Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
DETAILED ACTION
The instant application having Application No. 18216271 has a total of 20 claims pending in the application, of which claims 11 and 19 have been withdrawn.
I. ACKNOWLEDGEMENT OF REFERENCES CITED BY APPLICANT
Information Disclosure Statement
As required by M.P.E.P 609(c), the applicant’s submissions of the Information Disclosure Statements dated 1/14/26, 7/17/24, 7/19/23 are acknowledged by the examiner and the cited references have been considered in the examination of the claims now pending. As required by M.P.E.P 609 C(2), a copy of the PTOL-1449 initialed and dated by the examiner is attached to the instant office action.
Any references lined through were not received by the office and could not be considered.
II. REJECTIONS NOT BASED ON PRIOR ART
Claim Rejections – 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claims 1-10, 12-18 and 20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to an abstract idea without significantly more.
Step 1: Claim 1 and 5 are process type claims. Claim 13 is a machine type claim. Therefore, claims 1-10, 12-18 and 20 are directed to either a process, machine, manufacture or composition of matter.
As per claim 1,
2A Prong 1:
“determine first prompt data representing a task to generate an image, the first prompt data including a first plurality of examples usable to cause generation of the image” An Artist mentally or with pen and paper prepares descriptions of what they want to draw, and examples of different ideas).
“processing … the first prompt data to generate first … natural language” The Artist mentally or with pencil and paper considers their writing and edits it to get it in a state they want it to be in.
“Processing … the first … natural language data to generate the first image data” The Artist mentally or with pencil and paper draws a picture based upon their writing.
“Processing … the first image data to determine that the first image data corresponds to a target image” The Artist mentally or with pencil and paper looks at their drawing to see if it meets the description.
“based on the first image data corresponding to the target image, determining a second plurality of examples including the first … natural language data and at least a portion of the first plurality of examples” The Artist mentally or with pencil and paper updates their writing to contain new suggestions or ideas for drawing.
“determining training da including at least the second plurality of examples” The Artist mentally or with pencil and paper considers the updates and keeps the ones they are interested in using.
“using the training data to … for image generation” The Artist mentally or with pencil and paper uses the new ideas to update or draw further images).
2A Prong 2: This judicial exception is not integrated into a practical application.
Additional elements:
Computer implemented, machine-generated (mere instructions to apply the exception using a generic computer component);
“Using a larger language model (LLM)”, “a first machine learning model”, “A second machine learning model”, “a third machine learning model” (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: Claims describe generic machine learning models without any further limitations or details which make them anything more than generic, off the shelf machine learning models);
2B: The claim does not include additional elements individually or in combination that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
Computer implemented, machine-generated (mere instructions to apply the exception using a generic computer component)
“Using a larger language model (LLM)”, “a first machine learning model”, “A second machine learning model”, “a third machine learning model” (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: Claims describe generic machine learning models without any further limitations or details which make them anything more than generic, off the shelf machine learning models);
As per claim 2,
2A Prong 1:
“determining a first value representing a semantic difference between the first … natural language data and a first example natural language input of the first plurality of examples” The Artist mentally or with pencil and paper calculates differences between examples and other changes to the text.
“determining a second value representing a semantic difference between the first … natural language data and a second example natural language input of the first plurality of examples” The Artist mentally or with pencil and paper calculates differences between examples and other changes to the text.
“Determining a first semantic diversity value for a first list, the first list including the first example natural language input, the second example natural language input, and the first … natural language data, wherein the firs semantic diversity value is an average of the first value and the second value” The Artist mentally or with pencil and paper makes a list of the values and performs the calculation to average the values.
“Determining a third value representing a semantic difference between the first example natural language input and the second example natural language input” The Artist mentally or with pencil and paper calculates differences between examples.
“determining a fourth value representing a semantic difference between the first example natural language input and a third example natural language input of the first plurality of examples” The Artist mentally or with pencil and paper calculates differences between examples.
Determining a second semantic diversity value for a second list, the second list including the first example natural language input, the second example natural language input, and the third example natural language input, wherein the second semantic diversity value is an average of the third value and the fourth value” The Artist mentally or with pencil and paper makes a list of the values and performs the calculation to average the values.
“Based at least in part on the first semantic diversity value being greater than the second semantic diversity value, determining the second plurality of examples to include the first … natural language data, the first example natural language input, and the second example natural language input, and to exclude the third example natural language input” The Artist mentally or with pencil and paper examines the values and decides which examples to include and which to discard.
2A Prong 2: This judicial exception is not integrated into a practical application.
Additional elements:
Computer implemented, machine-generated (mere instructions to apply the exception using a generic computer component);
2B: The claim does not include additional elements individually or in combination that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
Computer implemented, machine-generated (mere instructions to apply the exception using a generic computer component)
As per claims 3-4, these claims contain similar general machine learning models and similar mental steps to claim 1, and are rejected for similar reasons.
As per claim 5,
2A Prong 1:
“determining first prompt data including a first task and a first plurality of examples” An Artist mentally or with pen and paper prepares descriptions of what they want to draw, and examples of different ideas).
“processing … the first prompt data to generate first … data” The Artist mentally or with pencil and paper considers their writing and edits it to get it in a state they want it to be in.
“Processing … the first … data data to generate the first model output” The Artist mentally or with pencil and paper draws a picture based upon their writing.
“Determining … the first model output data corresponds to a target output” The Artist mentally or with pencil and paper looks at their drawing to see if it meets the description.
“based on the first model output corresponding to the target output, determining a second plurality of examples including the first … data and at least a portion of the first plurality of examples, the second plurality of examples to be used for model training for at least the first task” The Artist mentally or with pencil and paper updates their writing to contain new suggestions or ideas for drawing, and considers how to use them in future drawings.
2A Prong 2: This judicial exception is not integrated into a practical application.
Additional elements:
Computer implemented, machine-generated (mere instructions to apply the exception using a generic computer component);
“Using a larger language model (LLM)”, “a first machine learning model”, “A second machine learning model”, (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: Claims describe generic machine learning models without any further limitations or details which make them anything more than generic, off the shelf machine learning models);
2B: The claim does not include additional elements individually or in combination that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
Computer implemented, machine-generated (mere instructions to apply the exception using a generic computer component)
“Using a larger language model (LLM)”, “a first machine learning model”, “A second machine learning model” (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: Claims describe generic machine learning models without any further limitations or details which make them anything more than generic, off the shelf machine learning models);
As per claims 6-7, 10 and 12, these claims contain additional generic machine learning models and mental steps to claim 5, and are rejected for similar reasons.
As per claims 8-9, these claims contain similar mental steps to claim 5 and are rejected for similar reasons.
As per claim 13,
2A Prong 1:
“determine first prompt data including a first task and a first plurality of examples” An Artist mentally or with pen and paper prepares descriptions of what they want to draw, and examples of different ideas).
“process … the first prompt data to generate first … data” The Artist mentally or with pencil and paper considers their writing and edits it to get it in a state they want it to be in.
“Process … the first … data to generate the first model output” The Artist mentally or with pencil and paper draws a picture based upon their writing.
“Determine … the first model output corresponds to a target output” The Artist mentally or with pencil and paper looks at their drawing to see if it meets the description.
“based on the first model output corresponding to the target output, determine a second plurality of examples including the first … data and at least a portion of the first plurality of examples, the second plurality of examples to be used for model training for at least the first task” The Artist mentally or with pencil and paper updates their writing to contain new suggestions or ideas for drawing, and considers how to use them in future drawings.
2A Prong 2: This judicial exception is not integrated into a practical application.
Additional elements:
At least one processor, at least one memory, (mere instructions to apply the exception using a generic computer component);
“Using a larger language model (LLM)”, “a first machine learning model”, “A second machine learning model”, (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: Claims describe generic machine learning models without any further limitations or details which make them anything more than generic, off the shelf machine learning models);
2B: The claim does not include additional elements individually or in combination that are sufficient to amount to significantly more than the judicial exception.
Additional elements:
At least one processor, at least one memory, (mere instructions to apply the exception using a generic computer component)
“Using a larger language model (LLM)”, “a first machine learning model”, “A second machine learning model” (Adding the words “apply it” (or an equivalent) with the judicial exception, or mere instructions to implement an abstract idea on a computer, or merely uses a computer as a tool to perform an abstract idea - see MPEP 2106.05(f) – Examiner’s note: Claims describe generic machine learning models without any further limitations or details which make them anything more than generic, off the shelf machine learning models);
As per claims 14-15, 18 and 20, these claims contain additional generic machine learning models and mental steps to claim 5, and are rejected for similar reasons.
As per claims 16-17, these claims contain similar mental steps to claim 5 and are rejected for similar reasons.
Claim Rejections - 35 USC § 112
III. REJECTIONS BASED ON PRIOR ART
Examiners Note: Some rejections will be followed by an ‘EN’ that will denote an examiners note. This will be placed to further explain a rejection.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1, 4-6, 12, 13-14, and 20 are rejected under 35 U.S.C. 103 as being unpatentable over Hao et al (“Optimizing Prompts for Text-to-Image Generation””) in view of Dong et al (“I2T2I: Learning Text to Image Synthesis with Textual data Augmentation”).
As per claim 1, Zhou discloses, “A computer-implemented method comprising:” (Pg.7, particularly “learning from human preference” section; EN: this denotes this being machine learning, which will inherently include some form of computer to perform the machine learning).
“Determining first prompt data representing a task to generate an image” (Pg.2 particularly Section 2; EN: this denotes the prompts being used for the task of creating images). “the first prompt data including a first plurality of examples to generate an image” (Pg.2-3, particularly section 2.1; EN: this denotes the prompts being user inputs and manually engineered examples).
“Processing, using a large language model (LLM), the first prompt” (Pg.2, particularly section 2; EN: this denotes it being a GPT model, a type of LLM). “to generate first machine generated natural language data” ( Pg.2-3, particularly section 2.1; EN: this denotes using the GPT to optimize the prompts).
“Processing, using a first machine learning model configured to generate image data, the first machine generated natural language data to determine first image data” (Pg.3, particularly section 2.2; EN: this denotes using the prompts with text to image models in order to create images).
“processing, using a second machine learning model, the first image data to determine that the first image data corresponds to a target image” (Pg.3, particularly section 2.2; EN: this denotes using a CLIP model to determine if the image properly matches the prompts).
“Based on the first image data corresponding to the target image, determining a second plurality of examples including the first machine generated natural language data…” (Pg.4-5, particularly section 3.3; EN: this denotes the system identifying optimized prompts).
However, Hao fails to explicitly disclose, “…and at least portion of the first plurality of examples”, “determining training data including at least the second plurality of examples; and using the training data to configure a third machine learning model for image data generation.”
Dong discloses, “…and at least portion of the first plurality of examples” (Pg.2017, particularly section 3.1; EN: this denotes using original samples as well as new synthetizes examples to train the model).
“determining training data including at least the second plurality of examples; and using the training data to configure a third machine learning model for image data generation” (Abstract; EN: this denotes using synthetic data created by other models, such as the data of the Hao reference, in order to improve text to image generation).
Hao and Dong are analogous art because both involve text to image generation.
Before the effective filing date it would have been obvious to one skilled in the art of text to image generation to combine the work of Hao and Dong in order to make use of synthetically created data to improve text to image generation.
The motivation for doing so would be to “generate better multi-categories images using MSCOCO than the state-of the-art” (Dong, Abstract) or in the case of Hao, allow the system to use the newly created prompts and images to better train text-to-image models and improve their image generation.
Therefore before the effective filing date it would have been obvious to one skilled in the art of text to image generation to combine the work of Hao and Dong in order to make use of synthetically created data to improve text to image generation.
As per claim 4, Hao discloses, “determining second prompt data representing the task to generate an image, the second prompt data including the second plurality of examples” (Figure 1 and associated paragraphs; EN: this denotes optimizing the prompts via reinforcement learning, which includes taking steps with the prompts, optimizing, then repeating the process to reach an optimized prompt).
“processing, using the LLM, the second prompt data to generate second machine generated natural language data” ( Pg.2-3, particularly section 2.1; EN: this denotes using the GPT to optimize the prompts, which will be repeated with the reinforcement learning process).
“processing, using the first machine learning model, the second machine generated natural language data to determine second image data” (Pg.3, particularly section 2.2; EN: this denotes using the prompts with text to image models in order to create images).
“processing, using the second machine learning model, the second image data to determine that the second image data corresponds to the target image” (Pg.3, particularly section 2.2; EN: this denotes using a CLIP model to determine if the image properly matches the prompts).
“based on the second image data corresponding to the target image, determining a third plurality of examples including the first machine generated natural language data, the second machine generated natural language data, and at least a portion of the second plurality of examples” (Pg.4-5, particularly section 3.3; EN: this denotes the system identifying optimized prompts).
As per claims 5 and 13, Hao discloses, “A computer implemented method comprising:” (Pg.7, particularly “learning from human preference” section; EN: this denotes this being machine learning, which will inherently include some form of computer to perform the machine learning).
“Determining first prompt data including a first task” (Pg.2 particularly Section 2; EN: this denotes the prompts being used for the task of creating images). “and a first plurality of examples” (Pg.2-3, particularly section 2.1; EN: this denotes the prompts being user inputs and manually engineered examples).
“processing, using a large language model (LLM), the first prompt data” (Pg.2, particularly section 2; EN: this denotes it being a GPT model, a type of LLM). “to determine first machine-generated data” ( Pg.2-3, particularly section 2.1; EN: this denotes using the GPT to optimize the prompts).
“Processing, using a first machine learning model, the first machine generated data to determine first model output” (Pg.3, particularly section 2.2; EN: this denotes using the prompts with text to image models in order to create images).
“determining, using a second machine learning model, that the first model output corresponds to a target output” (Pg.3, particularly section 2.2; EN: this denotes using a CLIP model to determine if the image properly matches the prompts).
“based on the first model output corresponding to the target output, determining a second plurality of examples including the first machine generated data…” (Pg.4-5, particularly section 3.3; EN: this denotes the system identifying optimized prompts).
Hao fails to explicitly disclose, “and at least a portion to he first plurality of examples, the second plurality of examples to be used for model training for at least the first task”
Dong discloses, “and at least a portion to he first plurality of examples” (Pg.2017, particularly section 3.1; EN: this denotes using original samples as well as new synthetizes examples to train the model).
“the second plurality of examples to be used for model training for at least the first task” (Abstract; EN: this denotes using synthetic data created by other models, such as the data of the Hao reference, in order to improve text to image generation).
Hao and Dong are analogous art because both involve text to image generation.
Before the effective filing date it would have been obvious to one skilled in the art of text to image generation to combine the work of Hao and Dong in order to make use of synthetically created data to improve text to image generation.
The motivation for doing so would be to “generate better multi-categories images using MSCOCO than the state-of the-art” (Dong, Abstract) or in the case of Hao, allow the system to use the newly created prompts and images to better train text-to-image models and improve their image generation.
Therefore before the effective filing date it would have been obvious to one skilled in the art of text to image generation to combine the work of Hao and Dong in order to make use of synthetically created data to improve text to image generation.
As per claims 6 and 14, Hao discloses, “processing, using the second machine learning model, the first model output to determine a value representing a likelihood of the first model output corresponding to the target output” (Pg.3, particularly section 2.2; EN: this denotes using a CLIP model to determine if the image properly matches the prompts).
As per claims 12 and 20, Hao discloses, “Wherein the first machine learning model is configured to generate the image” (Pg.2 particularly Section 2; EN: this denotes the prompts being used for the task of creating images).
“wherein the first model output represents the image” (Pg.2 particularly Section 2; EN: this denotes the prompts being used for the task of creating images).
Claim Rejections - 35 USC § 103
Claims 3, 7-10, and 15-18 are rejected under 35 U.S.C. 103 as being unpatentable over Hao et al (“Optimizing Prompts for Text-to-Image Generation””) in view of Dong et al (“I2T2I: Learning Text to Image Synthesis with Textual data Augmentation”) and further in view of Strobelt et al (“Interactive and Visual Prompt Engineering for Ad-hoc Task Adaption with Large Language Models”).
As per claim 3, Hao discloses, “processing, using the second machine learning model, the first image data to determine a first value representing a likelihood that the first image data corresponds to the target image” (Pg.3, particularly section 2.2; EN: this denotes using a CLIP model to determine if the image properly matches the prompts).
“determining a second value representing a likelihood of a first example natural language input of the first plurality of examples resulting in generation of second image data that corresponds to the target image” (Pg.3, particularly section 2.2; EN: this denotes using a CLIP model to determine if the image properly matches the prompts. Pg.5, particularly the second paragraph; EN: this denotes comparing the original prompts and their generation as well).
“Determining that the first value is greater than the second value” (Pg.5, particularly the second paragraph; EN: this denotes comparing the original prompts and their generation as well, and finding that the generated prompts perform better).
Hao fails to explicitly disclose, “based on determining that the first value is greater than the second value, determining the second plurality of examples to include the first machine generated natural language data and to exclude the first example natural language input.”
Strobelt discloses, “based on determining that the first value is greater than the second value, determining the second plurality of examples to include the first machine generated natural language data and to exclude the first example natural language input” (Pg.1153, particularly section 6.2; EN: this denotes evaluating prompts and discarding the ones that underperform).
Hao and Strobelt are analogous art because both involve prompt engineering.
Before the effective filing date it would have been obvious to one skilled in the art of prompt engineering to combine the work Hao and Strobelt in order to discard prompts that don’t perform as well.
The motivation for doing so would be because “We explore prompts that introduce four answer choices … Figure 11 shows that ‘choose between A, B, C, and D’ consistently gives worse results than the other two variations, independently of how the input is introduced (q1). We discard this variation” (Strobelt, Pg.1153, section 6.2) or in the case of Hao, allow the system to discard and not use prompts which under perform.
Therefore before the effective filing date it would have been obvious to one skilled in the art of prompt engineering to combine the work Hao and Strobelt in order to discard prompts that don’t perform as well.
As per claims 7 and 15, Hao discloses, “wherein a first example of the first plurality of examples is associated with a first score representing a likelihood of a second model output, generated using the first example, corresponding to the target output” (Pg.3, particularly section 2.2; EN: this denotes using a CLIP model to determine if the image properly matches the prompts. Pg.5, particularly the second paragraph; EN: this denotes comparing the original prompts and their generation as well).
“the method further comprises: determining, based on the value, a second score associated with the first machine generated data” (Pg.3, particularly section 2.2; EN: this denotes using a CLIP model to determine if the image properly matches the prompts).
Hao fails to explicitly disclose, “based at least in part on the first score and the second score, determining the second plurality of examples to include the first machine generated data and to exclude the first example”.
Strobelt discloses, “based at least in part on the first score and the second score, determining the second plurality of examples to include the first machine generated data and to exclude the first example” (Pg.1153, particularly section 6.2; EN: this denotes evaluating prompts and discarding the ones that underperform).
Hao and Strobelt are analogous art because both involve prompt engineering.
Before the effective filing date it would have been obvious to one skilled in the art of prompt engineering to combine the work Hao and Strobelt in order to discard prompts that don’t perform as well.
The motivation for doing so would be because “We explore prompts that introduce four answer choices … Figure 11 shows that ‘choose between A, B, C, and D’ consistently gives worse results than the other two variations, independently of how the input is introduced (q1). We discard this variation” (Strobelt, Pg.1153, section 6.2) or in the case of Hao, allow the system to discard and not use prompts which under perform.
Therefore before the effective filing date it would have been obvious to one skilled in the art of prompt engineering to combine the work Hao and Strobelt in order to discard prompts that don’t perform as well.
As per claim 8 and 16, Hao fails to explicitly disclose, “determining a first value representing a semantic difference between the first machine generated data and a first example of the first plurality of examples”, “Determining a second value representing asemantic difference between the first machine generated data and a second example of the first plurality of examples”, “determining a third value representing a semantic difference between the first example and the second example”, “based at least in part on the first value, the second value, and the third value, determining the second plurality of examples to include the first machine generated data, and the first example, and to exclude the second example.”
Strobelt discloses, “determining a first value representing a semantic difference between the first machine generated data and a first example of the first plurality of examples” (Pg.1153, particularly section 6.2, figure 11 and associated paragraphs; EN: this denotes looking at various inputs for differences, and scoring them in order to determine which perform well and which do not).
“Determining a second value representing asemantic difference between the first machine generated data and a second example of the first plurality of examples” (Pg.1153, particularly section 6.2, figure 11 and associated paragraphs; EN: this denotes looking at various inputs for differences, and scoring them in order to determine which perform well and which do not).
“determining a third value representing a semantic difference between the first example and the second example” (Pg.1153, particularly section 6.2, figure 11 and associated paragraphs; EN: this denotes looking at various inputs for differences, and scoring them in order to determine which perform well and which do not).
“based at least in part on the first value, the second value, and the third value, determining the second plurality of examples to include the first machine generated data, and the first example, and to exclude the second example” (Pg.1153, particularly section 6.2; EN: this denotes evaluating prompts and discarding the ones that underperform).
Hao and Strobelt are analogous art because both involve prompt engineering.
Before the effective filing date it would have been obvious to one skilled in the art of prompt engineering to combine the work Hao and Strobelt in order to discard prompts that don’t perform as well.
The motivation for doing so would be because “We explore prompts that introduce four answer choices … Figure 11 shows that ‘choose between A, B, C, and D’ consistently gives worse results than the other two variations, independently of how the input is introduced (q1). We discard this variation” (Strobelt, Pg.1153, section 6.2) or in the case of Hao, allow the system to discard and not use prompts which under perform.
Therefore before the effective filing date it would have been obvious to one skilled in the art of prompt engineering to combine the work Hao and Strobelt in order to discard prompts that don’t perform as well.
As per claims 9 and 17, Hao fails to explicitly disclose, “Wherein the first plurality of examples is an ordered list of examples with a first example being last in the ordered list, and the method further comprises: determining the second plurality of examples by replacing the first example with the first machine generated data in the ordered list”
Strobelt discloses, “Wherein the first plurality of examples is an ordered list of examples with a first example being last in the ordered list, and the method further comprises” (Pg.1153, particularly section 6.2, figure 11 and associated paragraphs; EN: this denotes looking at various inputs for differences, and scoring them in order to determine which perform well and which do not. Here figure 11 denotes an ordered list of the various forms with the poor performers at the end of the list).
“determining the4 second plurality of examples by replacing the first example with the first machine generated data in the ordered list” (Pg.1153, particularly section 6.2; EN: this denotes evaluating prompts and discarding the ones that underperform).
Hao and Strobelt are analogous art because both involve prompt engineering.
Before the effective filing date it would have been obvious to one skilled in the art of prompt engineering to combine the work Hao and Strobelt in order to discard prompts that don’t perform as well.
The motivation for doing so would be because “We explore prompts that introduce four answer choices … Figure 11 shows that ‘choose between A, B, C, and D’ consistently gives worse results than the other two variations, independently of how the input is introduced (q1). We discard this variation” (Strobelt, Pg.1153, section 6.2) or in the case of Hao, allow the system to discard and not use prompts which under perform.
Therefore before the effective filing date it would have been obvious to one skilled in the art of prompt engineering to combine the work Hao and Strobelt in order to discard prompts that don’t perform as well.
As per claims 10 and 18, Hao discloses, “determining second prompt data including the first task and the second plurality of examples” (Figure 1 and associated paragraphs; EN: this denotes optimizing the prompts via reinforcement learning, which includes taking steps with the prompts, optimizing, then repeating the process to reach an optimized prompt).
“processing, using the LLM, the second prompt data to determine second machine generated data” ( Pg.2-3, particularly section 2.1; EN: this denotes using the GPT to optimize the prompts, which will be repeated with the reinforcement learning process).
“processing, using the first machine learning model, the second machine generated data to determine second model output” (Pg.3, particularly section 2.2; EN: this denotes using the prompts with text to image models in order to create images).
“determining that the second model does not correspond to the target output” (Pg.3, particularly section 2.2; EN: this denotes using a CLIP model to determine if the image properly matches the prompts).
Hao fails to explicitly disclose, “based on the second model output not corresponding to the target output, discarding the second machine generated data.”
Strobelt discloses, “based on the second model output not corresponding to the target output, discarding the second machine generated data” (Pg.1153, particularly section 6.2; EN: this denotes evaluating prompts and discarding the ones that underperform).
Hao and Strobelt are analogous art because both involve prompt engineering.
Before the effective filing date it would have been obvious to one skilled in the art of prompt engineering to combine the work Hao and Strobelt in order to discard prompts that don’t perform as well.
The motivation for doing so would be because “We explore prompts that introduce four answer choices … Figure 11 shows that ‘choose between A, B, C, and D’ consistently gives worse results than the other two variations, independently of how the input is introduced (q1). We discard this variation” (Strobelt, Pg.1153, section 6.2) or in the case of Hao, allow the system to discard and not use prompts which under perform.
Therefore before the effective filing date it would have been obvious to one skilled in the art of prompt engineering to combine the work Hao and Strobelt in order to discard prompts that don’t perform as well.
Allowable Subject Matter
Claim 2 will not be receiving an art rejection. It would not be obvious to one of ordinary skill in the art to take in the particular combination of calculations comparisons, and averages and taking the actions described based on those calculations in the claim. Should the claims be amended to overcome the rejection under U.S.C. 101, the claim would be found allowable over the prior art if it were rolled up into the independent claim with any intervening claims.
Conclusion
The examiner requests, in response to this Office action, support be shown for language added to any original claims on amendment and any new claims. That is, indicate support for newly added claim language by specifically pointing to page(s) and line no(s) in the specification and/or drawing figure(s). This will assist the examiner in prosecuting the application.
When responding to this office action, Applicant is advised to clearly point out the patentable novelty which he or she thinks the claims present, in view of the state of the art disclosed by the references cited or the objections made. He or she must also show how the amendments avoid such references or objections See 37 CFR 1.111(c).
Any inquiry concerning this communication or earlier communications from the examiner should be directed to BEN M RIFKIN whose telephone number is (571)272-9768. The examiner can normally be reached Monday-Friday 9 am - 5 pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Alexey Shmatov can be reached at (571) 270-3428. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/BEN M RIFKIN/ Primary Examiner, Art Unit 2123
Prosecution Insights