DETAILED ACTION
This action is in response to Applicant’s Request for Continued Examination ("Response”) received on February 26, 2026 in response to the Office Action dated January 27, 2026. This action is made Non-Final.
Claims 1-18 are pending.
Claims 1, 17, and 18 are independent claims.
Claims 1-18 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 .
Applicant’s Response
In Applicant’s Response, Applicant amended claims 1, 17, and 18, and submitted arguments against the prior art in the Office Action dated January 27, 2026.
Claim Rejections - 35 USC § 103
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.
Claim(s) 1, 5-9, 11-14, 17, and 18 is/are rejected under 35 U.S.C. 103 as being unpatentable over Verma et al., US Publication 2021/0174163 (“Verma”), in view of Hacmon et al., US Publication 2022/0076080 (“Hacmon”), in view of Donovan, US Publication 2023/0114826 (“Donovan”), and further in view of Linardatos, Pantelis, Vasilis Papastefanopoulos, and Sotiris Kotsiantis. "Explainable ai: A review of machine learning interpretability methods." Entropy 23.1 (2020): 18 (“Linardatos”).
Claim 1:
Verma teaches or suggests a computer-implemented method to perform … artificial intelligence (AI) control, the method comprising:
receiving, by a computing system comprising one or more computing devices, a user request for output data from a user (see Fig. 1, 2, 6, 7; para. 0005 - cache decision maker, coupled to the client interface, configured to analyze the client request, and, based at least in part on the analysis, select either the first AI model or the second AI model to respond to the request; para. 0014 - selection of an appropriate AI model, out of two or more possible models, to respond to a client request; para. 0021 - client may be, for example, a human user, or, for example, in what is a more common scenario, a computer program that accesses an AI-enhanced web solution; para. 0036 - cache decision maker 215 is trained to recognize the types of inputs (e.g., user requests).);
determining, by the computing system, context data descriptive of a context associated with the user request ... associated with the user request (see para. 0020 - cache decision maker decides, on a per request basis, whether it is better to use the simpler model at the edge, or to use the complex model from the cloud; para. 0023 - classifier model 121 of input analyzer 120 is trained to recognize the types of inputs where the simple model is suitable, and those for which it is not; para. 0029 - accessed by a client application to submit requests, e.g., images of potatoes on a conveyor belt from a potato greening inspection application, or images acquired by a UAV of a bridge surface from a department of highways inspection application; para. 0030 - input analyzer component 240 receives information from the client interface component; para. 0037 - a range of input images. the images are aligned in a specific position; para. which accent is used in a specific speech sample, and depending on the specific accent used, it can direct a speech to text conversion request to one of the specialized. Para. 0040 – requests originating from a location.);
executing, by the computing system, a … AI controller to select one or more selected AI systems from a plurality of available AI systems, each of the plurality of AI systems comprising one or more AI components, wherein the one or more selected AI systems are selected with the ... AI controller (see para. 0005 - cache decision maker, coupled to the client interface, configured to analyze the client request, and, based at least in part on the analysis, select either the first AI model or the second AI model to respond to the request; para. 0014 - selection of an appropriate AI model, out of two or more possible models, to respond to a client request; para. 0038 - cache decision maker 215 checks incoming requests and decides whether to use the simple AI model 215 at the edge device 220 (known as a "cache hit"), or the more complex AI model 221 in the cloud 230. both the simple AI model 221 and the complex AI model 231 are in the same location, such as, for example, in an AI enabled load-balancer.);
generating, by the computing system, a respective output from each of the one or more selected AI systems based on the user request (see Fig. 1, 2, 6, 7; para. 0014 - selection of an appropriate AI model, out of two or more possible models, to respond to a client request; para. 0021 - client may be, for example, a human user, or, for example, in what is a more common scenario, a computer program that accesses an AI-enhanced web solution; para. 0037 - AI models which are used to examine images to check if a manufactured product is defective or good may be considered; para. 0047 - where a response is provided to the client request from the first Al model; para. 0048 - where a response to the client request is provided from the second AI model; para. 0050 - a request to detect any defects; para. 0053 - where a response is provided to the client from the selected simple Al model.); and
providing, by the computing system, the output data to the user based on the respective output from one or more of the one or more selected AI systems (see Fig. 1, 2, 6, 7; para. 0014 - selection of an appropriate AI model, out of two or more possible models, to respond to a client request; para. 0016 - device may be used to receive images and make decisions based on the image, using various AI models trained to detect a problem or one
or more defects; para. 0017 - check parts after manufacture; para. 0018 - inspect bridges for cracks; para. 0021 - client may be, for example, a human user, or, for example, in what is a more common scenario, a computer program that accesses an AI-enhanced web solution; para. 0037 - AI models which are used to examine images to check if a manufactured product is defective or good may be considered; para. 0047 - where a response is provided to the client request from the first Al model; para. 0048 - where a response to the client request is provided from the second AI model; para. 0050 - a request to detect any defects; para. 0053 - where a response is provided to the client from the selected simple Al model.).
Though Verma teaches selecting an AI model based on whether an AI model is appropriate, suitable, or qualified, Hacmon more specifically teaches or suggests … responsible artificial intelligence … responsible AI … responsible AI (see Fig. 1; para. 0002 - to evaluate the fairness and robustness of any AI-based ML models of systems and applications, several steps are required: it is required to detect and measure various properties of ethical bias from different ethical points of view. Then, it is required to aggregate those different ethical perspectives to one final fairness score. This final fairness score provides the data scientists with quantitative estimation, i.e., assessment for the fairness of the examined ML model and can assist them in evaluations and comparisons of different models; para. 0007 -over twenty-five fairness measurements in the literature, each examines the ML model from a different ethical point of view; para. 0008 - detecting an underlying bias and the fairness level of an ML model, which can be integrated into Continuous Integration/Continuous Delivery processes; para. 0015 - final fairness score aggregation module for aggregating the executed tests results into a final fairness score of the examined model and dataset; para. 0038 - system (in the form of a plugin, for example) can be integrated into a larger system, for examining the fairness and robustness of ML models which try to fulfill various AI-based tasks.).
Accordingly, it would have been obvious to one having ordinary skill before the effective filing date of the claimed invention to modify the system and method, taught in Verma, to include ... responsible artificial intelligence … responsible AI … responsible AI for the purpose of efficiently evaluating AI models based on ethical considerations to produce a model score, enabling model comparisons, as taught by Hacmon (0002, 0007, 0015, and 0038).
Donovan further teaches or suggests wherein the context comprises societal context ...; ... based on a responsibility objective associated with the societal context (see para. 0127 - virtue scoring model 322 is trained, for example, via training data 326 to generate a virtue score 324 corresponding to a particular virtue in response to content data 320 such as analysis data 314, AI output data of an underlying AI process to be analyzed; para. 0129 - A responsibility scoring model trained to generate a virtue score 324 indicating a responsibility score or other metric that indicates, for example, how well underlying AI or other content is addressing legal and/or ethical principles; para. 0130 - equitability scoring model trained to generate a virtue score 324 indicating an equitability score or other metric, that indicates, for example, an amount (or lack of) bias in the underlying AI or other content data; para. 0134 - user customized virtues, trained for example, based on results from user defined surveys to generate other virtue scores 324 that are different than those listed above and address a particular user problem or concern.).
Accordingly, it would have been obvious to one having ordinary skill before the effective filing date of the claimed invention to modify the system and method, taught in Verma, to include wherein the context comprises societal context ...; ... based on a responsibility objective associated with the societal context for the purpose of efficiently evaluating AI models based on ethical, legal, and particular users concerns to produce a model score, improving underlying model analysis and enabling model comparisons, as taught by Donovan (0002, 0007, 0015, and 0038).
Linardatos further teaches or suggests wherein the one or more selected AI systems are trained to handle sensitive requests for the societal context, wherein the plurality of available AI systems comprises different AI systems trained for different societal contexts (see 3.3 - machine learning systems are increasingly adopted in real life applications, any inequities or discrimination that are promoted by those systems have the potential to directly affect human lives. Machine Learning Fairness is a sub-domain of machine learning interpretability that focuses solely on the social and ethical impact of machine learning algorithms by evaluating them in terms impartiality and discrimination. fairness of a machine learning system has been evaluated by checking the models’ predictions and errors across certain demographic segments, for example, groups of a specific ethnicity or gender. techniques have been developed both to remove bias from training data and from model predictions and to train models that learn to make fair predictions in the first place. biases regarding ethnicity, gender, disability status, marital status, or any other demographic. While straightforward and efficient when it comes to selecting the most fair model. to diminish the impact of the sensitive features in the model’s decisions, the sensitive features along with their most correlated features are removed before training. This forces the model to learn from and, therefore, base its decisions on other attributes, thus not being biased. design of suitable models and the development of a sound theoretical background that would explicitly address critical issues, such as social stratification and unequal access, in order for true fairness to be achieved. regulate the conditional probability distribution of the sensitive feature values given that the label is positive, by simply boosting the probability of the disadvantaged sensitive feature values given the positive label, while, at the same time, decreasing the probability of the favoured sensitive feature values given the positive label. training a different model for every sensitive attribute value. The case where a sensitive feature has two values, and, therefore, two models were trained, was illustrated: one model was developed using only the rows that had a favoured sensitive value, while another model only utilised the rows that had a discriminated sensitive value. The different models are part of a meta-model, where discrimination is mitigated by adjusting the conditional probability.).
Accordingly, it would have been obvious to one having ordinary skill in the art before the effective filing date of the claimed invention to combine the teachings of Linardatos with those of Verma. One would have been motivated to do so for the purpose of efficiently developing models specific to sensitive issues that may then be used in a manner that increases fairness and reduces bias or discrimination, as taught by Linardatos (3.3).
Claim(s) 17 and 18:
Claim(s) 17 and 18 correspond to Claim 1, and thus, Verma, Hacmon, Donovan, and Linardatos teach or suggest the limitations of claim(s) 17 and 18 as well.
Claim 5:
Verma further teaches or suggests wherein the context data comprises user location data and/or user demographic data associated with the user (see para. 0020 - cache decision maker decides, on a per request basis, whether it is better to use the simpler model at the edge, or to use the complex model from the cloud; para. 0023 - classifier model 121 of input analyzer 120 is trained to recognize the types of inputs where the simple model is suitable, and those for which it is not; para. 0029 - accessed by a client application to submit requests, e.g., images of potatoes on a conveyor belt from a potato greening inspection application, or images acquired by a UAV of a bridge surface from a department of highways inspection application; para. 0030 - input analyzer component 240 receives information from the client interface component; para. 0037 - a range of input images. the images are aligned in a specific position; para. 0039 - which accent is used in a specific speech sample, and depending on the specific accent used, it can direct a speech to text conversion request to one of the specialized. accent-specific simple models; Para. 0040 – requests originating from a location.).
Claim 6:
Verma further teaches or suggests wherein the … AI controller comprises a machine-learned model (see para. 0023 - Input analyzer 120 performs this task using classifier model 121, which is a third AI model that is trained to determine, for a given client request, whether a simpler local model would provide results compatible with the counterpart complex model stored in the cloud. As described in detail below, classifier model 121 of input analyzer 120 is trained to recognize the types of inputs where the simple model is suitable, and those for which it is not; para. 0027 - which are trained to, and then used to, make inference decisions in response to client requests. a training data generator component 245 of cache decision maker application; para. 0030 - the third AI model used by the input analyzer component 240 to make its decision may be trained; para. 0032 - cache decision maker 215 itself includes a fast AI model that is trained to determine whether the simple AI model 221 would provide results compatible with the complex AI model; para. 0043 - create the input to train the decision maker; para. 0036 – cache decision maker 215 is trained to recognize the types of inputs (e.g., user requests); para. 0040 - the training approaches described above may be used to train the cache decision maker to determine which type of inputs can be proficiently handled by each of the specialized models; para. 0042 - a classifier model provided in cache decision maker 215 may be trained to classify each input instance.).
Hacmon further teaches or suggests responsible AI (see Fig. 1; para. 0002 - to evaluate the fairness and robustness of any AI-based ML models of systems and applications, several steps are required: it is required to detect and measure various properties of ethical bias from different ethical points of view. Then, it is required to aggregate those different ethical perspectives to one final fairness score. This final fairness score provides the data scientists with quantitative estimation, i.e., assessment for the fairness of the examined ML model and can assist them in evaluations and comparisons of different models; para. 0007 -over twenty-five fairness measurements in the literature, each examines the ML model from a different ethical point of view; para. 0008 - detecting an underlying bias and the fairness level of an ML model, which can be integrated into Continuous Integration/Continuous Delivery processes; para. 0015 - final fairness score aggregation module for aggregating the executed tests results into a final fairness score of the examined model and dataset; para. 0038 - system (in the form of a plugin, for example) can be integrated into a larger system, for examining the fairness and robustness of ML models which try to fulfill various AI-based tasks.).
Accordingly, it would have been obvious to one having ordinary skill before the effective filing date of the claimed invention to modify the system and method, taught in Verma, to include responsible AI for the purpose of efficiently evaluating AI models based on ethical considerations to produce a model score, enabling model comparisons, as taught by Hacmon (0002, 0007, 0015, and 0038).
Claim 7:
Verma further teaches or suggests wherein the … AI controller comprises a set of deterministic rules (see para. 0005 - cache decision maker, coupled to the client interface, configured to analyze the client request, and, based at least in part on the analysis, select either the first AI model or the second AI model to respond to the request; para. 0014 - selection of an appropriate AI model, out of two or more possible models, to respond to a client request; para. 0038 - cache decision maker 215 checks incoming requests and decides whether to use the simple AI model 215 at the edge device 220 (known as a "cache hit"), or the more complex AI model 221 in the cloud 230. both the simple AI model 221 and the complex AI model 231 are in the same location, such as, for example, in an AI enabled load-balancer.).
Hacmon further teaches or suggests responsible AI (see Fig. 1; para. 0002 - to evaluate the fairness and robustness of any AI-based ML models of systems and applications, several steps are required: it is required to detect and measure various properties of ethical bias from different ethical points of view. Then, it is required to aggregate those different ethical perspectives to one final fairness score. This final fairness score provides the data scientists with quantitative estimation, i.e., assessment for the fairness of the examined ML model and can assist them in evaluations and comparisons of different models; para. 0007 -over twenty-five fairness measurements in the literature, each examines the ML model from a different ethical point of view; para. 0008 - detecting an underlying bias and the fairness level of an ML model, which can be integrated into Continuous Integration/Continuous Delivery processes; para. 0015 - final fairness score aggregation module for aggregating the executed tests results into a final fairness score of the examined model and dataset; para. 0038 - system (in the form of a plugin, for example) can be integrated into a larger system, for examining the fairness and robustness of ML models which try to fulfill various AI-based tasks.).
Accordingly, it would have been obvious to one having ordinary skill before the effective filing date of the claimed invention to modify the system and method, taught in Verma, to include responsible AI for the purpose of efficiently evaluating AI models based on ethical considerations to produce a model score, enabling model comparisons, as taught by Hacmon (0002, 0007, 0015, and 0038).
Claim 8:
Verma further teaches or suggests wherein the plurality of available AI systems respectively comprise a plurality of machine-learned models that have been respectively trained using a plurality of different training datasets that are respectively associated with a plurality of different contexts (see para. 0016 - device may be used to receive images and make decisions based on the image, using various AI models trained to detect a problem or one or more defects; para. 0020 - cache decision maker decides, on a per request basis, whether it is better to use the simpler model at the edge, or to use the complex model from the cloud; para. 0023 - classifier model 121 of input analyzer 120 is trained to recognize the types of inputs where the simple model is suitable, and those for which it is not; para. 0029 - accessed by a client application to submit requests, e.g., images of potatoes on a conveyor belt from a potato greening inspection application, or images acquired by a UAV of a bridge surface from a department of highways inspection application; para. 0030 - input analyzer component 240 receives information from the client interface component; para. 0037 - a range of input images. the images are aligned in a specific position; para. which accent is used in a specific speech sample, and depending on the specific accent used, it can direct a speech to text conversion request to one of the specialized; par. 0039 – AI service such as a speech to text converter is considered. The complex speech to text converter is trained to recognize sound samples corresponding to many different accents; para. 0040 – requests originating from a location; para. 0041 - first model may be based on a CNN, a second model may first use principal component analysis to reduce input images to a set of feature vectors and then use a decision tree, a third model may use a recurrent neural network, a fourth model may be trained to process images that are taken in bright sunlight, and a fifth model may be trained for images that are taken under shady conditions.).
Claim 9:
Verma further teaches or suggests wherein the plurality of available AI systems respectively comprise a plurality of machine-learned models that have been respectively fine-tuned using a plurality of different finetuning datasets that are respectively associated with a plurality of different contexts (see para. 0016 - device may be used to receive images and make decisions based on the image, using various AI models trained to detect a problem or one or more defects; para. 0020 - cache decision maker decides, on a per request basis, whether it is better to use the simpler model at the edge, or to use the complex model from the cloud; para. 0023 - classifier model 121 of input analyzer 120 is trained to recognize the types of inputs where the simple model is suitable, and those for which it is not; para. 0029 - accessed by a client application to submit requests, e.g., images of potatoes on a conveyor belt from a potato greening inspection application, or images acquired by a UAV of a bridge surface from a department of highways inspection application; para. 0030 - input analyzer component 240 receives information from the client interface component; para. 0037 - a range of input images. the images are aligned in a specific position; para. which accent is used in a specific speech sample, and depending on the specific accent used, it can direct a speech to text conversion request to one of the specialized; par. 0039 – AI service such as a speech to text converter is considered. The complex speech to text converter is trained to recognize sound samples corresponding to many different accents; para. 0040 – requests originating from a location; para. 0041 - first model may be based on a CNN, a second model may first use principal component analysis to reduce input images to a set of feature vectors and then use a decision tree, a third model may use a recurrent neural network, a fourth model may be trained to process images that are taken in bright sunlight, and a fifth model may be trained for images that are taken under shady conditions.).
Claim 11:
Verma further teaches or suggests wherein the plurality of available AI systems respectively comprise a plurality of different model prompts that are respectively associated with a plurality of different contexts (see para. 0016 - Images from the cameras are fed to an edge device that processes the images using one or more AI models; para. 0023 - Input analyzer 120 performs this task using classifier model 121, which is a third AI model that is trained to determine, for a given client request, whether a simpler local model would provide results compatible with the counterpart complex model stored in the cloud. As described in detail below, classifier model 121 of input analyzer 120 is trained to recognize the types of inputs where the simple model is suitable, and those for which it is not; para. 0027 - which are trained to, and then used to, make inference decisions in response to client requests. a training data generator component 245 of cache decision maker application; para. 0030 - the third AI model used by the input analyzer component 240 to make its decision may be trained; para. 0032 - cache decision maker 215 itself includes a fast AI model that is trained to determine whether the simple AI model 221 would provide results compatible with the complex AI model; para. 0043 - create the input to train the decision maker; para. 0036 – cache decision maker 215 is trained to recognize the types of inputs (e.g., user requests); para. 0039 - cache decision maker 215 is trained to determine which accent is used in a specific speech sample, and depending on the specific accent used, it can direct a speech to text conversion request to one of the specialized, accent-specific simple models, or to the more complex speech to text converter; para. 0040 - the training approaches described above may be used to train the cache decision maker to determine which type of inputs can be proficiently handled by each of the specialized models; para. 0042 - a classifier model provided in cache decision maker 215 may be trained to classify each input instance.).
Claim 12:
Verma further teaches or suggests wherein the plurality of available AI systems comprise a plurality of different and distinct AI systems (see para. 0016 - device may be used to receive images and make decisions based on the image, using various AI models trained to detect a problem or one or more defects; para. 0020 - cache decision maker decides, on a per request basis, whether it is better to use the simpler model at the edge, or to use the complex model from the cloud; para. 0023 - classifier model 121 of input analyzer 120 is trained to recognize the types of inputs where the simple model is suitable, and those for which it is not; para. 0029 - accessed by a client application to submit requests, e.g., images of potatoes on a conveyor belt from a potato greening inspection application, or images acquired by a UAV of a bridge surface from a department of highways inspection application; para. 0030 - input analyzer component 240 receives information from the client interface component; para. 0037 - a range of input images. the images are aligned in a specific position; para. which accent is used in a specific speech sample, and depending on the specific accent used, it can direct a speech to text conversion request to one of the specialized; par. 0039 – AI service such as a speech to text converter is considered. The complex speech to text converter is trained to recognize sound samples corresponding to many different accents; para. 0040 – requests originating from a location; para. 0041 - first model may be based on a CNN, a second model may first use principal component analysis to reduce input images to a set of feature vectors and then use a decision tree, a third model may use a recurrent neural network, a fourth model may be trained to process images that are taken in bright sunlight, and a fifth model may be trained for images that are taken under shady conditions.).
Claim 13:
Verma further teaches or suggests wherein the plurality of available AI systems comprise a single AI system that is dynamically adaptable to a plurality of different and distinct behaviors (see para. 0016 - device may be used to receive images and make decisions based on the image, using various AI models trained to detect a problem or one or more defects; para. 0020 - cache decision maker decides, on a per request basis, whether it is better to use the simpler model at the edge, or to use the complex model from the cloud; para. 0023 - classifier model 121 of input analyzer 120 is trained to recognize the types of inputs where the simple model is suitable, and those for which it is not; para. 0029 - accessed by a client application to submit requests, e.g., images of potatoes on a conveyor belt from a potato greening inspection application, or images acquired by a UAV of a bridge surface from a department of highways inspection application; para. 0030 - input analyzer component 240 receives information from the client interface component; para. 0037 - a range of input images. the images are aligned in a specific position; para. which accent is used in a specific speech sample, and depending on the specific accent used, it can direct a speech to text conversion request to one of the specialized; par. 0039 – AI service such as a speech to text converter is considered. The complex speech to text converter is trained to recognize sound samples corresponding to many different accents; para. 0040 – requests originating from a location; para. 0041 - first model may be based on a CNN, a second model may first use principal component analysis to reduce input images to a set of feature vectors and then use a decision tree, a third model may use a recurrent neural network, a fourth model may be trained to process images that are taken in bright sunlight, and a fifth model may be trained for images that are taken under shady conditions.).
Claim 14:
Verma further teaches or suggests wherein providing, by the computing system, the output data to the user based on the respective output from one or more of the one or more selected AI systems comprises providing as the output data to the user all of the respective outputs from all of the selected AI systems (see Fig. 1, 2, 6, 7; para. 0014 - selection of an appropriate AI model, out of two or more possible models, to respond to a client request; para. 0016 - device may be used to receive images and make decisions based on the image, using various AI models trained to detect a problem or one or more defects; para. 0017 - check parts after manufacture; para. 0018 - inspect bridges for cracks; para. 0021 - client may be, for example, a human user, or, for example, in what is a more common scenario, a computer program that accesses an AI-enhanced web solution; para. 0037 - AI models which are used to examine images to check if a manufactured product is defective or good may be considered; para. 0047 - where a response is provided to the client request from the first Al model; para. 0048 - where a response to the client request is provided from the second AI model; para. 0050 - a request to detect any defects; para. 0053 - where a response is provided to the client from the selected simple Al model.).
Claim(s) 2 is/are rejected under 35 U.S.C. 103 as being unpatentable over Verma, in view of Hacmon, in view of Donovan, in view of Linardatos, and further in view of Meinders, US Publication 2019/0318262 (“Meinders”).
Claim 2:
Verma does not explicitly disclose querying, by the computing system, the user for the context data; and receiving, by the computing system, the context data from the user responsive to said querying.
Meinders further teaches or suggests querying, by the computing system, the user for the context data; and receiving, by the computing system, the context data from the user responsive to said querying (see para. 0053 – selecting an algorithm and/or model based on the received input; para. 0057 - if the user wishes to build an AI system with a binary classifier as the output medium, the artificial intelligence tool will select a machine learning algorithm with a binary classifier; para. 0061 - output can be an AI system; para. 0117 - user is then prompted to enter a "consideration input" and/or an intent at step where they can add cultural context, ethics, etc. (any variable that should be considered
in the design process). These input considerations will be output at the end, and can also be used to highlight information throughout the design that might be relevant to that consideration. There are several benefits to entering the "consideration input", the primary benefit is to build ethics, culture and bias controls into the design
Accordingly, it would have been obvious to one having ordinary skill before the effective filing date of the claimed invention to modify the system and method, taught in Verma, to include querying, by the computing system, the user for the context data; and receiving, by the computing system, the context data from the user responsive to said querying for the purpose of efficiently soliciting information about or from a user for use in model creation, improving model design and effectiveness, as taught by Meinders (0061).
Claim(s) 3 is/are rejected under 35 U.S.C. 103 as being unpatentable over Verma, in view of Hacmon, in view of Donovan, in view of Linardatos, and further in view of Levy, US Patent 11,423,897 (“Levy”).
Claim 3:
Verma further teaches or suggests wherein determining, by the computing system, the context data descriptive of the context associated with the user request (see para. 0020 - cache decision maker decides, on a per request basis, whether it is better to use the simpler model at the edge, or to use the complex model from the cloud; para. 0023 - classifier model 121 of input analyzer 120 is trained to recognize the types of inputs where the simple model is suitable, and those for which it is not; para. 0029 - accessed by a client application to submit requests, e.g., images of potatoes on a conveyor belt from a potato greening inspection application, or images acquired by a UAV of a bridge surface from a department of highways inspection application; para. 0030 - input analyzer component 240 receives information from the client interface component; para. 0037 - a range of input images. the images are aligned in a specific position; para. which accent is used in a specific speech sample, and depending on the specific accent used, it can direct a speech to text conversion request to one of the specialized. Para. 0040 – requests originating from a location.).
Verma does not explicitly disclose comprises automatically inferring the context data.
Levy teaches or suggests automatically inferring the context data (see col. 2, lines 21-36 - generate a response suited for a particular purpose … response data for a particular type of user request having a particular contextual intent and/or to provide response data that includes significant words. inferred based on the context in which the request was made (e.g., the words/phrases used in the request, past purchase history, past searches, or any suitable data associated with the user and/or a group of users that are similar to the user).
Accordingly, it would have been obvious to one having ordinary skill before the effective filing date of the claimed invention to modify the system and method, taught in Verma, to include automatically inferring the context data for the purpose of efficiently generating content suited for a particular purpose using request context, improving requested responses, as taught by Levy (col. 2).
Claim(s) 4 is/are rejected under 35 U.S.C. 103 as being unpatentable over Verma, in view of Hacmon, in view of Donovan, in view of Linardatos, and further in view of Bhaskar et al., US Publication 2021/0194888 (“Bhaskar”).
Claim 4:
Verma further teaches or suggests wherein determining the context data and said executing the … AI controller to select the one or more selected AI systems are performed in response to a determination, by the computing system (see see para. 0020 - cache decision maker decides, on a per request basis, whether it is better to use the simpler model at the edge, or to use the complex model from the cloud; para. 0023 - classifier model 121 of input analyzer 120 is trained to recognize the types of inputs where the simple model is suitable, and those for which it is not; para. 0029 - accessed by a client application to submit requests, e.g., images of potatoes on a conveyor belt from a potato greening inspection application, or images acquired by a UAV of a bridge surface from a department of highways inspection application; para. 0030 - input analyzer component 240 receives information from the client interface component; para. 0037 - a range of input images. the images are aligned in a specific position; para. which accent is used in a specific speech sample, and depending on the specific accent used, it can direct a speech to text conversion request to one of the specialized. Para. 0040 – requests originating from a location.).
Verma does not explicitly disclose responsible AI … that the user request comprises a sensitive request.
Hacmon further teaches or suggests responsible AI (see Fig. 1; para. 0002 - to evaluate the fairness and robustness of any AI-based ML models of systems and applications, several steps are required: it is required to detect and measure various properties of ethical bias from different ethical points of view. Then, it is required to aggregate those different ethical perspectives to one final fairness score. This final fairness score provides the data scientists with quantitative estimation, i.e., assessment for the fairness of the examined ML model and can assist them in evaluations and comparisons of different models; para. 0007 -over twenty-five fairness measurements in the literature, each examines the ML model from a different ethical point of view; para. 0008 - detecting an underlying bias and the fairness level of an ML model, which can be integrated into Continuous Integration/Continuous Delivery processes; para. 0015 - final fairness score aggregation module for aggregating the executed tests results into a final fairness score of the examined model and dataset; para. 0038 - system (in the form of a plugin, for example) can be integrated into a larger system, for examining the fairness and robustness of ML models which try to fulfill various AI-based tasks.).
Accordingly, it would have been obvious to one having ordinary skill before the effective filing date of the claimed invention to modify the system and method, taught in Verma, to include responsible AI for the purpose of efficiently evaluating AI models based on ethical considerations to produce a model score, enabling model comparisons, as taught by Hacmon (0002, 0007, 0015, and 0038).
Bhaskar further teaches or suggests that the user request comprises a sensitive request (see para. 0003 - determining, by the client device, that the requested content includes sensitive information based on a profile associated with the client device and a context in which the content is being accessed; para. 0164 - applicable to the client device and/or user may be identified. identify applicable security policy for a given request (e.g., a request sent by a client device) based on information within the request (e.g., information identifying the device, the user, or the organization; para. 0165 - it can be determined if the user is permitted to view the sensitive information based on the applicable security policy, the device profile, and/or the context of the access request.).
Accordingly, it would have been obvious to one having ordinary skill before the effective filing date of the claimed invention to modify the system and method, taught in Verma, to include responsible AI for the purpose of efficiently forming a particular response where the request is determined to be a sensitive request, improving response appropriateness, as taught by Bhaskar (0164 and 0165).
Claim(s) 10 is/are rejected under 35 U.S.C. 103 as being unpatentable over Verma, in view of Hacmon, in view of Donovan, in view of Linardatos, and further in view of Khov, US Publication 2021/0329338 (“Khov”).
Claim 10:
Verma further teaches or suggests wherein the plurality of available AI systems respectively (see para. 0023 - Input analyzer 120 performs this task using classifier model 121, which is a third AI model that is trained to determine, for a given client request, whether a simpler local model would provide results compatible with the counterpart complex model stored in the cloud. As described in detail below, classifier model 121 of input analyzer 120 is trained to recognize the types of inputs where the simple model is suitable, and those for which it is not; para. 0027 - which are trained to, and then used to, make inference decisions in response to client requests. a training data generator component 245 of cache decision maker application; para. 0030 - the third AI model used by the input analyzer component 240 to make its decision may be trained; para. 0032 - cache decision maker 215 itself includes a fast AI model that is trained to determine whether the simple AI model 221 would provide results compatible with the complex AI model; para. 0043 - create the input to train the decision maker; para. 0036 – cache decision maker 215 is trained to recognize the types of inputs (e.g., user requests); para. 0040 - the training approaches described above may be used to train the cache decision maker to determine which type of inputs can be proficiently handled by each of the specialized models; para. 0042 - a classifier model provided in cache decision maker 215 may be trained to classify each input instance.).
Khov further teaches or suggests comprise a plurality of classification filters that have been respectfully designed to detect and filter undesirable content associated with a plurality of different contexts (see para. 0031 - defined data can be associated with, e.g., offensive, inappropriate (e.g., curse words, offensive vocalizations, explicit visual material, etc.), and/or undesirable content. The content recognition system can attempt to match the identified content with the defined data or some portion of the defined data; para. 0032 - the content recognition system may utilize classification models, such as appropriately trained neural networks, to detect groups of frames or individual frames with undesirable content. Recognizing certain kinds of undesirable content, such as violent actions or profane words may utilize a model trained with temporal information such as a recurrent neural network or long short-term memory (LSTM) neural network or gated recurrent unit (GRU) neural network; para. 0064 – content modification system 124 can utilize the classification (e.g., offensive, inappropriate, undesirable, etc .) of the content to filter (e.g., replace, remove, obfuscate) the content; para. 0100 - content modification system 124 can utilize the classification (e.g., offensive, inappropriate, undesirable, etc.) of the content to filter (e.g., replace, remove, obfuscate) the content.).
Accordingly, it would have been obvious to one having ordinary skill before the effective filing date of the claimed invention to modify the system and method, taught in Verma, to include comprise a plurality of classification filters that have been respectfully designed to detect and filter undesirable content associated with a plurality of different contexts AI for the purpose of efficiently identifying content for modification based on specific classifications, improving output appropriateness, as taught by Khov (0031, 0032, 0064, 0100).
Claim(s) 15 is/are rejected under 35 U.S.C. 103 as being unpatentable over Verma, in view of Hacmon, in view of Donovan, in view of Linardatos, and further in view of Reznik, US Publication 2021/0026906 (“Reznick”).
Claim 15:
Verma further teaches or suggests wherein providing, be the computing system, the output data to the user based on the respective output from one or more of the one or more selected AI systems comprises: … outputs from all of the selected AI systems (see Fig. 1, 2, 6, 7; para. 0014 - selection of an appropriate AI model, out of two or more possible models, to respond to a client request; para. 0016 - device may be used to receive images and make decisions based on the image, using various AI models trained to detect a problem or one
or more defects; para. 0017 - check parts after manufacture; para. 0018 - inspect bridges for cracks; para. 0021 - client may be, for example, a human user, or, for example, in what is a more common scenario, a computer program that accesses an AI-enhanced web solution; para. 0037 - AI models which are used to examine images to check if a manufactured product is defective or good may be considered; para. 0047 - where a response is provided to the client request from the first Al model; para. 0048 - where a response to the client request is provided from the second AI model; para. 0050 - a request to detect any defects; para. 0053 - where a response is provided to the client from the selected simple Al model.).
Reznick further teaches or suggests filtering all of the respective outputs … to generate a final subset of one or more outputs, and providing the final subset as the output data to the user (see Fig. 17; para. 0051 - an algorithm or Artificial Intelligence (AI) method may comprise a search tool, the algorithm or AI method being used to search data and/or to generate data (i.e., a search tool may, in some embodiments, comprise a data generation tool); para. 0069 - refine or modify a set of search results, or pre-emptively effect a list of search results to be output by search software, by providing feedback such as indication(s) that one or more particular search result(s) are considered undesirable and removing from a listing of search results the indicated search result(s) along with additional search result(s) that are determined to be associated with, corresponding to, relevant or likely to be undesirable for the same reason(s) as the indicated search result(s ); para. 0134 - identified by the server or software application as being undesirable based on the user's indication of undesirable search results; para. 0186 - search tool may be programmed to make an assumption or inference that these search results are undesirable and offer to remove (or automatically remove, in accordance with some embodiments) similar search results; para. 0187 - search tool comprises AI/machine learning (e.g., an algorithm that may be trained and evolves over time) and utilizes this to recognize patterns or information from the user's passive input that indicates certain search results or types of search results.).
Accordingly, it would have been obvious to one having ordinary skill before the effective filing date of the claimed invention to modify the system and method, taught in Verma, to include filtering all of the respective outputs … to generate a final subset of one or more outputs, and providing the final subset as the output data to the user for the purpose of efficiently applying AI and machine learning to filter results to obtain an optimized set, improving responses to user requests, as taught by Reznik (0186 and 0187).
Claim(s) 16 is/are rejected under 35 U.S.C. 103 as being unpatentable over Verma, in view of Hacmon, in view of Donovan, in view of Linardatos, and further in view of Mital et al., US Publication 2020/0159856 (“Mital”).
Claim 16:
Mital more specifically teaches or suggests wherein: the user request comprises a search query; and the output data comprises one or more search results responsive to the search query (see Fig. 1, 5; para. 0006 - part of a search session, receiving user input at a search engine. One or more searches on a set of data using the user input. Search results are provided from the one or more searches to a user. Based on a history of the search session, suggestions are provided in a user interface of AI models that could be applied to expand potential search results for the search session. User input is received at the user interface selecting one or more of the suggested AI model. The one or more selected AI models are applied to expand the set of data. Search results to the user based on searching the expanded set of data; para. 0015 - search engine user interface can provide suggested augmentation artificial intelligence (AI) models to the user that can be applied to the search session. In particular, the AI models can operate on certain data to produce additional augmentation data that can be refined, indexed, and searched to provide additional searchable data to the user that were not previously available; para. 0016 – An augmentation AI model produces data that augments the input data according to a predetermined augmentation goal of the augmentation AI model.).
Accordingly, it would have been obvious to one having ordinary skill before the effective filing date of the claimed invention to modify the system and method, taught in Verma, to include wherein: the user request comprises a search query; and the output data comprises one or more search results responsive to the search query for the purpose of efficiently augmenting search results using AI models with a particular augmentation goal based on user search inputs, expanding and augmenting search results, as taught by Mital (0015 and 0016).
Response to Arguments
Applicant’s further arguments have been considered but are not persuasive because the arguments do not correspond to the rationales as used in the current rejection.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Andrew T McIntosh whose telephone number is (571)270-7790. The examiner can normally be reached M-Th 8:00am-5:30pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Tamara Kyle can be reached at 571-272-4241. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/ANDREW T MCINTOSH/Primary Examiner, Art Unit 2144
