Deep Blue’ holds a special place in many people’s hearts. In 1996, Garry Kasparov, then world chess champion, squared off against an opponent that went by the name ‘Deep Blue’, a unique opponent — a computer powered with the ability to evaluate 200 million positions per second. Kasparov won 4-2 and they faced again a year later.

This time it was a different outcome, ‘Deep Blue’ won 3.5-2.5 and signalled the age of machines becoming so much superior to humans in chess. AlphaZero, developed by DeepMind, through self-play, learnt to beat the best computer chess engines.

In Garry Kasparov’s words, the difference between the best human chess player and AlphaZero is similar to the difference between Usain Bolt (top speed: ~25 mph) and a Ferrari (top speed: 200 mph).

Analogously, humans have competed in a race to discover new materials. Modern life is built on our ability to invent new materials that offer unprecedented new capability:

Inventing new materials

Silicon for transistors that enabled the computing era, Li-ions as the storage mechanism in batteries that has powered the portable electronics revolution, Silicon, again, for solar cells enabling the renewable power industry, to name a few.

Those that have invented new materials have typically gone on to win Nobel Prizes, the ultimate recognition in science.

The role of robots

Now, a new kind of competitor has emerged: Robots that perform experiments, along with machine learning, which can analyse the outcome of the experiment and decide what to do next, acting as the ‘brains’ of the scientist. These setups can perform quite sophisticated experiments, typically reserved for scientists with advanced training.

Such robotic apparatus have now been built for photovoltaics, organic synthesis, and photocatalysis; my team has built two such robots for batteries.

These robotic setups have now found materials with better performance in shorter time with little to no human intervention. I think we have reached the ‘Deep Blue’ moment for materials discovery. This reference to ‘Deep Blue’ is intentional as the current hardware requires hand-engineering the design space akin to the hand-coded rules used by Deep Blue.

Humans’ role

This begs the question, how should humans participate in materials innovation in the future? A year after his loss to Deep Blue, Garry Kasparov invented a new form of chess called ‘Centaur Chess’.

Centaur is a mythical creature that is half-human, half-horse with the human providing direction and intellect and the horse providing the ability to proceed fast.

The notion of ‘Centaur Chess’ is to combine human intuition along with the computer’s ability for brute computation. Eventually, the hope is that human-plus-machine is greater than just human or machine alone.

Centaur materials

In the same way, ‘Centaur Materials’ appear to be the best path forward. The aim of ‘Centaur Materials’ is to combine human intuition in identifying the problem to solve and select the region of materials space where a discovery might be made. Then, turn to machine to do brute force computation, in this case, evaluate material performance through robotic experimentation, followed by navigating the highly complex decision landscape and decide what to do next. Taking to the robotic approach rather than fighting it will help materials innovation scale.

Looking ahead

The next decade will see new material design spaces fully optimised autonomously by robotic experimentation that will yield breakthrough materials well-beyond human intuition, knowledge or reasoning.

Only time will tell whether one of these robotic setups will go on to win the Nobel Prize, but I would not bet against it.

The writer is an Associate Professor of Mechanical Engineering at Carnegie Mellon

Related Topics
comment COMMENT NOW