July 24, 2018
Furthering the evolution of chemical robotics, Researchers at Glasgow University have built an AI robot that discovers new molecules on its own.
Lee Cronin, Regius Professor at Glasgow says, “It’s a bit like a robotic bartender that mixes cocktails. In our lab, the robot mixes together the chemicals in just the right way – shaken, not stirred – and explores what happens.”
Cronin further explains that it is not just about the mixing of chemicals, the role of AI and machine learning enables the system to detect the right reaction, which can lead it to the creation of new molecules.
To enable the robot to do so, the researchers have trained the system with about 10% of all the tasks involved in molecule research. The training simply involved shadowing of the human chemists in the lab, and tagging of all the discoveries they have made. The machine learning algorithm will then allow the system to learn the pattern of trial and error and deeper research in promising reactions.
After training, when the robot is set to work independently, it begins by creating a new database, as explained by the professor in his interview with the Nature publications. The new database will be mostly empty except the name of the ingredients that it will randomly add and mix to find the probability of creating new molecules, and it does it all in the real-time.
It is notable how this system will open up new dimensions in chemical robotics and AI. Till now, the Artificially Intelligent systems used in Chemist labs were performing predefined steps fed by human programmers and chemists. Now, when it is proven how they can actively contribute in the areas that hitherto been considered limited to the explorative intelligence of humans, the role of AI in research can become more significant, optimizing R&D costs through more automated processes.
This would mean revolutionary changes in the entire sphere; as Cronin says, “There are more molecules possible to be made than there are atoms in the Universe. There are so many molecules we need to search for and make, there’s no way we could even scratch the surface without some help.”
Automation will certainly make these researches safer and cheaper, leading to better drugs research and possibly more economic medications. The optimization will further get help from Prof. Cronin’s other chemistry initiatives like Deep Matter Group which works on the digitizing the chemical world.
At present, the robot’s predictions have had 80% accuracy. With more training, it will improve, so will the pace of research with reduced expansion costs. Cronin says it will free many chemists from performing the recurrent manual tasks in the labs, and instead allow them to focus on doing actual chemistry.
This is not the first time Machine learning has made the job of chemists and chemical detection easier. In the first quarter of the year, WADA or World Anti-Doping Agency has announced research and development in AI for the improvement in their anti-doping efforts. To get more information on this, read our full report here.
Resources:
https://www.nature.com/articles/d41586-018-05776-9
https://www.nature.com/articles/s41586-018-0307-8
https://www.bbc.com/news/uk-scotland-44872432
