We hear more and more talk about artificial intelligence (AI), and sometimes with concern, for the fear that this type of technology could undermine the human capacity for achievement or even take it over. But a study published by a team of researchers from the University of Münster in Germany, in Nature magazine explains better than a lot of talk about what AI can do for us; namely, carry out often boring and repetitive tasks which require a considerable expense of time and energy for humans. In this case, the context were the chemical pathways to produce new molecules for industry use in the quickest and most advantageous way.
The German researchers primed their computers on almost all the known reactions, about 12.4 million, and then asked these devices to suggest the best solutions, or rather the most efficient chemistry reactions, to synthesize 9 molecules using a system of artificial intelligence developed by the chemist Marwin Segler and other computer scientists from the university. At the end of the process, the researchers carried out a blind test by providing the AI-suggested solutions to 45 “judges” made up of German and Chinese chemists, along with solutions created by human experts from the university, without specifying the origins.
The evaluators did not notice a difference between the suggested proposals, demonstrating that those originating from AI were as valid as the others. Since the 60s, industries have been trying to follow this path, but only recently these applications have proved to be efficient. According to experts, the software designed by Segler is particularly advanced, because it can “learn” data on its own and has no need of human input to implement it. Many firms are buying programs like these; for example, Merck has bought a Korean program called Chematica for an undisclosed sum, and other big names from the industry are doing the same, to improve the research phases of new medicines.