Cooking with AI
At the University of Toronto, Ted Sargent runs a test kitchen of types. Him and his colleagues develop recipes, measure and mix ingredients carefully, and then assess its outcome. The combination of ingredients mostly – if not always – turn out to be inedible.
Fortunately, they are not making food. Sargent’s team cooks with carbon dioxide. Their goal is to invent recipes to ‘upgrade’ greenhouse gases into more useful raw materials. Instead of releasing the pollutant into the air, or capturing it and sequestering it underground, factories and power plants of the future could use renewable energy sources to convert the carbon dioxide into materials that can be sold.
One promising recipe involves electrically zapping carbon dioxide with other reactants to change it into the six-atom molecule ethylene, which is composed of two carbon atoms and two hydrogen atoms. It’s a raw material involved in the making of common plastics such as Ziploc bags.
It’s about a $60 billion market. It’s a pretty valuable commodity chemical.
The real significance of Sargent’s work is that the process of cooking is used with artificial intelligence. Sargent’s colleagues came across new recipes for producing ethylene by using new AI and supercomputer-driven techniques.
Sargent collaborated with Zachary Ulissi of Carnegie Mellon University, who specialises in using algorithms to create new materials. Ulissi stimulated 12,229 microscopic close-ups of 244 different crystals to zero in on the most promising contenders in the making of ethylene. They wanted to find materials that would easily stick to carbon monoxide molecules. Ulissi used a supercomputer to run a small proportion of the stimulations, then trained a machine-learning algorithm with those supercomputing results, and the algorithm learned to make the remaining stimulations efficiently.

Electrical set ups like these are used to chemically upgrade carbon dioxide into new useful raw materials in the University of Toronto. Image Credit: Wired
These computer-based methods provide researchers with a swift and flexible strategy for discovering new materials. Both teams have not found their winning ingredients yet because of the large supply of electricity needed – this means that producing ethylene from carbon dioxide would not be profitable. The team are working to design more economically viable recipes. Nature published an article last week reporting the discovery of multiple new materials – electrocatalysts. They allow for faster and more energy-efficient recipes for the conversion of ethylene from carbon dioxide. These catalysts could be the secret ingredient which enables the technology to become more scalable.
We need to decrease our carbon footprint, but we don’t want to do so at the expense of the increasing prosperity of people around the world.
To find their catalysts, both Ulissi and Sargent’s team used a public database called the Materials Project, which contains data about more than 12,000 inorganic chemical compounds. Both teams knew from previous experience that materials containing copper made good catalysis, so they searched the site specifically for a non reactive alloy made of copper. From the provided list of 244 crystals, an aluminium-copper alloy seemed to be the best fit. After the algorithm made predictions for how evenly the two metals should be mixed and its best ratios, scientists synthesised these metals in the lab based on those predictions and fed the results back to the algorithm. After some time this process led them to uncover 17 efficient catalysts. (All of this work was completed before the lab shut down two months ago due to the Covid-19 pandemic).

Supercomputing techniques and AI is used to create this new copper-aluminium alloy for speeding up the conversion of ethylene from carbon dioxide, a chemical used to make plastic. Image Credit: Wired
More scientists are now relying on computing tools to help create new raw materials. Even with AI and supercomputers, it took Ulissi and Sargent’s team approximately three years to identify these new catalysts, test them, and publish the results.
It is said that incorporating robots to the workflow should increase material discovery. Still, with robots, material discovery will need human oversight. Even with stand mixers, food processors, and Instapots, the kitchen still needs a cook.