Electric vehicles can significantly reduce CO2 emissions, but car factories are running out of materials to make batteries. One of the key components, nickel, is projected to cause a supply shortfall by the end of the year. Scientists have recently discovered four new substances that could potentially help. But what is perhaps even more interesting is how these substances were found: the researchers relied on artificial intelligence to select useful chemicals from a list of more than 300 options. And they are not the only ones who turn to AI for scientific inspiration.
Setting up hypotheses is a purely human field from the beginning. Now, however, scientists are beginning to ask machine learning to create original insights. Neural networks are designed that propose new hypotheses based on the patterns found by the networks in the data, rather than relying on human assumptions. Soon, many areas may turn to the muse of machine learning to speed up the scientific process and reduce human biases. co-author of a study on the search for batteries. He adds that the AI procedure helps to identify the chemical combinations that are worth examining, so that many more chemical areas can be covered more quickly.
The discovery of new substances is not the only area where machine learning can contribute. to science. Researchers are also applying neural networks to larger technical and theoretical issues. Renato Renner, a physicist at the Institute for Theoretical Physics in Zurich, hopes to one day use machine learning to develop a unified theory of how the universe works. However, before AI can reveal the true nature of reality, researchers need to address the notoriously difficult question of how neural networks built like the human nervous system make their decisions.
For new battery materials, Scientists performing such tasks have so far typically relied on database search tools, modeling, and their own intuitions about chemicals to select useful compounds. Instead, a team at the University of Liverpool used machine learning to simplify the creative process. The researchers developed a neural network that ranked chemical combinations according to their likelihood of producing a useful new substance. Scientists then used these rankings to guide their laboratory experiments. Four promising battery candidates have been identified without having to test all the substances on the list, saving months of experimentation and error
Hardware, software, tests, curiosities and colorful news from the IT world here clicked