Meet the company using AI to speed up everything from drug discovery to nuclear fusion reactors

London-based artificial intelligence research lab DeepMind—founded in 2010 and acquired by Alphabet in 2014—built its reputation by developing AI that can beat humans across a range of difficult games, including chess, poker, video game StarCraft, and Chinese strategy game Go. And though these accomplishments display AI’s cognitive and reasoning capabilities, providing useful benchmarks for the technology, DeepMind founder Demis Hassabis says he’s happy to see dizzying advances in AI and machine learning find practical applications. “The idea was never to do games as an end in itself, but to develop general-purpose algorithms using games as a convenient test platform, and apply them to important real-world problems,” he says. 

This past July, DeepMind proved it’s no longer just playing games, rolling out 200 million entries to AlphaFold, its open-access database of 3D protein structures, dwarfing the 350,000 protein structures that it launched with in summer 2021. The additional structures—basically, the entire protein universe—are a massive gift to scientists working in drug discovery, synthetic biology, and nanomaterials. Because proteins aren’t static, but in a constant state of folding and refolding themselves before settling into their final 3D structure, determining protein structures has been notoriously difficult and expensive. 

 

Read More : https://www.fastcompany.com/90849923/deepmind-alphafold-protein-database