Google said its researchers cut errors roughly 40-fold on its Willow quantum processor in a study published in Science, among research findings the company announced during October.
The technique addresses a task that previously threatened to consume up to 95% of a quantum computer’s resources, Google said in its Nov. 2 publication, “The latest AI news we announced in October 2026.”
Google, which sells AI services and computing infrastructure, said its quantum researchers demonstrated a technique called “magic state cultivation” on Willow. The process produces special quantum states needed for reliable calculations.
The researchers compared errors in those states with errors in states produced without cultivation. Google described the result as removing a major obstacle to building larger quantum computers that can work reliably despite errors.
The company did not give the study’s testing period, number of trials or absolute error rates in the announcement. The roughly 40-fold improvement describes a comparison between the states, rather than the performance of an entire quantum computer.
Google also announced findings from research with Beth Israel Deaconess Medical Center into AMIE, its conversational medical AI system. Researchers evaluated the system during actual primary care clinic visits, the company said.
AMIE gathered patient histories and prepared clinical summaries before urgent care appointments. Google said the system helped clinicians prepare in 75% of cases and matched doctors’ final lists of possible diagnoses 90% of the time.
The company identified the clinical research as its first publication in the main journal of The Lancet. It did not give a patient count or evaluation period, or specify how researchers measured help with preparation.
Those findings concerned preparation and possible diagnoses. Google did not provide figures for treatment outcomes in the announcement.
Other scientific work included an experimental Geospatial Reasoning agent and public access to Deep Review, a tool for technical papers. Google said it presented both at its annual Research@ 2026 event.
Google.org, the company’s philanthropic arm, awarded $36 million to 17 organizations using AI for scientific research. Google also said its teams were mapping forest types around the world.
In a separate hardware experiment, Google said its first Project Suncatcher prototype satellite generated a haiku using Gemma 4 on a Google AI processor in low-Earth orbit.
The company said the result confirmed that the hardware survived launch and could run AI tasks in space. Further testing will examine how the chips withstand radiation and extreme temperatures as Google explores orbital computing infrastructure.
