des Deutschen Volkes, Germany's most prestigious scholarship foundation.
Sign in to keep reading
One more chapter is free with a Lably account.
Work
In 1998 Neven became research professor of computer science at the University of Southern California at the Laboratory for Biological and Computational Vision. In 2003 he returned as the head of the Laboratory for Human-Machine Interfaces at USC's Information Sciences Institute.
Computer vision and machine learning
Face recognition, avatars and face filters
Neven co-founded two companies, Eyematic for which he served as CTO and Neven Vision which he initially led as CEO. At Eyematic he developed face recognition technology and real-time facial feature analysis for avatar animation. Teams led by Neven have repeatedly won top scores in government sponsored tests designed to determine the most accurate face recognition software. Face filters, now ubiquitous on mobile phones, were launched for the first time by Neven Vision on the networks of NTT DoCoMo and Vodafone Japan in 2003. Neven Vision also pioneered mobile visual search for camera phones. Neven Vision was acquired by Google in 2006.
Object recognition and adversarial images
At Google he managed teams responsible for advancing Google's visual search technologies. His team launched Google Goggles now Google Lens. The concept of adversarial patterns originated in his group when he tasked Christian Szegedy with a project to modify the pixel inputs of a deep neural network to lower the activity of select output nodes. The motivation was to use this technique for object localization which did not work out. But the idea gave rise to the fields of adversarial learning and DeepDream art. In 2013 his optical character recognition team won the ICDAR Robust Reading Competition by a wide margin and in 2014 the object recognition team won the ImageNet challenge.
Google Glass
Neven was a co-founder of the Google Glass project. His team completed the first prototype, codenamed Ant, in 2011.
Quantum Artificial Intelligence
In 2006 Neven started to explore the application of quantum computing to hard combinatorial problems arising in machine learning. In collaboration with D-Wave Systems he developed the first image recognition system based on quantum algorithms. It was demonstrated at SuperComputing07. At NIPS 2009 his team demonstrated the first binary classifier trained on a quantum processor.
In 2012 together with Pete Worden at NASA Ames he founded the Quantum Artificial Intelligence Laboratory. In 2014 he invited John M. Martinis and his group at UC Santa Barbara to join the lab to start a fabrication facility for superconducting quantum processors. The Quantum Artificial Intelligence team performed the first experimental demonstration of a scalable simulation of a molecule.
In 2016 the team formulated an experiment to demonstrate quantum supremacy. Quantum supremacy was then declared by Google in October 2019.
In 2023 Quantum AI researchers demonstrated that quantum error correction works in practice by showing for the first time that the error of a logical qubit decreases when increasing the number of physical qubits it is composed of.
Google's quantum processors have been used to study the physics of quantum many body states that otherwise are challenging to prepare in a laboratory such as time crystals, traversable wormholes and non-Abelian anyons.
In 2024, Google Quantum AI unveiled the Willow processor, a 105-qubit superconducting quantum chip. With Willow, the team realized a milestone in quantum error correction by achieving performance well below the surface code threshold, demonstrating that errors can be suppressed exponentially as the number of qubits is scaled. On the random circuit sampling benchmark, Willow completed the task in five minutes, running 1025 times faster than the world's fastest classical supercomputer.
Subsequently, in 2025, the team demonstrated the first verifiable quantum advantage utilizing the "Quantum Echoes" algorithm. Executed on the Willow processor, the algorithm computes Out-of-Time-Ordered Correlators (OTOC) to measure quantum dynamics and many-body interference. The computational task was completed approximately 13,000 times faster than the fastest known classical algorithm on the Frontier supercomputer, marking a step toward practical applications such as modeling molecular structures in nuclear magnetic resonance (NMR) spectroscopy.
Neven's law
Neven's law states that the performance of quantum computers improves at a doubly exponential rate.