Google to Accelerate Materials Discovery with DeepMind AI

On Wednesday, Google published a study on how DeepMind artificial intelligence (AI) predicted the structure of over 2 million materials.

The Alphabet-owned AI research company reported that nearly 400,000 of DeepMind’s hypothetical material designs are almost ready for production. After successfully creating the batch of new materials, the firm will focus on utilizing the study for real-world technologies.

For example, the Google DeepMind AI can help develop component substitutes for better-performing car batteries and computer chips. It may also be able to create alternative materials for solar panels that offer the same performance but without the waste.

Material discovery and synthesis are among the high-priced and time-consuming research and development (R&D). For instance, it took researchers about two decades to make lithium-ion batteries that are safe and efficient for commercial use.

DeepMind research scientist Ekin Dogus Cubuk said autonomous experimentation and synthesis can cut the timeline to a few years. He emphasized that their AI leveraged data from approximately 50,000 known materials to create more than 2 million theoretical models.

In addition, Google plans to make DeepMind’s research data publicly available to speed up breakthroughs in material discovery. Materials project director Kristin Persson lauded the move, saying it will encourage more companies to delve into material research.

With the DeepMind AI predicting the stability of new materials, synthesis will become faster and cheaper. Such an outcome could encourage firms to invest in creating ideal materials over working around existing ones.

DeepMind Positions Google as AGI Technology Leader

DeepMind’s breakthrough will help cement Google’s status as a powerhouse in artificial general intelligence (AGI) technology. Hypothetically, a fully realized AGI system can perform any task a human can accomplish.

AGI is of a broader scope than generative AI (GenAI), which changed the content landscape last year with ChatGPT. GenAI learns by being fed information by humans through large language model (LLM) data sets.

In contrast, AGI, also known as general AI, can detect things it does not know and seek answers for itself. Such capabilities make DeepMind superior to ChatGPT in data analysis and providing comprehensive insights.

However, a true AGI, which is a complete substitute for a human, is yet to be seen. Nevertheless, Google investors anticipate DeepMind eventually evolving into the world’s first true AGI.

User Review
0 (0 votes)


Leave a Reply