AI Robots Will Learn to Conduct Science Experiments on Their Own


Brian Storey is a scientist with the Toyota Research Institute, leading a project with AI machines running experiments. But the robotic arm doesn’t just pick bottles, adds a precise number of drops in test tubes and carries the tests to find the perfect chemical makeup for car batteries with a high-capacity. Brian is looking for a way to make the AI take over the experiments and not only perform the tests but also do the whole planning. He said the following:

“It’s automating not only the manual part of doing the experiment but also the planning part.”

With so much raw information and huge amounts of data, the head of the Carnegie Mellon’s machine learning department agreed that ‘we just cannot handle the amount of data anymore’, so using robots that can ‘eat’ terabytes of information and process is the way to go.

Science is Still Conducted By Humans, Not Run By Robots

So far, everyone believed that science can only be conducted and thought by humans, but robots are now helping us sift through tons of information, making scientific discoveries easier. For example, the discovery of the Higgs boson, knowns as the God particle, the AI used an algorithm that helped to search through billions of particles tracks that have been generated in the Large Hadron Collider (Switzerland).

One thing is clear, robots are not taking over science. They’re making researchers’ work easier. A collaborator on the Toyota project, Barnabás Póczos, who is a Carnegie Mellon machine learning professor stated that:

“I can easily imagine cases in which AI would recommend experiments to try to synthesize a chemical molecule that you wouldn’t think possible, but the AI will be able to do it.”

However, AI cannot generate new predictions, so the only way to discover new things is for the humans to work together with the AI.

In February, a trial collaboration with software maker Euretos was announced by the Dutch publisher Elsevier. They will use AI to read and assess millions of scientific articles that have been peer-reviewed. Then, the AI will suggest hypotheses in biochemistry field of work. From these hypotheses, academics will start choosing the ones that are promising.

“The vision is that the discussion becomes a much more automated process,” says Arie Baak, Euretos co-founder.


Recommended For You

Leave a Reply

Your email address will not be published. Required fields are marked *