Nvidia Corporation

Sony Corporation

Sony Pictures Entertainment, Inc.

Other

Jonathan Cohen

David Patrick Luebke

Peter Schuyler Shirley

Patrice Castonguay

Chris Allen

Samuli Matias Laine

Timo Oskari Aila

David Kirk McAllister

Morgan McGuire

Rev Lebaredian

Ratin Kumar

Michael Garland

Santanu Dutta

Eric Enderton

Ryan Leary

William N Bell

Simon Yuen

Michael Clair Houston

Frank Jargstorff

Michael Robert Phillips

Sarah Tariq

Eric Young

Roy Kim

Approaches presented herein can provide for the performance of specific types of tasks using a large model, without a need to retrain the model. Custom endpoints can be trained for specific types of tasks, as may be indicated by the specification of one or more guidance mechanisms. A guidance mechanism can be added to or used along with a request to guide the model in performing a type of task with respect to a string of text. An endpoint receiving such a request can perform any marshalling needed to get the request in a format required by the model, and can add the guidance mechanisms to the request by, for example, prepending one or more text strings (or text prefixes) to a text-formatted request. A model receiving this string can process the text according to the guidance mechanisms. Such an approach can allow for a variety of tasks to be performed by a single model.

Natural language processing applications using large language models

In various examples, a virtually animated and interactive agent may be rendered for visual and audible communication with one or more users with an application. For example, a conversational artificial intelligence (AI) assistant may be rendered and displayed for visual communication in addition to audible communication with end-users. As such, the AI assistant may leverage the visual domain—in addition to the audible domain—to more clearly communicate with users, including interacting with a virtual environment in which the AI assistant is rendered. Similarly, the AI assistant may leverage audio, video, and/or text inputs from a user to determine a request, mood, gesture, and/or posture of a user for more accurately responding to and interacting with the user.

Conversational AI platform with rendered graphical output

A special-purpose processing system, a method of carrying out sharing special-purpose processing resources and a graphics processing system. In one embodiment, the special-purpose processing system includes: (1) a special-purpose processing resource and (2) a Representational State Transfer (ReST) application programming interface operable to process data using the special-purpose processing resource in response to stateless commands based on a standard protocol selected from the group consisting of: (2a) a standard network protocol and (2b) a standard database query protocol.

System and method for sharing special-purpose processing resources

A system, method, and computer program product are provided for performing graph aggregation. In use, a graph with a plurality of vertices and a plurality of edges is identified. Additionally, aggregation is performed on the vertices and edges of the graph by computing a graph matching, where such graph matching is performed in a data-parallel manner.

System, method, and computer program product for performing graph aggregation

A system, method, and computer program product are provided for categorizing a plurality of vertices of a graph into independent sets. A random number is assigned to each vertex in the graph and the assigned number of each vertex is compared to the assigned numbers each of the neighbors of the vertex, where all vertices in the graph that have an assigned number greater than the assigned numbers of each of their neighbors are added to a first independent set, and all vertices in the graph that have an assigned number less than the assigned numbers of each of their neighbors are added to a second independent set separate from the first independent set.

System, method, and computer program product for performing graph coloring

A system, method, and computer program product are provided for categorizing a plurality of vertices of a graph. A predetermined plurality of random numbers is assigned to each vertex of the plurality of vertices, a determination is made whether each of the assigned predetermined plurality of random numbers of a single vertex is greater than a corresponding random number of the assigned predetermined plurality of random numbers of each of the neighbors of the single vertex, and in response to the determination, one of the assigned random numbers is selected from a group of assigned random numbers of the single vertex.

A system, method, and computer program product are provided for performing graph aggregation. In use, a graph with a plurality of vertices and a plurality of edges is identified. Additionally, graph matching is performed on the vertices and edges of the graph by computing a graph matching, wherein the performance of the graph matching is optimized.

System, method, and computer program product for performing graph matching

A system, method, and computer program product are provided for performing shadowing utilizing shadow maps and ray tracing. In operation, one or more shadow maps are rendered for at least one light source. Additionally, low confidence pixels associated with the one or more shadow maps are determined. Furthermore, shadow rays associated with the low confidence pixels are traced.

System, method, and computer program product for performing shadowing utilizing shadow maps and ray tracing

A system, method, and computer program product are provided for reducing noise in an image using depth-based on sweeping over image samples. In use, each noisy pixel of an image having noise is identified. Additionally, for each noisy pixel, at least one sample included in each of a plurality of neighboring pixels to the noisy pixel is identified. Furthermore, the samples are swept over at least partially in a depth-based order to identify a value for the noisy pixel that reduces the noise.

System, method, and computer program product for reducing noise in an image using depth-based sweeping over image samples

A method of simulation comprises controlling an avatar in an environment. Movement of graphical elements is simulated in a fluid coordinate frame surrounding said avatar, wherein said graphical elements in said fluid coordinate frame obey a first rule set. Said graphical elements and a first region surrounding said fluid coordinate frame are animated, wherein said graphical elements in said first region obey a second rule set. Said fluid coordinate frame moves in response to said controlling of said avatar. In an embodiment, a blending region blends the movement of graphical elements inside the fluid coordinate frame and outside the fluid coordinate frame.

Region of interest tracking for fluid simulation

A method for detecting object collisions in a simulation, which includes identifying a plurality of objects moving along a path within a simulation area, and defining a grid comprising defined regions which individually define a region within which any of the plurality of objects could potentially occupy. For each of the objects, the method further includes identifying which of the defined regions that each of the plurality of object occupies for at least a portion of a time step, and for each of the objects, determining an associated potential collision set by identifying objects of the plurality of objects which occupy common regions of the defined regions during any portion of the time step. In addition, for each of the objects, the method further includes determining an actual collision set comprising objects with which a given object will collide during the time step based upon location parameters of objects included in the potential collision set.

Method for detecting collisions among large numbers of particles

A method for use in computer graphics includes receiving data defining an explicit surface representation of three-dimensional (3D) geometry and casting a ray into the explicit surface representation. For each point where the ray intersects the explicit surface, a direction of the ray is compared to a direction of a surface normal of the explicit surface at that point, and then a count is incremented or decremented based on a first rule that uses a result of the comparison at that point. Then, for one or more voxels corresponding to the ray, data is generated for each voxel that indicates the voxel is either inside or outside of the explicit surface based on a second rule that uses a value of the count at that voxel. Another method includes defining a narrow band of voxels near the explicit surface, and then for each voxel in the narrow band, approximating a distance between the voxel and a closest point on the explicit surface by assuming a set of two-dimensional (2D) distances associated with the voxel and a portion of the explicit surface form a triangle. For each voxel outside of the narrow band, a distance to a closest point on the explicit surface is approximated by interpolating the approximated distances in the narrow band.

Method for converting polygonal surfaces to levelsets

A method for use in computer graphics includes receiving data representative of a plurality of particles. For each particle, the method further includes calculating a time at which a collision will occur based on a current trajectory of the particle. For pairs of particles that will collide with each other, another operation includes changing a direction of the particles to avoid a collision, wherein the changed directions approximate directions that would result from the collision. For particles in static contact, another operation includes allowing particles to push each other without bouncing. For particles in static contact that were previously moved by being pushed, the method also does not allow the particles to move during future collisions occurring during a time step.

Method for simulating large numbers of spherical bodies interacting

A method for use in rendering includes receiving an input particle system, an instancing program, and a number indicating a maximum number of particles to be stored in memory, providing an input particle count representative of at least a portion of the input particle system to at least one operator for the instancing program, running the at least one operator in a prediction mode to generate an output particle count, comparing the output particle count to the number indicating a maximum number of particles to be stored in memory, and spatially splitting a bounding box representative of the input particle count in response to the output particle count being greater than the number indicating a maximum number of particles to be stored in memory.

Method for predictively splitting procedurally generated particle data into screen-space boxes

A data visualization system and method are provided herein.

Growing community of inventors

Ann Arbor, MI, United States of America

Jonathan Cohen