Granted Patent B2

US 12,455,918 · App. 18/652,223 · Granted Oct 28, 2025

Inventors: Tamay Aykut (Pacifica, CA); Christopher Benjamin Kuhn (Munich, DE)

Assignee: Sureel Inc.

G06F16/45G06F16/438G06N3/0455G06N3/0475G06Q30/0208

Abstract

Claims (115)

1 . A method comprising:

determining, by one or more processors, that a training of an artificial intelligence has been completed, to create a trained artificial intelligence;

determining, by the one or more processors, an attribution vector created during the training of the artificial intelligence;

determining, by the one or more processors, that the trained artificial intelligence has received an input;

generating, by the trained artificial intelligence, an output based on the input;

while generating the output based on the input, determining:

activations of neural pathways in the trained artificial intelligence; and

an amount of individual activations;

after determining that the trained artificial intelligence has completed generating the output, creating a dynamic attribution vector based on the activations of the neural pathways and the amount of the individual activations, the dynamic attribution vector indicating an influence of individual content creators on the output;

modifying the attribution vector based at least in part on the dynamic attribution vector;

determining, by the one or more processors, an attribution determination for individual content creators of multiple content creators based on the attribution vector and based on determining one or more of the multiple content creators that contributed at least a threshold amount during the training, wherein a particular creator having a contribution less than the threshold amount does not receive a creator attribution; and

initiating, by the one or more processors, providing compensation to one or more of the multiple content creators based at least in part on the attribution determination;

wherein training the artificial intelligence to create the trained artificial intelligence comprises:

selecting a content creator from multiple content creators to create a selected content creator;

selecting a plurality of content items associated with the selected content creator to create selected content items;

training the artificial intelligence using the selected content items;

determining a measure of a proximity between:

the artificial intelligence trained using the selected content items, and the artificial intelligence prior to initiating the training;

determining a creator influence of the content creator on the trained artificial intelligence based at least in part on the measure of the proximity; and

including the creator influence in the attribution vector.

2 . The method of claim 1 , further comprising:

determining a transformativeness factor of the trained artificial intelligence relative to the artificial intelligence based at least in part on the measure of the proximity; and

determining the attribution vector based at least in part on the transformativeness factor.

3 . The method of claim 1 , further comprising:

determining a change caused by individual content items of the selected content items associated with the selected content creator to weights of the artificial intelligence during the training;

determining a number of activations caused by individual content items of the selected content items; and

determining a contribution of the selected content creator based at least in part on multiplying the change to the weights by the number of activations.

4 . The method of claim 1 , further comprising:

determining a contribution of individual content items of the selected content items associated with the selected content creator on gradients of the artificial intelligence during the training;

aggregating contributions of the selected content items; and

determining a contribution of the selected content creator based at least in part on the aggregating.

5 . The method of claim 1 , wherein the training further comprises:

determining a content item influence of individual content items of the plurality of content items associated with the content creator using a converging function comprising a loss minimization algorithm that uses a direction and a learning rate to reduce a cost function close to zero.

6 . The method of claim 1 , wherein the training further comprises:

determining a number of activations caused to individual parameters of the artificial intelligence;

determining a relative contribution of the selected content items of individual content creators based at least in part on a change in parameters of the trained artificial intelligence caused by the multiple content items with their respective activations;

after completing the training to create the trained artificial intelligence, generating a static attribution vector based on the relative contribution of the selected content items of individual content creators; and

creating the attribution vector based at least in part on the static attribution vector.

7 . The method of claim 1 , further comprising:

storing the attribution vector in a blockchain.

8 . A server comprising:

one or more processors; and

a non-transitory memory device to store instructions executable by the one or more processors to perform operations comprising:

determining that a training of an artificial intelligence has been completed, creating a trained artificial intelligence;

determining an attribution vector created during the training of the artificial intelligence;

determining that the trained artificial intelligence has received an input;

generating, by the trained artificial intelligence, an output based on the input;

determining an attribution determination for individual content creators of multiple content creators based on the attribution vector and based on identifying one or more of the multiple content creators that contributed at least a threshold amount during the training, wherein a particular creator having a contribution less than the threshold amount does not receive a creator attribution; and

initiating providing compensation to one or more of the multiple content creators based at least in part on the attribution determination;

wherein training the artificial intelligence to create the trained artificial intelligence comprises:

selecting a content creator from multiple content creators to create a selected content creator;

selecting a plurality of content items associated with the selected content creator to create selected content items;

training the artificial intelligence using the selected content items;

determining a measure of a proximity between:

the artificial intelligence trained using the selected content items, and

the artificial intelligence prior to initiating the training;

determining a creator influence of the content creator on the trained artificial intelligence based at least in part on the measure of the proximity;

determining a content item influence of individual content items of the plurality of content items associated with the content creator using a converging function comprising a loss minimization algorithm that uses a direction and a learning rate to reduce a cost function close to zero; and

including the creator influence and the content item influence in the attribution vector.

9 . The server of claim 8 , the operations further comprising:

determining a transformativeness factor of the trained artificial intelligence relative to the artificial intelligence based at least in part on the measure of the proximity; and

determining the attribution vector based at least in part on the transformativeness factor.

10 . The server of claim 8 , the operations further comprising:

while generating the output, determining:

activations of neural pathways in the trained artificial intelligence; and

an amount of individual activations.

11 . The server of claim 10 , the operations further comprising:

after determining that the trained artificial intelligence has completed generating the output, creating a dynamic attribution vector based on the activations of the neural pathways and the amount of the individual activations, the dynamic attribution vector indicating an influence of the individual content creators on the output; and

creating the attribution vector based at least in part on the dynamic attribution vector.

12 . The server of claim 8 , the operations further comprising:

storing the attribution vector in a blockchain.

13 . The server of claim 8 , wherein the training further comprises:

determining a contribution of individual content items of the selected content items associated with the selected content creator on gradients of the artificial intelligence during the training;

aggregating contributions of the selected content items; and

determining a contribution of the selected content creator based at least in part on the aggregating.

14 . The server of claim 8 , wherein the training further comprises:

determining a change caused by individual content items of the selected content items associated with the selected content creator to weights of the artificial intelligence during the training;

determining a number of activations caused by individual content items of the selected content items; and

determining a contribution of the selected content creator based at least in part on multiplying the change to the weights by the number of activations.

15 . A non-transitory computer-readable memory device to store instructions executable by one or more processors to perform operations comprising:

determining that a training of an artificial intelligence has been completed, creating a trained artificial intelligence;

after determining that the training of the artificial intelligence has been completed, generating a static attribution vector based on a relative contribution of selected content items of individual content creators;

determining an attribution vector created during the training of the artificial intelligence, the attribution vector created based at least in part on the static attribution vector;

determining that the trained artificial intelligence has received an input;

generating, by the trained artificial intelligence, an output based on the input;

determining an attribution determination for individual content creators of multiple content creators based on the attribution vector and based on identifying one or more of the multiple content creators that contributed at least a threshold amount during the training, wherein a particular creator having a contribution less than the threshold amount does not receive a creator attribution; and

initiating providing compensation to one or more of the multiple content creators based at least in part on the attribution determination;

wherein training the artificial intelligence to create the trained artificial intelligence comprises:

selecting a content creator from multiple content creators to create a selected content creator;

selecting a plurality of content items associated with the selected content creator to create selected content items;

training the artificial intelligence using the selected content items;

determining a number of activations caused to individual parameters of the artificial intelligence;

determining the relative contribution of the selected content items of individual content creators based at least in part on a change in parameters of the trained artificial intelligence caused by multiple content items with their respective activations;

determining a measure of a proximity between:

the artificial intelligence trained using the selected content items, and

the artificial intelligence prior to initiating the training;

determining a creator influence of the content creator on the trained artificial intelligence based at least in part on the measure of the proximity; and

including the creator influence in an attribution vector.

16 . The non-transitory computer-readable memory device of claim 15 , the operations further comprising:

while generating the output, determining:

activations of neural pathways in the trained artificial intelligence; and

an amount of individual activations.

17 . The non-transitory computer-readable memory device of claim 16 , further comprising:

after determining that the trained artificial intelligence has completed generating the output, creating a dynamic attribution vector based on the activations of the neural pathways and the amount of the individual activations, the dynamic attribution vector indicating an influence of the individual content creators on the output; and

creating the attribution vector based at least in part on the dynamic attribution vector.

18 . The non-transitory computer-readable memory device of claim 15 , wherein the training further comprises:

determining a content item influence of individual content items of the plurality of content items associated with the content creator using a converging function comprising a loss minimization algorithm that uses a direction and a learning rate to reduce a cost function close to zero.

19 . The non-transitory computer-readable memory device of claim 15 , wherein the training further comprises:

determining a contribution of individual content items of the selected content items associated with the selected content creator on gradients of the artificial intelligence during the training;

aggregating contributions of the selected content items; and

determining a contribution of the selected content creator based at least in part on the aggregating.

20 . The non-transitory computer-readable memory device of claim 15 , wherein the training further comprises:

determining a change caused by individual content items of the selected content items associated with the selected content creator to weights of the artificial intelligence during the training;

determining a number of activations caused by individual content items of the selected content items; and

determining a contribution of the selected content creator based at least in part on the change to the weights and the number of activations.

Assignments (1)

Continuity (6)

References Cited (12)