Google LLC

How AI Models Understand Language Using 'Attention'

This patent describes a neural network architecture, known as a Transformer, that uses a "self-attention" mechanism to process sequences of information, like words in a sentence, by weighing the importance of different parts of the input.

How Google's Patent Scores Authors and Posts on Messaging Systems

Google's 2013 patent describes a system for ranking authors and their posts on a messaging platform based on user interactions and subscriptions, influencing content visibility.

How Digital Maps Are Built From Small Image Pieces

Google's 2007 patent on how to assemble small map images, called tiles, into a larger map view on your device, enabling smooth zooming and panning.

Google's System for Auto-Searching Text on Your Device

Google's 2015 patent describes a system that automatically finds and displays relevant information online based on text it detects on your device, even without you asking it to.

How Touchscreen Gestures Use Physics to Feel Real

Google's patent on making touchscreen gestures feel natural by applying simulated physics, like friction or magnetism, to items on your screen.

How Google Displays Knowledge Panels Next to Search Results

A method for automatically assembling and displaying information boxes on search result pages by pulling data from multiple sources based on user query patterns.

How Google Ads Detect Which Web Page You Are Viewing

A method for web browsers to identify the correct webpage address to show relevant advertisements, even when the ad code is hidden inside a frame.

How Google Automatically Groups Photos and Contextual Data into Moments

A method for automatically grouping photos from front and rear cameras with related data like audio, weather, and location to create a rich, interactive digital memory.

How Google Distributed Machine Learning Across Many Computers

A 2003 Google patent describing a way to build machine learning models by splitting the work across a large network of computers rather than a single machine.

How Projection Neural Networks Speed Up AI Predictions

A method for making artificial intelligence models faster and more efficient by using fixed, non-trainable projections to simplify complex data before processing.

How to Shrink Large AI Models Using Knowledge Distillation

A method for teaching small, efficient AI models to mimic the complex decision-making patterns of much larger, more powerful neural networks.

How Smart Home Devices Automatically Connect to Utility Energy Programs

A system that lets smart home devices like thermostats enroll in energy-saving programs without the user needing to manually provide their utility account number.

How Search Engines Personalize Results Based on Your Browsing History

Google's patent for narrowing down search results by prioritizing websites that match your personal interests and past browsing habits.

How Social Networks Automatically Create Groups Based on Time and Place

Google's patent describes a system that automatically groups social media users together based on where they are or when they are posting, allowing for instant, context-aware content sharing.

How a Centralized Broker Handles Online Shopping Transactions

A system where a third-party broker handles payments and shipping logistics for online merchants, allowing customers to checkout without entering payment details on every individual site.

Predicting User Interests Based on Who Looks at Whose Profile

A method for predicting what a user might be interested in by analyzing the web of connections created when people view each other's social media profile pages.

How Google Uses User Feedback to Rank Websites by Quality

A method for Google to judge a website's quality by looking at how users interact with its search results across different categories of search queries.