Bosch Presented AI Young Researcher Award

Bosch has presented the AI Young Researcher Award, endowed with 50,000 euros, for the first time. Gergely Neu, scientist at Pompeu Fabra University in Barcelona, won the jury over with his basic research in reinforcement learning (RL), an area of AI.

Company's CDO aqn CTO Michael Bolle congratulated the 34-year-old honoree on his excellent achievements as he presented him with the award in Renningen. “This award is a way for the Bosch Center for Artificial Intelligence to recognize the exceptional achievements of young researchers in artificial intelligence,“ Bolle said. “Gergely Neu’s research plays a major role in making AI more robust, more reliable, and more understandable.“

The five-person jury of researchers from academia and industry reviewed submissions from across Europe and honored Neu’s application as the most promising, not least due to his research on probability theory. Neu’s work focuses on known “multi-armed bandit problems“ through which algorithms learn to find their way through countless situations that can be combined in myriad ways.

Neu considers a two-way exchange between academic and industry AI research to be essential, as mutual benefitting is the only way for each area to expand its knowledge.“In recent years, many talented AI researchers have been leaving academia for lucrative jobs in industry, so prizes like the Bosch AI Young Researcher Award play an important part in increasing the prestige of traditional academic careers,“ Neu said.

The winner plans to invest the 50,000 euro prize money towards amplifying his group’s current collaborations and creating new ones by inviting guest researchers to his laboratory and by enabling his own team to be visiting researchers at other laboratories and to attend conferences. For more than ten years, the Hungarian scientist has been conducting research in the field of reinforcement learning. He investigates at what point existing RL algorithms reach their limits and why, using his findings to develop robust algorithms that perform reliably.

In AI applications such as highly automated vehicles, automated financial trading systems, or intelligent power grids, reliability and stable performance are essential. Together with his team, Neu develops algorithms under conditions that are as realistic as possible: he exposes them to new environments and unknown situations, asks them to make vast numbers of decisions, and then examines the factors that led to success or failure. Unlike most empirical research in RL, Neu has decided to try to understand RL’s underlying logic and functions in order to pave the way for robust generalizing methods. Solving this would ultimately advance all fields of application.