DETAILED ACTION
This office action is in response to the application filed on December 1, 2023.
Claims 1-51 have been cancelled. Claims 52-71 are pending and have been examined. Claims 52-71 are rejected.
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 .
Information Disclosure Statement
Acknowledgment is made of the information disclosure statements filed December 29, 2023, which comply with 37 CFR 1.97. As such, the information disclosure statements have been placed in the application file and the information referred to therein has been considered by the examiner.
Claim Objections
Claim(s) 66 and 68 are objected to because of the following informalities:
In Claim 66, the recitation of “comparison of the generated prediction with obtained prediction” is grammatically incorrect and appears to be missing the word “an” before “obtained prediction”. It appears these recitations should read “comparison of the generated prediction with an obtained prediction”. Appropriate correction is required.
In Claim 68, the recitation of “and calculate robustness metric for the AI system in the testing state based on the number of correct predictions or the number of incorrect” is grammatically incorrect and appears to be missing the word “a” before “robustness metric”. It appears these recitations should read “and calculate a robustness metric for the AI system in the testing state based on the number of correct predictions or the number of incorrect”. Appropriate correction is required.
Claim Rejections - 35 USC § 112
The following is a quotation of 35 U.S.C. § 112(b):
(b) CONCLUSION – The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. § 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claim(s) 53, 68, and 70 are rejected under 35 U.S.C. 112(b) as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
Regarding Claim 53’s limitation "process outputs generated by other ones of the self-assessment entities in a human-consumable format”, there is insufficient antecedent basis for this limitation in the claim, rendering the claim indefinite because Claim 52 (which Claim 53 is dependent on) only mentions a self-verification entity, such as “memory circuitry to store program code of a set of self-assessment entities including a self-verification entity”, where it is not clear what “other ones of the self-assessment entities” in Claim 53 is referencing from Claim 52.
Regarding Claim 68’s limitation "and calculate robustness metric for the AI system in the testing state based on the number of correct predictions or the number of incorrect”, there is insufficient antecedent basis for this limitation in the claim, rendering the claim indefinite because it is not clear whether “the number of incorrect” in Claim 68 is referencing a number of incorrect predictions, a number of incorrect data items, or some other quantity.
Regarding Claim 70, the limitation "wherein the set of AIMER functions includes an AI system quality manager (AISQM), and execution of the instructions is to cause the compute node to operate the AISQM to: calculate a quality metric for the AI system based on the fingerprint, the accuracy metric, and the robustness metric; and issue one or more remedial actions to the AI system when the quality metric is below a threshold” does not clearly set the metes and bounds of the patent protection desired. There is insufficient antecedent basis for this limitation in the claim, rendering the claim indefinite because “the accuracy metric” and “the robustness metric” are not defined in Claim 69, which Claim 70 is dependent on.
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 52-71 are rejected under 35 U.S.C. 101 because the claimed invention is directed to non-statutory subject matter.
According to the USPTO guidelines, a claim is directed to non-statutory subject matter if:
Step 1: The claim does not fall within one of the four statutory categories of invention (process, machine, manufacture, or composition of matter) – see MPEP 2106.03, or,
Step 2: The claim recites a judicial exception, e.g. an abstract idea, without reciting additional elements that amount to significantly more than the judicial exception, as determined using the following analysis – see MPEP 2106.04:
Step 2A, Prong 1: Does the claim recite an abstract idea, law of nature, or natural phenomenon?
Step 2A, Prong 2: Does the claim recite additional elements that integrate the judicial exception into a practical application?
Step 2B: Does the claim recite additional elements that amount to significantly more than the judicial exception? - see MPEP 2106.05
MPEP 2106.04(a)(2)(I) states: “The mathematical concepts grouping is defined as mathematical relationships, mathematical formulas or equations, and mental processes.”
MPEP 2106.04(a)(2)(III) states: “Accordingly, the “mental processes” abstract idea grouping is defined as concepts performed in the human mind, and examples of mental processes include observations, evaluations, judgements, and opinions.
Further, the MPEP states: “The courts do not distinguish between mental processes that are performed entirely in the human mind and mental processes that require a human to use a physical aid (e.g. pen and paper or a slide run) to perform the claim limitation.
Using the two-step inquiry, it is clear that Claims 52-71 are each directed to non-statutory subject matter as shown below:
With respect to Claim 52:
Step 1: The claim is directed to a method, also known as a process, which is one of the four statutory categories of patentable subject matter.
Step 2A, Prong 1: A judicial exception is recited in the claim as it recites mental processes, which are abstract ideas:
“and stop or pause the operation of the AI engine circuitry when biased predictions are generated by the AI engine circuitry based on the predefined test dataset.” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“An apparatus of an artificial intelligence (AI) system, comprising: AI engine circuitry to operate a machine learning (ML) model;” (AI engine circuitry to operate a machine learning (ML) model is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process - see MPEP §§ 2106.04(d) and MPEP §§ 2106.05(f)(2).)
“memory circuitry to store program code of a set of self-assessment entities including a self-verification entity;” (Memory circuitry dataset is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process - see MPEP §§ 2106.04(d) and MPEP §§ 2106.05(f)(2). Using memory circuitry to store program code of a set of self-assessment entities including a self-verification entity is akin to storing data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).)
“and processor circuitry connected to the memory circuitry, wherein the processor circuitry is to operate the self-verification entity to: cause the AI engine circuitry to operate the ML model using a predefined test dataset,” (Processor circuitry that operates a self-verification entity to cause an AI engine circuitry to operate an ML model using a predefined test dataset is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process - see MPEP §§ 2106.04(d) and MPEP §§ 2106.05(f)(2).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. AI engine circuitry to operate a machine learning (ML) model is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process - see MPEP §§ 2106.04(d) and MPEP §§ 2106.05(f)(2).
Memory circuitry dataset is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process - see MPEP §§ 2106.04(d) and MPEP §§ 2106.05(f)(2). Using memory circuitry to store program code of a set of self-assessment entities including a self-verification entity is akin to storing data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).)
Processor circuitry that operates a self-verification entity to cause an AI engine circuitry to operate an ML model using a predefined test dataset is an additional element that amounts to adding the words “apply it” (or an equivalent) with the judicial exception, or merely uses a computer in its ordinary capacity as a tool to perform an existing process - see MPEP §§ 2106.04(d) and MPEP §§ 2106.05(f)(2).)
With respect to Claim 53:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 52.
Step 2A, Prong 2: The claim(s) do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the set of self-assessment entities includes a risk-related information (RRI) processing entity, and the processor circuitry is to operate the RRI processing entity to:” (A set of self-assessment entities that includes a risk-related information (RRI) processing entity, and where the processor circuitry is to operate the RRI processing entity generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
“process outputs generated by other ones of the self-assessment entities in a human-consumable format;” (Processing outputs generated by other ones of self-assessment entities in a human-consumable format which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).)
“and present the processed outputs to an authorized user.” (Presenting processed outputs to an authorized user is akin to outputting and presenting data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).)
Step 2B: The claim(s) do not recite additional elements that amount to significantly more than the judicial exception. A set of self-assessment entities that includes a risk-related information (RRI) processing entity, and where the processor circuitry is to operate the RRI processing entity generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h). Processing outputs generated by other ones of self-assessment entities in a human-consumable format which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).). Presenting processed outputs to an authorized user is akin to outputting and presenting data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).
With respect to Claim 54:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 52. An additional judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
determine trade-offs between risks of using the ML model versus functionality or efficiencies of operating the ML model.” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the set of self-assessment entities includes a risk mitigation entity, and the processor circuitry is to operate the risk mitigation entity to: (A set of self-assessment entities that includes a risk mitigation entity, and where the processor circuitry is to operate the risk mitigation entity generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. A set of self-assessment entities that includes a risk mitigation entity, and where the processor circuitry is to operate the risk mitigation entity generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).
With respect to Claim 55:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 52. An additional judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
orchestrate internal processes of the AI system and orchestrate interactions between individual components of a set of components of the AI system or the ML model.” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the set of self-assessment entities includes an AI system management entity, and the processor circuitry is to operate the AI system management entity to: (A set of self-assessment entities that includes an AI system management entity, and where the processor circuitry is to operate the AI system management entity generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. A set of self-assessment entities that includes an AI system management entity, and where the processor circuitry is to operate the AI system management entity generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).
With respect to Claim 56:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 55. An additional judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
detect, during the operation of the AI system, a malfunctioning component of the set of components;” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
“and replace the malfunctioning component with another component that fulfills a same or similar function as the malfunctioning component.” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the set of self-assessment entities includes an AI system redundancy entity, and the processor circuitry is to operate the AI system redundancy entity to: (A set of self-assessment entities that includes an AI system redundancy entity, and where the processor circuitry is to operate the AI system redundancy entity generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. A set of self-assessment entities that includes an AI system redundancy entity, and where the processor circuitry is to operate the AI system redundancy entity generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).
With respect to Claim 57:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 55. An additional judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
“and issue the action to one or more components of the set of components to be executed by the one or more components.” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the set of self-assessment entities includes a human oversight entity, and the processor circuitry is to operate the human oversight entity to: (A set of self-assessment entities that includes a risk-related information (RRI) processing entity, and where the processor circuitry is to operate the RRI processing entity generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
provide information about potential biases in predictions generated by the ML model to an authorized user via a user interface;” (Providing information generated by the ML model to a user is akin to the outputting of data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).)
“receive a selected action based on the provided information;” (Receiving a selected action based on the provided information is akin to the inputting of data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. A set of self-assessment entities that includes a risk-related information (RRI) processing entity, and where the processor circuitry is to operate the RRI processing entity generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h). Providing information generated by the ML model to a user is akin to the outputting of data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g). Receiving a selected action based on the provided information is akin to the inputting of data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).
With respect to Claim 58:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 52. An additional judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
track interactions with the AI system, wherein the interactions include one or more of user activity, behavior of the AI system or the ML model, and information on training or testing the ML model;” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
“and logging the tracked interactions in one or more records.” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the set of self-assessment entities includes a record keeping entity, and the processor circuitry is to operate the record keeping entity to: (A set of self-assessment entities that includes a record keeping entity, and where the processor circuitry is to operate the record keeping entity generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. A set of self-assessment entities that includes a record keeping entity, and where the processor circuitry is to operate the record keeping entity generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).
With respect to Claim 59:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 52.
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the AI engine circuitry comprises one or more of an inference engine, a recommendation engine, a reinforcement learning agent, a neural network engine, a neural co-processor, a hardware accelerator, a graphics processing unit, or a general-purpose processor.” (An AI engine circuitry comprising one or more of an inference engine, a recommendation engine, a reinforcement learning agent, a neural network engine, a neural co- processor, a hardware accelerator, a graphics processing unit, or a general-purpose processor generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. An AI engine circuitry comprising one or more of an inference engine, a recommendation engine, a reinforcement learning agent, a neural network engine, a neural co- processor, a hardware accelerator, a graphics processing unit, or a general-purpose processor generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).
With respect to Claim 60:
Step 1: Claim 60 is directed to an apparatus, which is one of the four statutory categories of patentable subject matter.
Step 2A, Prong 1: A judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
“A non-transitory computer readable medium (NTCRM) comprising instructions for operating a set of artificial intelligence (AI) system monitoring, evaluation, and reporting (AIMER) functions including an AI risk management system function (AIRMS), wherein execution of the instructions by one or more processors is to cause a compute node to operate the AIRMS to:
“and issuing one or more corrective actions to the AI system when the monitored outputs include potential biases, wherein the one or more corrective actions include adjusting one or more parameters to reduce or eliminate the potential biases.” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
monitor outputs generated by an AI system implemented by the compute node;” (Monitoring outputs generated by an AI system implemented by the compute node is considered passing data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. Monitoring outputs generated by an AI system implemented by the compute node is considered passing data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).
With respect to Claim 61:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 60. An additional judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
“wherein execution of the instructions is to cause the compute node to operate the AIRMS to:
monitor inputs provided to the AI system;” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
“and issuing one or more other corrective actions to the AI system when the monitored inputs include potential errors, wherein the one or more other corrective actions include adjusting one or more parameters of the inputs to correct the potential errors.” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application.
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception.
With respect to Claim 62:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 60. An additional judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
validate an input dataset before the input dataset is provided to the AI system;” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
“and tag the input dataset with a digital certificate when the input dataset is properly validated, (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
and wherein the AI system is to verify the input dataset using the digital certificate, and generate a prediction using the input dataset when the input dataset is properly verified.” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the set of AIMER functions includes a data verification component (DVC), and execution of the instructions is to cause the compute node to operate the DVC to: (A set of AIMER functions that include a data verification component (DVC), and execution of the instructions is to cause the compute node to operate the DVC generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. A set of AIMER functions that include a data verification component (DVC), and execution of the instructions is to cause the compute node to operate the DVC generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).
With respect to Claim 63:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 60. An additional judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
“process the inputs, the outputs, and the internal states to generate statistics or metrics related to the inputs, the outputs, and the internal states;” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the set of AIMER functions includes an entity for record keeping (ERK), and execution of the instructions is to cause the compute node to operate the ERK to: (A set of AIMER functions that include an entity for record keeping (ERK), and execution of the instructions is to cause the compute node to operate the ERK generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
obtain inputs to the AI system, outputs generated by the AI system, and internal states corresponding to the inputs or the outputs;” (Obtaining inputs, outputs, and internal states corresponding to the inputs or the outputs is akin to the receiving and outputting of data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).)
“and store the inputs, the outputs, the internal states, and the statistics or metrics in a local or remote database.” (Storing inputs, outputs, internal states, and statistics or metrics in a local remote database is akin to the storing of data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. Obtaining inputs, outputs, and internal states corresponding to the inputs or the outputs is akin to the receiving and outputting of data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g). Storing inputs, outputs, internal states, and statistics or metrics in a local remote database is akin to the storing of data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).
With respect to Claim 64:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 60. An additional judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
generate transparency data including one or more of capability information of the AI system, maintenance and care information related to the AI system, self-assessment information related to the AI system, and historic data related to the operation of the AI system;” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
“and generate user interface data to present the transparency data, wherein the user interface data includes one or more of text data, image data, audio data, and video data;” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the set of AIMER functions includes an entity for transparency and information (ETI), and execution of the instructions is to cause the compute node to operate the ETI to: (A set of AIMER functions that include an entity for transparency and information (ETI), and execution of the instructions is to cause the compute node to operate the ETI generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
“and send the user interface data to an authorized user.” (Sending data to a user adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. A set of AIMER functions that include an entity for transparency and information (ETI), and execution of the instructions is to cause the compute node to operate the ETI generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h). Sending data to a user adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).
With respect to Claim 65:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 60. A judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
perform a self-verification process on a prediction generated by the AI system before the prediction is provided to an external entity, (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
wherein the self-verification process includes: comparison of the generated prediction with one or more historical predictions; and determination of biases in the generated prediction based on a divergence of the generated prediction from the one or more historical predictions.” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the set of AIMER functions includes an entity for AI output self-verification (EAIOSV), and execution of the instructions is to cause the compute node to operate the EAIOSV to: (A set of AIMER functions that include an entity for AI output self-verification (EAIOSV), and execution of the instructions causes the compute node to operate the EAIOSV generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. A set of AIMER functions that include an entity for AI output self-verification (EAIOSV), and execution of the instructions causes the compute node to operate the EAIOSV generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).
With respect to Claim 66:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 65. An additional judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
“comparison of the generated prediction with obtained prediction;” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
“and determination of biases in the generated prediction based on a divergence of the
generated prediction from the obtained prediction.” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
“wherein the self-verification process includes: operation of an alternation function to change one or more parameters of the AI system;” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“obtaining a prediction from the AI system with the changed one or more parameters;” (Obtaining a prediction from an AI system with changed one or more parameters is akin to the outputting of a value/information, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. Obtaining a prediction from an AI system with changed one or more parameters is akin to the outputting of a value/information, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).
With respect to Claim 67:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 60. An additional judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
place the AI system in a testing state;” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
“compare outputs generated by the AI system in the testing state with known outputs for the test dataset;” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
“and calculate an accuracy metric for the AI system in the testing state based on a number of correct predictions in the generated outputs or a number of incorrect predictions in the generated outputs.” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the set of AIMER functions includes an accuracy verification entity (AVE), and execution of the instructions is to cause the compute node to operate the AVE to: (A set of AIMER functions that include an accuracy verification entity (AVE), and execution of the instructions is to cause the compute node to operate the AVE generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
“provide a test dataset to the AI system in the testing state;” (Providing a test dataset to an AI system in a testing state is akin to the inputting of data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. A set of AIMER functions that include an accuracy verification entity (AVE), and execution of the instructions is to cause the compute node to operate the AVE generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h). Providing a test dataset to an AI system in a testing state is akin to the inputting of data, which adds insignificant extra-solution activity to the judicial exception, as discussed in MPEP 2106.05(g).
With respect to Claim 68:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 67. An additional judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
modify the test dataset to include one or more erroneous data items;” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
“compare outputs generated by the AI system in the testing state with known outputs for the test dataset;” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
“and calculate robustness metric for the AI system in the testing state based on the number of correct predictions or the number of incorrect, (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the number of correct predictions includes one or more correctly identified errors based on the one or more erroneous data items and the number of incorrect predictions includes one or more unidentified errors based on the one or more erroneous data items.” (A number of correct predictions including one or more correctly identified errors based on one or more erroneous data items and a number of incorrect predictions including one or more unidentified errors based on the one or more erroneous data items generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
“wherein the set of AIMER functions includes a robustness verification entity (RVE), and execution of the instructions is to cause the compute node to operate the RVE to: (A set of AIMER functions that include a robustness verification entity (RVE), and execution of the instructions causes the compute node to operate the RVE generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. A number of correct predictions including one or more correctly identified errors based on one or more erroneous data items and a number of incorrect predictions including one or more unidentified errors based on the one or more erroneous data items generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h). A set of AIMER functions that include a robustness verification entity (RVE), and execution of the instructions causes the compute node to operate the RVE generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).
With respect to Claim 69:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 60. An additional judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
generate a fingerprint for the AI system based on one or more inputs to the AI system, outputs generated by the AI system, and one or more internal system states of the AI system;” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
“and encrypt data to be conveyed between the AI system and the set of AIMER functions or communicated to external devices.” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the set of AIMER functions includes cryptographic engine (CE), and execution of the instructions is to cause the compute node to operate the CE to: (A set of AIMER functions that include a cryptographic engine (CE), and execution of the instructions causes the compute node to operate the CE generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. A set of AIMER functions that include a cryptographic engine (CE), and execution of the instructions causes the compute node to operate the CE generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).
With respect to Claim 70:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 69. An additional judicial exception is recited in the claims as they recite mental processes, which are abstract ideas:
calculate a quality metric for the AI system based on the fingerprint, the accuracy metric, and the robustness metric;” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
“and issue one or more remedial actions to the AI system when the quality metric is below a threshold.” (Covers mental concepts that could be practically performed in the human mind or through the use of a pencil and paper, including observation, evaluation, and judgement - see MPEP 2106.04.)
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the set of AIMER functions includes an AI system quality manager (AISQM), and execution of the instructions is to cause the compute node to operate the AISQM to: (A set of AIMER functions that include an AI system quality manager (AISQM), and execution of the instructions is to cause the compute node to operate the AISQM generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. A set of AIMER functions that include an AI system quality manager (AISQM), and execution of the instructions is to cause the compute node to operate the AISQM generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).
With respect to Claim 71:
Step 2A, Prong 1: Inherits the limitations and abstract ideas from Claim 60.
Step 2A, Prong 2: The claims do not recite additional elements that integrate the judicial exception into a practical application:
“wherein the AI system comprises one or more of an inference engine, a recommendation engine, a reinforcement learning agent, a neural network engine, a neural co-processor, a hardware accelerator, a graphics processing unit, or a general-purpose processor.” (An AI system comprising one or more of an inference engine, a recommendation engine, a reinforcement learning agent, a neural network engine, a neural co- processor, a hardware accelerator, a graphics processing unit, or a general-purpose processor generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).)
Step 2B: The claims do not recite additional elements that amount to significantly more than the judicial exception. An AI system comprising one or more of an inference engine, a recommendation engine, a reinforcement learning agent, a neural network engine, a neural co- processor, a hardware accelerator, a graphics processing unit, or a general-purpose processor generally links the use of the abstract idea to a particular technological environment or field of use – see MPEP 2106.05(h).
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.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or non-obviousness.
Claim(s) 52-55 and 59 are rejected under 35 U.S.C. 103 as being unpatentable over Sinn et al., (US Patent Application Number US12242613B2 filed on September 30, 2020, hereinafter “Sinn”), in view of Birnbaum et al., (US Patent Application Number US20200058382A1 filed on October 5, 2018, hereinafter “Birnbaum”).
With respect to Claim 52:
Sinn teaches:
“An apparatus of an artificial intelligence (AI) system, comprising:
AI engine circuitry to operate a machine learning (ML) model;” (Column 15, Lines 4-13 recite a machine learning component residing on a processer (AI engine circuitry) associated with operating the machine learning (ML) model.)
“memory circuitry to store program code of a set of self-assessment entities including a self-verification entity;” (Column 14, Lines 52-58 recite a processing unit that communicates with the memory, which is used to operate a machine learning models service that includes several components. Column 15, Lines 4-13 further recite how these set of components work together (including an adversarial operation evaluator) to evaluate and determine the level of robustness of a machine learning model against adversarial whitebox operations (a set of self-assessment entities including a self-verification entity).)
“and processor circuitry connected to the memory circuitry, wherein the processor circuitry is to operate the self-verification entity to: cause the AI engine circuitry to operate the ML model using a predefined test dataset,” (Column 14, Lines 52-58 recite a processing unit that communicates with the memory, which is used to operate a machine learning models service. Column 1, Lines 29-36 further recite that the embodiment may contain a machine learning model and a data set (predefined data set) used for testing the machine learning model.)
Sinn does not appear to explicitly disclose:
“and stop or pause the operation of the AI engine circuitry when biased predictions are generated by the AI engine circuitry based on the predefined test dataset.”
However, Birnbaum teaches:
“and stop or pause the operation of the AI engine circuitry when biased predictions are generated by the AI engine circuitry based on the predefined test dataset.” (Paragraph 0034 recites a framework that automatically removes an AI model from deployment based on a comparison, such as a model being deactivated. Paragraph 0035 further recites models with excessive bias may be automatically removed from use.)
It would have been obvious to a person having ordinary skill in the art (PHOSITA) before the effective filing date of the present application to combine the teachings of Sinn and the teachings of Birnbaum, which are both in the same field of invention. A PHOSITA would be motivated to combine the self-assessment architecture from Sinn with the bias-detection technique from Birnbaum in order to improve system reliability by utilizing Sinn’s existing architecture that checks for adversarial robustness issues and adding a bias-detection check that automatically halts operations before biased predictions cause future issues, meaning the combined teachings would be able to catch two defect types in advance instead of one.
With respect to Claim 53:
Sinn and Birnbaum combined teach:
“wherein the set of self-assessment entities includes a risk-related information (RRI) processing entity, and the processor circuitry is to operate the RRI processing entity to: (Column 14, Lines 53-58 from Sinn recites a machine learning models service containing several different components, including a reasoner component. Column 15, Lines 58-64 from Sinn further recite a reasoner component may control any component associated with the automated evaluation of the machine learning models service based on diagnostic information obtained from each of the components.)
process outputs generated by other ones of the self-assessment entities in a human-consumable format;” (Column 15, Lines 58-64 from Sinn recites a reasoner component may control any component associated with the automated evaluation of the machine learning models service based on diagnostic information obtained from each of the components (process outputs generated by other ones of the self-assessment entities). Column 15, Lines 14-19 from Sinn recite the reasoner component, along with the other components, may generate an evaluation summary (in a human-consumable format) based on evaluating and determining the level of robustness of the machine learning model.)
“and present the processed outputs to an authorized user.” (Paragraph 0070 from Birnbaum recites storing a report for access by a user of the deployed model (and present the processed outputs to an authorized user).)
With respect to Claim 54:
Sinn and Birnbaum combined teach:
“wherein the set of self-assessment entities includes a risk mitigation entity, and the processor circuitry is to operate the risk mitigation entity to: (Paragraph 0052 from Birnbaum recites a system that includes computing systems configured to receive information from different entities, which then process, store, and display information to other entities over the network. Paragraph 0061 from Birnbaum further recites that the system behavior and performance may be monitored against various metrics, including monitoring the sensitivity of the system and the efficiency of the system.)
determine trade-offs between risks of using the ML model versus functionality or efficiencies of operating the ML model.” (Paragraph 0061 from Birnbaum recites the sensitivity of the trained system may be monitored to determine whether the system is capturing substantially all of the individuals from a particular population that should be included in a particular cohort, and further, the efficiency of the system may be monitored to determine an achieved percentage reduction in the number of individuals required to proceed to an abstraction process.)
With respect to Claim 55:
Sinn and Birnbaum combined teach:
“wherein the set of self-assessment entities includes an AI system management entity, and the processor circuitry is to operate the AI system management entity to: (Column 15, Lines 58-64 from Sinn recite a reasoner component that controls (manages) an overall workflow of the system, such as any component associated with automated evaluations of machine learning models service based on diagnostic information obtained from each component.)
orchestrate internal processes of the AI system and orchestrate interactions between individual components of a set of components of the AI system or the ML model.” (Column 20, Lines 4-12 from Sinn recite the reasoner may include a variety of interactions with each component of the automated evaluation of machine learning models system, such as the reasoner component and adversarial operation evaluator component performing diagnostic adversarial operation evaluations in order to detect gradient masking, sub-optimal loss functions, and optimizer configurations.)
With respect to Claim 59:
Sinn and Birnbaum combined teach:
“wherein the AI engine circuitry comprises one or more of an inference engine, a recommendation engine, a reinforcement learning agent, a neural network engine, a neural co-processor, a hardware accelerator, a graphics processing unit, or a general-purpose processor.” (Column 11, Lines 32-37 from Sinn recite a general-purpose computing device, where the components of the computing system may include one or more processors or processing units.)
Claim(s) 56 are rejected under 35 U.S.C. 103 as being unpatentable over Sinn et al., (US Patent Application Number US12242613B2 filed on September 30, 2020, hereinafter “Sinn”), in view of Birnbaum et al., (US Patent Application Number US20200058382A1 filed on October 5, 2018, hereinafter “Birnbaum”), in further view of Shteingart et al., (US Patent Application Number US20240303146A1 filed on March 7, 2023, hereinafter “Shteingart”).
With respect to Claim 56:
Sinn and Birnbaum combined do not appear to explicitly disclose:
“wherein the set of self-assessment entities includes an AI system redundancy entity, and the processor circuitry is to operate the AI system redundancy entity to:
detect, during the operation of the AI system, a malfunctioning component of the set of components;”
“and replace the malfunctioning component with another component that fulfills a same or similar function as the malfunctioning component.”
However, Shteingart teaches:
“wherein the set of self-assessment entities includes an AI system redundancy entity, and the processor circuitry is to operate the AI system redundancy entity to: (Paragraph 0041 recites the environment monitoring server comprises a cluster monitoring and mitigation module, a virtual infrastructure monitoring and mitigation module, along with other connected modules. Paragraph 0048 recites virtual infrastructure monitor deploys a storage component that requires replacement due to malfunctioning.)
detect, during the operation of the AI system, a malfunctioning component of the set of components;” (Paragraph 0004 recites a monitoring entity may detect the malfunction associated with malfunctioning components in the cluster computing environment.)
“and replace the malfunctioning component with another component that fulfills a same or similar function as the malfunctioning component.” (Paragraph 0048 recites virtual infrastructure monitor deploys a storage component that requires replacement due to malfunctioning. Paragraph 0049 further recites the cluster monitor then adds a new replacement storage component to the storage cluster and validates the operation of the replacement storage component through role synchronization.)
It would have been obvious to a person having ordinary skills in the art (PHOSITA) before the effective filing date of the present application to combine the teachings of Sinn and Birnbaum with the teachings of Shteingart. A PHOSITA would be motivated to combine the self-assessment and automated bias-monitoring techniques from Sinn and Birnbaum with the detection and replacement malfunctioning component technique from Shteingart in order to improve the availability and reliability of Sinn and Birnbaum’s combined framework and provide a recovery mechanism.
Claim(s) 57-58 are rejected under 35 U.S.C. 103 as being unpatentable over Sinn et al., (US Patent Application Number US12242613B2 filed on September 30, 2020, hereinafter “Sinn”), in view of Birnbaum et al., (US Patent Application Number US20200058382A1 filed on October 5, 2018, hereinafter “Birnbaum”), in further view of Lee et al., (US Patent Application Number US20200265356A1 filed on February 14, 2020, hereinafter “Lee”).
With respect to Claim 57:
Sinn and Birnbaum combined teach:
“provide information about potential biases in predictions generated by the ML model to an authorized user via a user interface;” (Paragraph 0070 from Birnbaum recites that when a comparison results in a difference between the selected subset and the second plurality of individuals greater than a second threshold, the processing device may transmit an alert about the possible presence of bias (provide information about potential biases in predictions generated by the ML model) to a user of the deployed ML model (authorized user).)
Sinn and Birnbaum combined do not appear to explicitly disclose:
“wherein the set of self-assessment entities includes a human oversight entity, and the processor circuitry is to operate the human oversight entity to:”
“provide information about potential biases in predictions generated by the ML model to an authorized user via a user interface;”
“receive a selected action based on the provided information;”
“and issue the action to one or more components of the set of components to be executed by the one or more components.”
However, Lee teaches:
“provide information about potential biases in predictions generated by the ML model to an authorized user via a user interface;” (Paragraph 0038 recites probabilistic assessment information is presented in a dashboard user interface.)
“wherein the set of self-assessment entities includes a human oversight entity, and the processor circuitry is to operate the human oversight entity to: (Paragraph 0192 recites the AI platform may include an event bus that is configured to monitor and evaluate events pertaining to algorithms executing on one or more servers, where users of the AI platform are presented with various user interfaces (e.g., on client devices of the users), such as dashboard and console user interfaces, which present results of execution of the models as well as governance information pertaining to the execution of the models.)
“receive a selected action based on the provided information;” (Paragraph 0189 recites users can investigate further with fully explainable rationales generated (e.g., by the 'Automated Algo and Data governance' solution of the AI platform), then confirm whether the risk rating from Algo-AI was correct, akin to receiving a selected action based on provided information.)
“and issue the action to one or more components of the set of components to be executed by the one or more components.” (Paragraph 0189 recites the user’s follow up action will be used to train the “Algo-AI” algorithm.)
It would have been obvious to a person having ordinary skills in the art (PHOSITA) before the effective filing date of the present application to combine the teachings of Sinn and Birnbaum with the teachings of Lee. A PHOSITA would be motivated to combine the self-verification and automated bias-monitoring techniques from Sinn and Birnbaum with the user interactive human oversight dashboard component from Lee in order to improve the system’s reliability by ensuring a user is alerted to specific biases or defects that need a decision through a graphical user interface.
With respect to Claim 58:
Sinn and Birnbaum combined do not appear to explicitly disclose:
“wherein the set of self-assessment entities includes a record keeping entity, and the processor circuitry is to operate the record keeping entity to:
track interactions with the AI system, wherein the interactions include one or more of user activity, behavior of the AI system or the ML model, and information on training or testing the ML model;”
“and logging the tracked interactions in one or more records”
However, Lee teaches:
“wherein the set of self-assessment entities includes a record keeping entity, and the processor circuitry is to operate the record keeping entity to: (Paragraph 0064 recites an IndeX module is configured to monitor the state of an algorithm as it runs. For example, this module may capture actual values of all variables being used or generated during the code's run.)
track interactions with the AI system, wherein the interactions include one or more of user activity, behavior of the AI system or the ML model, and information on training or testing the ML model;” (Paragraph 0064 recites the module also records code identification, date/time, state of code (if any), and the hash of the dataset used.)
“and logging the tracked interactions in one or more records.” (Paragraph 0099 recites index records may each have a unique identifier (e.g., a date and time stamp) and be stored in a record block in chronological order, where the records may form an immutable history log represented as a blockchain.)
It would have been obvious to a person having ordinary skills in the art (PHOSITA) before the effective filing date of the present application to combine the teachings of Sinn and Birnbaum with the teachings of Lee. A PHOSITA would be motivated to combine the self-verification and automated bias-monitoring techniques from Sinn and Birnbaum with the AI system monitoring technique from Lee in order to improve the auditability of Sinn and Birnbaum’s processes, providing historical information that can be referred to at a later time.
Claim(s) 60-61, 63-68, and 71 are rejected under 35 U.S.C. 103 as being unpatentable over Lee et al., (US Patent Application Number US20200265356A1 filed on February 14, 2020, hereinafter “Lee”), in view of Birnbaum et al., (US Patent Application Number US20200058382A1 filed on October 5, 2018, hereinafter “Birnbaum”), in further view of Sinn et al., (US Patent Application Number US12242613B2 filed on September 30, 2020, hereinafter “Sinn”).
With respect to Claim 60:
Lee teaches:
“A non-transitory computer readable medium (NTCRM) comprising instructions for operating a set of artificial intelligence (AI) system monitoring, evaluation, and reporting (AIMER) functions including an AI risk management system function (AIRMS), wherein execution of the instructions by one or more processors is to cause a compute node to operate the AIRMS to:
monitor outputs generated by an AI system implemented by the compute node;” (Paragraph 0080 recites AI Cronus monitors and records its behavior, state, messages, and so on from any machines that are executing a given algorithm, where the system cross references runtime data with expected information that was captured during development time extraction.)
Lee does not appear to explicitly disclose:
“and issuing one or more corrective actions to the AI system when the monitored outputs include potential biases,
wherein the one or more corrective actions include adjusting one or more parameters to reduce or eliminate the potential biases”
However, Birnbaum teaches:
“and issuing one or more corrective actions to the AI system when the monitored outputs include potential biases,” (Paragraph 0033 recites based on a comparison, an alert may be generated for a user of the model alerting the user to the possibility of bias. Paragraph 0034 further clarifies the framework may automatically remove the model from deployment based on the comparison and detection of bias.)
It would have been obvious to a person having ordinary skills in the art (PHOSITA) before the effective filing date of the present application to combine the teachings of Lee with the teachings of Birnbaum, which are both in the same field of invention. A PHOSITA would be motivated to combine the bias detection mechanisms from Birnbaum with the AI monitoring system of Lee in order trigger a corrective action once a bias is identified through the monitoring system, which improves the systems awareness and allows issues to be identified and resolved appropriately.
Lee and Birnbaum do not appear to explicitly disclose:
“wherein the one or more corrective actions include adjusting one or more parameters to reduce or eliminate the potential biases”
However, Sinn teaches:
wherein the one or more corrective actions include adjusting one or more parameters to reduce or eliminate the potential biases” (Column 15, Lines 35-41 recite the optimization component may adjust one or more adversarial operation hyperparameters targeting the loss function towards optimal performance with respect to the adversarial operation objective. Column 19, Lines 11-24 recite the loss function composer may attempt to improve the consistency and convergence of the loss function in order to improve convergence and adversarial operation performance.)
It would have been obvious to a person having ordinary skills in the art (PHOSITA) before the effective filing date of the present application to combine the teachings of Lee and Birnbaum with the teachings of Sinn, which are all in the same field of invention. A PHOSITA would be motivated to combine the bias detection and removal framework from Lee and Birnbaum with the parameter adjustment mechanism from Sinn in order to improve the model behavior correction process when bias is detected and allow it to stay functioning through parameter changes instead of full removal of the model.
With respect to Claim 61:
Lee, Birnbaum, and Sinn combined teach:
“wherein execution of the instructions is to cause the compute node to operate the AIRMS to:
monitor inputs provided to the AI system;” (Paragraph 0080 from Lee recites AI Cronus monitors and records its behavior, state, messages, and so on from any machines that are executing a given algorithm, where the system cross references runtime data with expected information that was captured during development time extraction.)
“and issuing one or more other corrective actions to the AI system when the monitored inputs include potential errors, (Paragraph 0086 from Lee recites after an anomaly (error) is identified in a dataset, appropriate correction by the system to the dataset is implemented to correct the anomaly.)
Lee and Birnbaum do not appear to explicitly disclose:
wherein the one or more other corrective actions include adjusting one or more parameters of the inputs to correct the potential errors.”
However, Sinn teaches:
wherein the one or more other corrective actions include adjusting one or more parameters of the inputs to correct the potential errors.” (Column 8, Lines 47-56 recite a reparameterization process, where inputs are adjusted through a parameterized function to correct a problem (error) with the input, akin to a corrective action.)
With respect to Claim 63:
Lee, Birnbaum, and Sinn combined teach:
“wherein the set of AIMER functions includes an entity for record keeping (ERK), and execution of the instructions is to cause the compute node to operate the ERK to: (Paragraph 0064 from Lee recites the IndeX module is configured to monitor the state of an algorithm as it runs.)
obtain inputs to the AI system, outputs generated by the AI system, and internal states corresponding to the inputs or the outputs;” (Paragraph 0080 from Lee recites the system monitors and records its behavior, state (internal states), messages, and so on from any machines that are executing a given algorithm, where at development time, the platform extracts meta information about the AI, including the AI's input, output, query, available states of AI, etc.)
“process the inputs, the outputs, and the internal states to generate statistics or metrics related to the inputs, the outputs, and the internal states;” (Paragraph 0099 from Lee recites the result or output of the algorithm (e.g., executable 806) is analyzed to determine values of key constraints used by the algorithm to determine the result. Paragraph 0075 further clarifies key constraints and metrics, including qualitative constraints, are generated by AI and analytics processes that are relevant to particular business objectives.)
“and store the inputs, the outputs, the internal states, and the statistics or metrics in a local or remote database.” (Paragraph 0099 from Lee recites gathered data, including captured inputs, outputs, values of key constraints, or metadata pertaining to execution of the algorithm are recorded in an index record that each have a unique identifier and are stored in a record block in chronological order. Paragraph 0046 further clarifies database servers and data stores are used.)
With respect to Claim 64:
Lee, Birnbaum, and Sinn combined teach:
“wherein the set of AIMER functions includes an entity for transparency and information (ETI), and execution of the instructions is to cause the compute node to operate the ETI to: (Paragraph 0051 from Lee recites the platform ensuring essential integrity, performance, and accountability of AI tasks, reducing or virtually eliminating the black box.)
generate transparency data including one or more of capability information of the AI system, maintenance and care information related to the AI system, self-assessment information related to the AI system, and historic data related to the operation of the AI system;” (Paragraph 0080 from Lee recites playback of an entire state or scenario handled by an algorithm (e.g., including incoming messages, outgoing messages, data used, variables used, and so on), akin to historic data, is another feature provided by the system.)
“and generate user interface data to present the transparency data, wherein the user interface data includes one or more of text data, image data, audio data, and video data;” (Paragraph 0075 from Lee recites a dashboard (user interface) including relationship visualizations and drill-down capabilities for key constraints and other data associated with business objectives accessible from various front-end tools deployed within an organization.)
“and send the user interface data to an authorized user.” (Paragraph 0075 from Lee recites the dashboard may include user interfaces and user interface elements to allow a user to visually associate a business objective with an AI output aggregator. Paragraph 0076 from Lee further recites the dashboard view displays a bank's existing processes, including the roles of relevant staff members and the layers of critical functions, report documents, references, and datasets. Under broadest reasonable interpretation, specific user roles exist and are provided access to the dashboard.)
With respect to Claim 65:
Lee, Birnbaum, and Sinn combined teach:
“wherein the set of AIMER functions includes an entity for AI output self-verification (EAIOSV), and execution of the instructions is to cause the compute node to operate the EAIOSV to: (Paragraph 0070 from Lee recites an entity that lives within an AI algorithm, where it monitors AI execution.)
perform a self-verification process on a prediction generated by the AI system before the prediction is provided to an external entity, (Paragraph 0079 from Lee recites a baseline output is known (e.g., based on human-defined desirable boundaries), where the AI periodically (e.g., at frequent intervals) communicates with other AI agents (or one or more other deterministic algorithms that can be benchmarked) to monitor AI execution.)
wherein the self-verification process includes: comparison of the generated prediction with one or more historical predictions;” (Paragraph 0068 from Birnbaum recites measuring a model’s current output against a previously-established reference, where a selected subset and a second plurality of individuals along the at least one characteristic is compared.)
“and determination of biases in the generated prediction based on a divergence of the generated prediction from the one or more historical predictions.” (Paragraph 0070 from Birnbaum recites when the comparison results in a difference between the selected subset and the second plurality of individuals greater than a second threshold, an alert is sent to a user of the deployed model. Paragraph 0069 further clarifies if the selected subset has a particular survival probability over time and the second plurality of individuals has a particular survival probability over time that is within a range defined by the particular survival probability over time of the selected subset, the difference may be determined as zero. Otherwise, the distance between the survival probabilities over time may be determined and averaged over time to compute the difference.)
With respect to Claim 66:
Lee, Birnbaum, and Sinn combined teach:
“wherein the self-verification process includes: operation of an alternation function to change
one or more parameters of the AI system;” (Column 15, Lines 35-41 from Sinn recites the optimization component may adjust, modify, and/or tune one or more adversarial operation hyperparameters targeting the loss function towards optimal performance with respect to the adversarial operation objective.)
“obtaining a prediction from the AI system with the changed one or more parameters;” (Column 17, Lines 61-65 from Sinn recite the outputs may include, for example, a modified targeted model (e.g., a modified targeted model), a loss function, an optimizer, one or more adversarial examples, and an evaluation summary.)
“comparison of the generated prediction with obtained prediction;” (Column 7, Lines 16-18 from Sinn recites a determination should be made relating to gradient masking, e.g., whether the gradients used in the adversarial operations are zero, infinite or not a number (NaN). Column 7, Lines 5-8 further recited if the gradient-free adversarial operations achieve significantly higher success rates than the gradient-based ones (comparison), this indicates that an adaptive adversarial operation needs to be devised.)
and determination of biases in the generated prediction based on a divergence of the generated prediction from the obtained prediction.” (Paragraph 0069 from Birnbaum recites if the selected subset has a particular survival probability over time and the second plurality of individuals has a particular survival probability over time that is within a range defined by the particular survival probability over time of the selected subset, the difference may be determined as zero. Otherwise, the distance between the survival probabilities over time may be determined and averaged over time to compute the difference.)
With respect to Claim 67:
Lee, Birnbaum, and Sinn combined teach:
“wherein the set of AIMER functions includes an accuracy verification entity (AVE), and execution of the instructions is to cause the compute node to operate the AVE to: (Column 1, Lines 33-37 from Sinn recite a level of robustness of a machine learning model against adversarial whitebox operations may be evaluated and determined by applying a data set used for testing the machine learning model, one or more adversarial operation objectives, an adversarial threat model, and a selected number of hyperparameters, akin to an accuracy verification entity or evaluation components using a test dataset.)
place the AI system in a testing state;” (Column 19, Lines 37-44 from Sinn recite the diagnostic adversarial operation evaluations (testing state) may include a suite of gradient-based, gradient-free and transfer adversarial operations in order to diagnose the machine learning model.)
“provide a test dataset to the AI system in the testing state;” (Column 21, Section 53-55 from Sinn recite a level of robustness of a machine learning model against adversarial whitebox operations may be evaluated and determined by applying a data set used for testing the machine learning model.)
“compare outputs generated by the AI system in the testing state with known outputs for the test dataset;” (Column 3, Lines 62-67 and Column 4, Lines 1-4 from Sinn recite comparing the classifier’s generated output against the known, ground-truth label for the test input.)
“and calculate an accuracy metric for the AI system in the testing state based on a number of correct predictions in the generated outputs or a number of incorrect predictions in the generated outputs.” (Column 7, Lines 5-8 from Sinn recite if the gradient-free adversarial operations achieve significantly higher success rates than the gradient-based ones, this indicates that an adaptive adversarial operation needs to be devised. Column 17, Lines 53-55 from Sinn recite the adversarial operation evaluator may apply end-to-end adversarial operations for diagnostic and final evaluation purposes. Column 19, Lines 64-67 and Column 20, Lines 1-3 from Sinn recite the percentage of test data points (accuracy metric) for which any of the performed adversarial operations (either a diagnostic or final one) achieved the adversarial operation objective is calculated.)
With respect to Claim 68:
Lee, Birnbaum, and Sinn combined teach:
“wherein the set of AIMER functions includes a robustness verification entity (RVE), and execution of the instructions is to cause the compute node to operate the RVE to: (Column 1, Lines 33-37 from Sinn recite a level of robustness of a machine learning model against adversarial whitebox operations may be evaluated and determined by applying a data set used for testing the machine learning model, one or more adversarial operation objectives, an adversarial threat model, and a selected number of hyperparameters, akin to an accuracy verification entity or evaluation components using a test dataset.)
modify the test dataset to include one or more erroneous data items;” (Column 3, Lines 63-67 and Column 4, Lines 1-10 from Sinn recite a modified input is created from an original input to specifically cause an error (include one or more erroneous data items.)
“compare outputs generated by the AI system in the testing state with known outputs for the test dataset;” (Column 3, Lines 65-67 from Sinn recite comparing the classifier’s actual output against the known/ground-truth label for the test data.)
“and calculate robustness metric for the AI system in the testing state based on the number of correct predictions or the number of incorrect, wherein the number of correct predictions includes one or more correctly identified errors based on the one or more erroneous data items and the number of incorrect predictions includes one or more unidentified errors based on the one or more erroneous data items.” (Column 7, Lines 5-8 from Sinn recite if the gradient-free adversarial operations achieve significantly higher success rates than the gradient-based ones, this indicates that an adaptive adversarial operation needs to be devised. Column 17, Lines 53-55 from Sinn recite the adversarial operation evaluator may apply end-to-end adversarial operations for diagnostic and final evaluation purposes. Column 19, Lines 64-67 and Column 20, Lines 1-3 from Sinn recite the calculation of a metric for which any of the performed adversarial operations (either a diagnostic or final one) achieved the adversarial operation objective, akin to comparing success and failure rates of adversarial operations against a model.)
With respect to Claim 71:
Lee, Birnbaum, and Sinn combined teach:
“wherein the AI system comprises one or more of an inference engine, a recommendation engine, a reinforcement learning agent, a neural network engine, a neural co-processor, a hardware accelerator, a graphics processing unit, or a general-purpose processor.” (Column 11, Lines 32-37 from Sinn recite a general-purpose computing device, where the components of the computing system may include one or more processors or processing units.)
Claim(s) 62 are rejected under 35 U.S.C. 103 as being unpatentable over Lee et al., (US Patent Application Number US20200265356A1 filed on February 14, 2020, hereinafter “Lee”), in view of Birnbaum et al., (US Patent Application Number US20200058382A1 filed on October 5, 2018, hereinafter “Birnbaum”), in view of Sinn et al., (US Patent Application Number US12242613B2 filed on September 30, 2020, hereinafter “Sinn”), in further view of Siravara et al., (US Patent Application Number US11481671B2 filed on May 16, 2019).
With respect to Claim 62:
Lee, Birnbaum, and Sinn combined teach:
“wherein the set of AIMER functions includes a data verification component (DVC), and execution of the instructions is to cause the compute node to operate the DVC to: (Paragraph 0062 from Lee recites DataX converts an organization's data dictionary into an easily searchable and navigable form, where it also takes information from the CodeX module to automatically trace data lineage, track sources, measure usage, and compare the latter against the organization's data governance policy.)
“and tag the input dataset with a digital certificate when the input dataset is properly validated, (Paragraph 0130 from Lee recites the block data structure includes a field for data integrity (e.g., a hash value of the data set). Under the broadest reasonable interpretation, this is akin to tagging an input data set with a digital certificate.)
Lee, Birnbaum, and Sinn combined do not appear to explicitly disclose:
“validate an input dataset before the input dataset is provided to the AI system;”
“and tag the input dataset with a digital certificate when the input dataset is properly validated,
“and wherein the AI system is to verify the input dataset using the digital certificate, and generate a prediction using the input dataset when the input dataset is properly verified.”
However, Siravara teaches:
“validate an input dataset before the input dataset is provided to the AI system;” (Column 4, Lines 40-43 recites calculating a file integrity value of the file, which involves calculating a hash value, and using a file integrity detection function to determine whether the file integrity value corresponds to a reference file integrity value of the file.)
“and tag the input dataset with a digital certificate when the input dataset is properly validated, (Column 4, Lines 40-43 recites calculating a file integrity value of the file, which involves calculating a hash value, and using a file integrity detection function to determine whether the file integrity value corresponds to a reference file integrity value of the file.)
and wherein the AI system is to verify the input dataset using the digital certificate, and generate a prediction using the input dataset when the input dataset is properly verified.” (Column 3, Lines 41-45 recite determining whether the file integrity value corresponds to a reference file integrity value of the file and performing an operation with the machine learning model based on determining that the file integrity value corresponds to the reference file integrity value of the file.)
It would have been obvious to a person having ordinary skills in the art (PHOSITA) before the effective filing date of the present application to combine the teachings of Lee, Birnbaum, and Sinn with the teachings of Siravara, which are all in the same field of invention. A PHOSITA would be motivated to combine the monitoring and AI behavior correction system from Lee, Birnbaum, and Sinn with the data validation and verification technique from Siravara in order to decrease the system’s vulnerability to corrupt data. This In turn would ensure the AI system only generates predictions from data that has been verified, improving the system’s availability.
Claim(s) 69-70 are rejected under 35 U.S.C. 103 as being unpatentable over Lee et al., (US Patent Application Number US20200265356A1 filed on February 14, 2020, hereinafter “Lee”), in view of Birnbaum et al., (US Patent Application Number US20200058382A1 filed on October 5, 2018, hereinafter “Birnbaum”), in view of Sinn et al., (US Patent Application Number US12242613B2 filed on September 30, 2020, hereinafter “Sinn”), in further view of Gottschlich et al., (US Patent Application Number US20190319977A1 filed on June 27, 2019).
With respect to Claim 69:
Lee, Birnbaum, and Sinn do not appear to explicitly disclose:
“wherein the set of AIMER functions includes cryptographic engine (CE), and execution of the instructions is to cause the compute node to operate the CE to:
generate a fingerprint for the AI system based on one or more inputs to the AI system, outputs generated by the AI system, and one or more internal system states of the AI system;”
“and encrypt data to be conveyed between the AI system and the set of AIMER functions or communicated to external devices.”
However, Gottschlich teaches:
“wherein the set of AIMER functions includes cryptographic engine (CE), and execution of the instructions is to cause the compute node to operate the CE to: (Paragraph 0029 recites the example system includes a communication interface, memory, and a fingerprint processor. The fingerprint processor includes a telemetry collector, a trace processor, a fingerprint extractor, a fingerprint analyzer, a fingerprint clusterer, and a fingerprint classifier.)
generate a fingerprint for the AI system based on one or more inputs to the AI system, outputs generated by the AI system, and one or more internal system states of the AI system;” (Paragraph 0111 recites a fingerprint extractor generates a first fingerprint from the captured application execution behavior and performance monitor information.)
“and encrypt data to be conveyed between the AI system and the set of AIMER functions or communicated to external devices.” (Paragraph 0060 recites the fingerprint classification vector is provided with other vectors as cloud telemetry information to the telemetry analyzer in a secure cloud.. It is understood that a secure cloud environment involves protected and encrypted transmissions of data.)
It would have been obvious to a person having ordinary skills in the art (PHOSITA) before the effective filing date of the present application to combine the teachings of Lee, Birnbaum, and Sinn with the teachings of Gottschlich, which are all in the same field of invention. A PHOSITA would be motivated to combine the bias detection and model parameter adjustment framework from Lee, Birnbaum, and Sinn with the fingerprint generation and data encryption technique from Gottschlich in order to allow the system to catch tampering through the use of fingerprints and improve the security of the system’s data when transmitted.
With respect to Claim 70:
Lee, Birnbaum, Sinn, and Gottschlich combined teach:
“wherein the set of AIMER functions includes an AI system quality manager (AISQM), and execution of the instructions is to cause the compute node to operate the AISQM to: (Column 15, Lines 58-64 from Sinn recite a reasoner component may control an overall workflow of the system such as, for example, any component associated with the automated evaluation of machine learning models service, based on diagnostic information obtained from each of the components of the automated evaluation of machine learning models service.)
calculate a quality metric for the AI system based on the fingerprint, the accuracy metric, and the robustness metric;” (Column 15, Lines 14-18 from Sinn recite the reasoner component, in association with the machine learning component, the adversarial operation evaluator, and the optimization component, may generate an evaluation summary based on evaluating and determining of the level of robustness of the machine learning model. (Paragraph 0111 from Gottschlich recites a fingerprint extractor generates a first fingerprint from the captured application execution behavior and performance monitor information.)
“and issue one or more remedial actions to the AI system when the quality metric is below a threshold.” (Paragraph 0070 from Birnbaum recites when the comparison results in a difference between the selected subset and the second plurality of individuals greater than the threshold, automatically remove the model from deployment, akin to issuing one or more remedial actions to the system when the quality metric is below a threshold.)
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Vibha Bhat whose telephone number is (571)-272-7091. The examiner can normally be reached on Monday – Thursday from 8:00 AM to 5:00 PM EST and every other Friday from 8:00 AM to 4:00 PM EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. See MPEP § 713.01. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at https://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Mariela Reyes, can be reached at telephone number (571)-270-1006. The fax phone number for the organization where this application or proceeding is assigned is (571)-273-8300. Information regarding the status of an application 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://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 (572)-272-1000.
/Vibha Bhat/Examiner
Art Unit 2142
/HAIMEI JIANG/Primary Examiner, Art Unit 2142
Prosecution Insights