AI’s new role in science: Supercomputer Emulator

Bishop: They are still computationally expensive. Moreover, because emulators are trained using data, they tend to be no more accurate than their training data. They may also give inaccurate results when faced with scenarios that differ significantly from the ones they were trained on.

\”I am a believer in the \”use-inspired fundamental research \”–[like] Pasteur’s work. He was a consultant to the brewing industry. Why was this beer always souring? \”He basically invented the field of microbiology.\” – Chris Bishop, Microsoft Research

Source:
https://spectrum.ieee.org/ai-for-science?share_id=7196524&socialux=facebook&utm_campaign=RebelMouse&utm_content=IEEE+Spectrum&utm_medium=social