Applying AI Research to Enhance Google Cloud Solutions
Google Cloud, a branch of Alphabet, provides businesses with a range of digital transformation solutions, including cloud computing, data analytics, and cutting-edge artificial intelligence (AI) and machine learning tools. At the recent Next ’22 developer and tech conference, Google Cloud showcased its latest advancements in these areas.
Over the past few years, we have collaborated with Google Cloud to apply our AI research to improve their core solutions. In this article, we will highlight some of these projects, including optimizing document understanding, enhancing the value of wind energy, and providing easier access to AlphaFold.
Supporting Document AI Innovation
Sharing knowledge through written documents has been essential throughout history, from ancient cuneiform tablets to modern printing presses. However, extracting valuable information from diverse documents poses challenges. Google Cloud’s Document AI tackles this issue by allowing users to upload various documents, such as invoices and tax forms, and make their digital, printed, or handwritten content extractable and searchable.
To make AI tools effective for document understanding, extensive training data is necessary. However, such data is often scarce, incomplete, or lacks proper annotations, hindering widespread AI adoption. Working closely with Google Cloud’s Document AI team, we have developed innovative machine learning models that require significantly less training data, enabling more accurate parsing of utility bills, purchase orders, and potentially other document types. We are also striving to improve the solution’s performance across different languages, making it accessible to customers in various industries and regions.
Enhancing Wind Energy Value
As part of our commitment to advancing science and benefiting humanity, we have focused on leveraging our AI research to support global sustainability initiatives. Together with Google Cloud Professional Services, we have collaborated to positively impact the wind energy sector and promote a carbon-free future.
Although wind farms play a crucial role in generating carbon-free electricity, their output is subject to weather fluctuations. To balance supply and demand on the electricity grid, wind farm operators rely on accurate energy generation forecasts. By developing a custom AI tool trained on weather forecasts and historical wind turbine data, we have helped improve the prediction of wind power output. Additionally, our model recommends delivery commitments for supplying electricity to the grid a day in advance.
ENGIE, a leading energy and renewables supplier, is currently piloting this technology in Germany. If successful, it could be implemented across Europe, making wind energy more economically attractive and reliable, and driving the adoption of renewable energy sources.
Accessing Breakthroughs with Vertex AI
From designing to deploying machine learning models, the development journey involves multiple stages and requires robust data infrastructure. Google Cloud’s Vertex AI simplifies this journey by providing a single platform where users can access machine learning tools for every step of the process.
After introducing our groundbreaking AlphaFold system, which accurately predicts protein structures, we integrated it into Vertex AI. This integration enables scientists to effectively run the AlphaFold prediction workflow by tracking experiments, optimizing hardware selection, and managing operations at scale.
To facilitate access to AlphaFold, we have partnered with Google Cloud to host the AlphaFold Protein Structure Database, which includes nearly all cataloged proteins known to science. This extensive database offers over 200 million proteins for bulk download, with billions of structures already accessed. It has become a valuable resource for the scientific community, driving progress in biology.
If you are interested in applying AI research to make a positive impact in the world, check out our open roles.