“Researchers are developing AI that will accurately predict atomic interactions faster than the compute-heavy simulations of today.”
The sun doesn’t always shine and to store energy[batteries] in its absence is a costly affair. So, with all this talk about going green, how is it possible to incentivise people towards futuristic yet unsustainable and expensive goals?
To address this challenge, Facebook AI and the Carnegie Mellon University (CMU) have announced the Open Catalyst Project, a collaboration intended to use AI to accelerate quantum mechanical simulations by 1,000x in order to discover new electrocatalysts needed for more efficient and scalable ways to store and use renewable energy.
The goal of the Open Catalyst Project, stated the Facebook AI team, is to discover low-cost catalysts in order to find solutions that are good alternatives of current solutions that are inefficient or rely on rare and expensive electrocatalysts like platinum, limiting their practicality.
FB and CMU have even released the Open Catalyst 2020 (OC20) dataset and are also providing baseline models for the community to benchmark approaches against state-of-the-art and compare progress.
Overview Of Open Catalyst Project
To achieve results that are dramatically faster than the compute-heavy simulations scientists rely on today, the team is developing AI that will accurately predict atomic interactions. “Calculations that take modern laboratories days, with the help of AI, could take seconds,” claimed the team at Facebook AI. “This has ramifications outside catalysis, and will enable scientists to rapidly explore and iterate on other challenges that involve quantum mechanics.”
Finding efficient storage techniques is also a material science problem. The properties change with change in chemistry; the way interactions take place at the atomic level is crucial. At the atomic scale, the number of combinations of interactions is difficult to predict. Scientists rely on quantum mechanical simulation tools such as density functional theory(DFT).
DFT is used to simulate the movement of atoms in a given scenario, estimating the energy of a system and attempting to find the configuration with the lowest energy state. This process is computationally intensive, taking hours or even days on high-end servers. DFT also scales poorly with an increase in the number of atoms. Where DFT fails, machine learning thrives. As DFT scales poorly, machine learning models can be trained using the results from initial DFT calculations and approximate the energy and forces of molecules based on past data.
The researchers have identified a use case for machine learning but now comes the real challenge—data. So, the team at FAIR and CMU have even addressed this problem by releasing a dataset called the OC20. The OC20 data set comprises over 1.3 million relaxations of molecular adsorptions onto surfaces, the largest data set of electrocatalyst structures to date.
“Producing this data set also required a substantial amount of engineering expertise and computing power. We ran DFT simulations on spare compute cycles over a period of four months. Facebook’s data centres will reach net zero emissions by the end of the year, making this a responsible and sustainable way to run the compute-intensive calculations necessary to build this data set,” said Facebook.
- Calculations that take modern laboratories days, with the help of AI, could only take a few seconds.
- Success could usher in the widespread adoption of renewable energy, as costs come down and impact on the grid is mitigated by better storage.
- The implications for water quality remediation, medical treatment development, advanced manufacturing, or geochemistry.
- Current baseline models are still far from being useful in practical applications, so there is still much to be accomplished to realise the renewable energy solutions are needed.
Know more about the Open Catalyst Project here.
Subscribe to our NewsletterGet the latest updates and relevant offers by sharing your email.
I have a master's degree in Robotics and I write about machine learning advancements. email:firstname.lastname@example.org