# US 2019/0318244 A1 · Methods and apparatus to provide machine programmed creative support to a user

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2019/0318244 A1
- Document ID: US20190318244A1
- Status: Published application
- Kind: A1
- Application number: 16/455,346
- Filed: 2019-06-27
- Published: 2019-10-17
- Earliest claimed: 2019-06-27
- Assignees: Intel Corporation
- Applicants: Intel Corporation
- Inventors: Ignacio Javier Alvarez; Javier Felip Leon; David Israel Gonzalez Aguirre; Javier Sebastian Turek; Luis Carlos Maria Remis; Justin Gottschlich
- Web page: https://patentgrep.com/patent/US20190318244A1

## CPC

- G06N 3/08: Learning methods
- G06N 3/006: based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
- G06N 3/045: Combinations of networks
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/047: Probabilistic or stochastic networks
- G06N 3/0475: Generative networks
- G06N 3/0895: Weakly supervised learning, e.g. semi-supervised or self-supervised learning
- G06N 3/09: Supervised learning
- G06N 3/092: Reinforcement learning
- G06N 5/045: Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

Methods, apparatus, systems and articles of manufacture to provide machine programmed creative support to a user are disclosed. An example apparatus include an artificial intelligence architecture to be trained based on previous inputs of the user; a processor to: implement a first machine learning model based on the trained artificial intelligence architecture; and predict a first action based on a current state of a computer program using the first machine learning model; implement a second machine learning model based on the trained artificial intelligence architecture; and predict a second action based on the current state of the computer program using the second machine learning model; and a controller to select a state based on the action that results in a state that is more divergent from the current state of the computer program.

## Claims

### Claim 1

An apparatus to generate an idea to inspire a user, the apparatus including:
an artificial intelligence architecture to be trained based on previous inputs of the user;
a processor to:
implement a first neural network based on the trained artificial intelligence architecture; and
predict a first action based on a current state of a computer program using the first neural network;
implement a second neural network based on the trained artificial intelligence architecture; and
predict a second action based on the current state of the computer program using the second neural network; and
a controller to select a state based on the action that results in a state that is more divergent from the current state of the computer program.

### Claim 2 (depends on 1)

The apparatus of claim 1, wherein the current state includes at least one of (A) items currently being rendered by the computer program or (B) a last input of the user.

### Claim 3 (depends on 1)

The apparatus of claim 1, wherein the artificial intelligence architecture is trained to predict the next action of the user based on previous inputs of the user with respect to previous states.

### Claim 4 (depends on 1)

The apparatus of claim 1, wherein the state is a first state, the artificial intelligence architecture including:
a variational autoencoder to:
convert a second state into a latent state;
generate a predicted state based on a predicted action;
a third neural network to:
predict a third action based on the latent state;
update based on a value for the third action; and
when the value for the third action is above a threshold, select the third action as the predicted action; and
a fourth neural network to:
estimate the value for the third action based on the latent state; and
update based on a reward for the previous predicted state based on the second state.

### Claim 5 (depends on 1)

The apparatus of claim 1, further including a comparator to (A) determine a first distance between a first state corresponding to the first action to the current state and (B) determine a second distance between a second state corresponding to the second action to the current state, the more divergent state being the state with the larger distance.

### Claim 6 (depends on 1)

The apparatus of claim 1, wherein the controller is to instruct a user interface to render the selected state to the user.

### Claim 7 (depends on 1)

The apparatus of claim 1, wherein the processor includes:
a first computer core to:
implement the first neural network based on the trained artificial intelligence architecture; and
predict the first action based on the current state of the computer program using the first neural network; and
a second computer core to:
implement the second neural network based on the trained artificial intelligence architecture; and
predict the second action based on the current state of the computer program using the second neural network.

### Claim 8 (depends on 7)

The apparatus of claim 7, wherein:
the first computer core is to generate a third action based on the selected state using the first neural network;
the second computer core is to generate a fourth action based on the selected state using the second neural network; and
the controller is to instruct a user interface to render the a subsequently selected state based on the action that is more divergent from the current state of the computer program.

### Claim 9 (depends on 7)

The apparatus of claim 7, wherein the artificial intelligence architecture is to deploy trained artificial intelligence information to the first and second computer core based on instructions from the controller.

### Claim 10 (depends on 9)

The apparatus of claim 9, wherein the controller is to transmit the instructions to the artificial intelligence architecture in response to at least one of a threshold amount of time or a trigger from the computer program.

### Claim 11 (depends on 1)

The apparatus of claim 1, wherein the artificial intelligence architecture is to train based on user feedback corresponding to the selected state.

### Claim 12

A method to generate an idea to inspire a user, the method including:
training, by executing an instruction with a processor, an artificial intelligence architecture based on previous inputs of the user;
implementing a first neural network based on the trained artificial intelligence architecture;
predicting, by executing an instruction with the processor, a first action based on a current state of a computer program using the first neural network;
implementing a second neural network based on the trained artificial intelligence architecture;
predicting, by executing an instruction with the processor, a second action based on the current state of the computer program using the second neural network; and
selecting, by executing an instruction with the processor, a state based on the action that results in a state that is more divergent from the current state of the computer program.

### Claim 13 (depends on 12)

The method of claim 12, wherein the current state includes at least one of (A) items currently being rendered by the computer program or (B) a last input of the user.

### Claim 14 (depends on 12)

The method of claim 12, wherein the artificial intelligence architecture is trained to predict the next action of the user based on previous inputs of the user with respect to previous states.

### Claim 15 (depends on 12)

The method of claim 12, wherein the state is a first state, further including:
converting a second state into a latent state;
generating a predicted state based on a predicted action;
predicting a third action based on the latent state;
updating based on a value for the third action;
when the value for the third action is above a threshold, selecting the third action as the predicted action;
estimating the value for the third action based on the latent state; and
updating based on a reward for the previous predicted state based on the second state.

### Claim 16 (depends on 12)

The method of claim 12, further including:
determining a first distance between a first state corresponding to the first action to the current state; and
determining a second distance between a second state corresponding to the second action to the current state, the more divergent state being the state with the larger distance.

### Claim 17 (depends on 12)

The method of claim 12, further including instructing a user interface to render the selected state to the user.

### Claim 18 (depends on 12)

The method of claim 12, further including:
generating a third action based on the selected state using the first neural network;
generating a fourth action based on the selected state using the second neural network; and
instructing a user interface to render the a subsequently selected state based on the action that is more divergent from the current state of the computer program.

### Claim 19 (depends on 12)

The method of claim 12, further including deploying trained artificial intelligence information to a first and second computer core based on instructions for a controller.

### Claim 20 (depends on 19)

The method of claim 19, further including transmitting the instructions to the artificial intelligence architecture in response to at least one of a threshold amount of time or a trigger from the computer program.

### Claim 21 (depends on 12)

The method of claim 12, further including training the artificial intelligence architecture based on user feedback corresponding to the selected state.

### Claim 22

A non-transitory computer readable storage medium comprising instructions which, when executed, cause a machine to at least:
train an artificial intelligence architecture based on previous inputs of the user;
implement a first machine learning model based on the trained artificial intelligence architecture;
predict a first action based on a current state of a computer program using the first machine learning model;
implement a second machine learning model based on the trained artificial intelligence architecture;
predict a second action based on the current state of the computer program using the second machine learning model; and
select a state based on the action that results in a state that is more divergent from the current state of the computer program.

### Claim 23 (depends on 22)

The computer readable storage medium of claim 22, wherein the current state includes at least one of (A) items currently being rendered by the computer program or (B) a last input of the user.

### Claim 24 (depends on 22)

The computer readable storage medium of claim 22, wherein the artificial intelligence architecture is trained to predict the next action of the user based on previous inputs of the user with respect to previous states.

### Claim 25 (depends on 22)

The computer readable storage medium of claim 22, wherein the state is a first state, the instructions to cause the machine to:
convert a second state into a latent state;
generate a predicted state based on a predicted action;
predict a third action based on the latent state;
update based on a value for the third action;
when the value for the third action is above a threshold, select the third action as the predicted action;
estimate the value for the third action based on the latent state; and
update based on a reward for the previous predicted state based on the second state.

### Claim 26

(canceled)

### Claim 27

(canceled)

### Claim 28

(canceled)

### Claim 29

(canceled)

### Claim 30

(canceled)

### Claim 31

(canceled)