Autonomous Robot Scientist Does in Days What It Would Take a Human Months to Complete

A free-roaming robot that works autonomously in normal lab conditions has been unveiled today—should human scientists be worried?

The 900 pound, 5 foot 7 tall machine can conduct experiments in darkness, has an infinite amount of patience and can work for more than 21 hours a day without needing to recharge. It does not care about timezones, or in fact, the concept of time at all.

Powered by algorithms, it wirelessly zips around a lab with ease, using an articulated arm to conduct tests using the same instruments as human researchers. Luckily for our species, however, the academic team behind the bot had altruism in mind.

"It's not meant to replace researchers, in the sense that it is doing different things," Professor Andrew Cooper, author of a new paper detailing the capabilities of the robot, told Newsweek during a telephone interview. "Also, it's important to stress that you still need the human scientists because the robot does the experiments but it doesn't decide what experiments to do.

"One of the hardest things about research is choosing what to work on, and the robot doesn't do that. So if you ask the robot to work on a pointless of uninteresting problem it would just do that forever. So we are not replacing the scientists, the aim is to have the ability for scientists to do much more ambitious and interesting chemistry. You still need the scientist. It's an additional team member, but you still need the team."

While that's clearly a win for humanity, the project still provides an tantalizing glimpse into how human researchers will work alongside robots in the future.

At its core, the robot was made to aid research into the development of photocatalysts, which are materials that help produce hydrogen from water using light.

The scientists said that unlike many automated systems, their robot is able to dispense solids and liquids with both "high accuracy and repeatability." After being programed, it can pick items up, put items down and even operate lab machinery.

It uses a combination of laser scanning and touch feedback rather than a vision system, meaning can operate in total darkness, perfect for light-sensitive reactions.

Crucially, the tech works. In an example, the robot was set up to explore hypotheses to potentially improve the performance of a polymeric photocatalyst. The machine ran 688 experiments in eight days—diligently working for 172 out of 192 hours.

Ultimately, it optimized the reaction conditions in three days. Experts said the process may take a human researcher "several months" to complete a similar task.

The findings, from experts at the University of Liverpool, U.K., are outlined in a paper published today in the journal Nature. According to the team, the robot could be used to hunt for materials for clean energy or new drug formulations by searching for unexplored chemical reactions.

Cooper told Newsweek the robot took about three years to construct, and cost roughly £100,000 ($125,000), a price that is not unusual compared to high-end lab instruments. "It would be expensive if it was going to mow your lawn but in the context of research we think it's actually considerably less expensive than other approaches," he said.

After working through the technical challenges of designing and producing the robot, the professor said the process of experimenting has been streamlined. "If you want to set up a new instrument into your experiment, you train the robot where that instrument is, that takes maybe a couple of hours, once it knows where it is and you have the software connection, then it can use that instrument," he said.

Looking to the future, the team will potentially add more AI functionality, including letting the robot learn from its own mistakes, getting smarter over time.

The biggest breakthrough, for now, is having a robot that integrates the use of multiple instruments rather than a machine hardwired for single tasks, Cooper said.

After being set up by a human, the university's mechanical creation can add solids or liquids, shake materials, shine light on samples to help encourage a chemical reaction and even transport the samples to another machine in the laboratory.

In some ways, the robot looks would fit in at a car facility over an academic institution— but the team has embraced the helper. "It really is autonomous," Cooper told Newsweek. "You just need a technician to load up the materials and off it goes."

Mobile robot chemist
This image shows a photograph of the mobile robot chemist loading a rack of experimental samples. Prof. Andrew Cooper

University of Liverpool PhD student Benjamin Burger, who built and programed the bot while working on the project, said in a statement: "The biggest challenge was to make the system robust. To work autonomously over multiple days, making thousands of delicate manipulations, the failure rate for each task needs to be very low. But once this is done, the robot makes far fewer mistakes than a human operator."

While still largely theoretical, Cooper told Newsweek he believes increased collaboration between robots to be the direction the research will take. In theory, multiple robots could run collaborative experiments from across the world at the same time.

In addition, it is very easy to see how scientists may come to increasingly rely on robots to maintain social distancing in labs throughout the COVID-19 pandemic.

And while it may be a scary prospect for some, we could only be a decade away from having laboratory-based robots that truly think for themselves, Cooper said. "It's hard to predict the timescale for this but it's not impossible to imagine a system where the computer brain actually does decide what experiments to do, so then not only is the robot doing the chemistry, it is choosing what to do," he said. "And I think that's a very exciting idea, because then not only don't you know what the results will be, you don't even know what the thing is going to do.

"You can imagine a scenario where the system is asking for certain chemicals and you don't know what its intent is. We are some way from that at the moment. But I can imagine that could happen in the next 10 years.

"Certainly, international collaboration we could potentially do now, but the bigger idea is autonomous science where the system chooses what to do and why. That's going to take a while but I will stick my neck out and say maybe in the next 10 years."

For now, the project expects to have two more of the robots in place by the end of 2020—possibly each being given specialized roles. The team will develop the AI to better communicate with human counterparts, maybe even via voice commands.

Reflecting on the project, Cooper noted the humanoid nature of the robot had surfaced interesting behaviors about the nature of a machine-human workplace.

"It's almost impossible, when you see the thing moving around, it's almost impossible not to anthropomorphize," he said. "I noticed people had started to call it him or her. It was kind of interesting because you don't normally call a lab instrument him or her. I think it's because of the way it moves... it looks sort of strangely human-like."

Cooper said that, ultimately, he had viewed the robot as being a specialized team member. "I referred to 'it' as 'it.' I'm not sure what that says about me," he joked.