Dropbox, Inc.

Openai Opco, LLC

Jie Tang

Brian E Smith

Rian Hunter

Aston Motes

David Euresti

Alicia Chen

Alex Sydell

Ilge Akkaya

Raul Gonzalez

Peter Zhokhov

Bowen Baker

Jeffrey Clune

Brandon Houghton

Adrien Ecoffet

Ramesh Balakrishnan

Francois Alexander Allain

Julie Tung

Josiah Boning

Joost Huizanga

Joost Huizinga

Disclosed herein are methods, systems, and computer-readable media for training a machine learning model to label unlabeled data and/or perform automated actions. In an embodiment, a method comprises receiving unlabeled digital video data, generating pseudo-labels for the unlabeled digital video data, the generating comprising receiving labeled digital video data, training an inverse dynamics model (IDM) using the labeled digital video data, and generating at least one pseudo-label for the unlabeled digital video data, wherein the at least one pseudo-label is based on a prediction, generated by the IDM, of one or more actions that mimic at least one timestep of the unlabeled digital video data. In some embodiments, the method further comprises adding the at least one pseudo-label to the unlabeled digital video data and further training the IDM or a machine learning model using the pseudo-labeled digital video data.

Using machine learning to train and use a model to perform automatic interface actions based on video and input datasets

The disclosed embodiments relate to a system that facilitates accessing audio files associated with a user from an online content management system. During operation, the online content management system receives a function call from a client to list audio files associated with the user. In response to the function call, the system filters through operations for files associated with the user to identify audio files. Next, the system returns metadata for the identified audio files associated with the user to the client. In some embodiments, filtering through the operations involves filtering an operation log to identify operations performed on audio files associated with the user, wherein the operations occurred since a preceding call to the function.

Accessing audio files from an online content management system

Disclosed are systems, methods, and non-transitory computer-readable storage media for prioritizing structural operations and distributing changes in a synced online content management system. An exemplary method can include receiving, from a first client device, a first structural operation for a particular content item and an account within the online content management system. The method can also include receiving, from a second client device, a competing structural operation for the particular content item and the account. The method can then include prioritizing the first structural operation from the first client device over the competing structural operation from the second client device based on predetermined factors. The method can then include executing the first structural operation for the particular content item and the account. The exemplary method can then include distributing the prioritized first structural operation to all synced devices for execution.

Prioritizing structural operations and distributing changes in a synced online content management system

A method, system, and manufacture for authorizing an untrusted client device for access on a content management system. The content management system receives a request from an untrusted client device to access content on the content management system. The content management system sends an authentication key to the untrusted client device. The content management system then receives the authentication key from a trusted client device. Based on the matching authentication key, the content management system transmits data to the untrusted client device in accordance with any additional instructions that the trusted client device may have sent.

Authorizing an untrusted client device for access on a content management system

A method and system for automatically rotating image files and preventing duplicate uploads in a remote file storage system is described. The system receives an image file. The system calculates a hash value associated with newly uploaded files. The system then automatically rotates the images based on orientation data associated with the files (if included). The system then compares the hash value(s) associated with the new file(s) with the hash values from the existing files stored on the remote file storage system. If the hash values of any of the new files to be uploaded match a hash value on the hash value list, then the system prevents the new files from being uploaded to the remote file storage system.

System and method for preventing duplicate uploads of modified photos in a synchronized content management system

A method and system for preventing duplicate file uploads in a remote content management system is described. The user device receives a hash value list associated with the files stored in the remote content management system. The user device calculates a hash value associated with new files to be uploaded. The system then compares the hash value(s) associated with the new file(s) to be uploaded with the hash value list received from the remote file storage system. If the hash values of any of the new files to be uploaded match a hash value on the hash value list, then the system prevents the new files from being uploaded to the remote file storage system.

System and method for preventing duplicate file uploads in a synchronized content management system

Growing community of inventors

San Francisco, CA, United States of America

Jie Tang