Many professionals might be at least slightly worried what the fast pace of development in robotics and AI could mean for their jobs in future years or decades. But a team of scientists at the University of Liverpool have this week published a study in the journal Nature that sounds very much like a case of turkeys voting for Christmas. They’ve designed an autonomous robot that does the job of a scientist.
The machine, named KUKA-1, is able to take its own decisions on which laboratory experiments to carry out. It then executes them in the lab using the same apparatus and processes that the PhD candidates who would usually do the experiments would. But unlike PhD candidates, KUKA-1 doesn’t get bored painstakingly setting up multiple repetitions of experiments. The robot can also work for 21.5 hours a day, needing just 2.5 hours to recharge its batteries, and can think in 10 dimensions.
KUKA-1 has already proven its credentials by helping to discover a new catalyst, having worked independently for eight days. During that time it carried out 688 experiments on the basis of around 98 million potential combinations of chemical ingredients.
KUKA-1 was designed by a team led by Professor Andrew Cooper, director of the Materials Innovation Factory at the University of Liverpool and co-author of the study in Nature. He was designed to help improve the efficiency of their attempts to develop photo-catalysts, which are materials that help produce hydrogen from water, using sunlight. If the process is ever perfected and made efficient, it would basically provide a limitless, clean source of energy.
He commented of KUKA-1’s contribution to the endeavour:
“You could argue it’s like doing a PhD in a week. We’re sort of seeing the robot as a team member, a kind of very socially maladaptive team member that works all the time and doesn’t join in with any social events.”
KUKA-1 has done such a job assisting Professor Cooper’s team that the university now plans to build two similar robots to help out in laboratories. They cost between £50,000 and £120,000 each to build, depending on their specifications. With post-doc researchers costing around £40,000-£45,000 to employ for a year and generally not wanting to work over 20 hours a day without holidays or sick leave, that represents good value for the university.
Robots like KUKA-1 could also help keep laboratory research flowing when social distancing measures restrict the human workforce doing experiments to around 25% of usual levels. KUKA-1’s movement is informed by 3D laser mapping technology Lidar, which is also use in driverless cars. An articulated arm with a gripper able to pick up glassware or operate machinery helps it conduct experiments. It can also weigh out solid substances, dispense liquids, run catalytic reaction experiments and measure results. Machine learning algorithms help it narrow down which combinations of chemicals might have the best chance of success as catalysts. That selection is then further refined based on the results of experiments already conducted.
Professor Cooper believes the huge number of experiments that could be run with the help of robot assistants like KUKA-1 could potentially revolutionise chemistry. But while PhD candidate lab assistants maybe should be worried, professors such as Professor Cooper don’t have to be just yet. KUKA-1’s failing is that it can’t frame a hypothesis.
As Professor Cooper explains:
“You could think of the robot as a physical search engine. It performs the search autonomously, but the hardest thing about research often is deciding what to work on, and the robot doesn’t do that . . . Currently, you still need the human researcher to frame the problem. So, we’re not making ourselves redundant — yet.”