AstraZeneca has teamed with artificial intelligence (AI) machine learning firm BenevolentAI to discover new targets for drugs to treat chronic kidney disease and idiopathic pulmonary fibrosis. The news came on 30 April, shortly after STAT reported that IBM would stop selling its machine learning AI tool Watson for Drug Discovery (Nat. Biotechnol. 33, 10–11, 2015), owing to poor financial performance. These opposing fates reflect the controversy that runs in the field over how useful AI-powered efforts may be for target discovery. GlaxoSmithKline was an early entrant into AI when, in 2013, it teamed up with IBM, and by 2017 many collaborations to use AI-based drug discovery followed. Most platforms use natural language systems trained in the life sciences to sift big data sets, which include ‘omics’ information on genes, RNA, proteins and metabolites as well as chemical databases and clinical data, to find new associations and patterns with which to generate new hypotheses. BenevolentAI’s platform in April released results from a study initiated a year earlier with Action Against AMD, which identified seven existing drugs that could be repurposed for acute macular degeneration. In March 2019, Exscientia, a Dundee, UK-based university spinout, announced a collaboration with Celgene to discover small molecule drug candidates in oncology and autoimmunity. Despite pharma’s buy-in, critics contend that AI has yet to accelerate drug discovery as promised, a view underlined by IBM’s limited success in drug discovery. For now, AI is blazing a trail in imaging diagnostics, with AI-based diagnostics for stroke, wrist fractures and diabetic retinopathy approved by the US Food and Drug Administration.