Using A.I. to Speed the Process of Scientific Discovery

Global Rush to Fund A.I.
March 10, 2020

Using A.I. to Speed the Process of Scientific Discovery

Running experiments with several variables often requires tiny, methodical tweaks to measurements, materials and inputs.

Running experiments with several variables often requires tiny, methodical tweaks to measurements, materials and inputs.

Graduate students might spend hundreds of tedious hours making small adjustments again and again until a solution is found—a waste of their cognitive abilities.

Increasingly, A.I. systems are being used in research labs to speed the process of scientific discovery.

Materials scientists at the University of British Columbia used a robot overseen by an A.I. algorithm to rapidly test a new kind of solar cell and log results. Based on what each experiment revealed, an algorithm determined what to change next. A process that might have taken 9–12 months was completed in five days.

Google’s DeepMind developed a way of testing and modeling the complex folding patterns of long chains of amino acids, solving a problem that has vexed scientists for many years. DeepMind’s system, AlphaFold, will allow scientists to synthesize new drugs to treat diseases and develop enzymes that might someday break down pollution.