The company is planning to apply an algorithm that is based on one of its core technologies called AlphaGo — software that beat top human players in a strategy game called Go — to ultimately build drugs.
What’s more, the technology needed to fold the proteins doesn’t require the same computing power of the version that beat the world champion Go player. Instead, researchers are basing their healthcare application off AlphaGo Zero, which used one-twelfth of the power of AlphaGo.
Unlike AlphaGo, AlphaGo Zero learned from scratch how to play the game. DeepMind’s latest project is using these same algorithmic principles in protein folding, a crucial step for drug discovery.
“It is the novel algorithms that really matter,” David Silver, PhD, the principal investigator on DeepMind’s Go project, told Bloomberg. “It is actually the algorithmic advances that lead to more progress than either compute power or data.”
More articles on artificial intelligence:
Massachusetts General uses machine learning tool to predict breast cancer
53% of CIOs say machine learning is a key focus in their digitization efforts
