# US 12,182,718 B2 · Creating diversity in artificial intelligence and machine learning > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,182,718 B2 - Document ID: US12182718B2 - Status: Granted - Kind: B2 - Application number: 18/434,587 - Filed: 2024-02-06 - Published: 2024-05-30 - Granted: 2024-12-31 - Earliest claimed: 2019-02-20 - Assignees: Sony Interactive Entertainment Inc. - Applicants: Sony Interactive Entertainment LLC - Inventors: Albhy Galuten - Web page: https://patentgrep.com/patent/US12182718B2 ## CPC - G06N 3/086: using evolutionary algorithms, e.g. genetic algorithms or genetic programming - G06F 18/214: Generating training patterns; Bootstrap methods, e.g. bagging or boosting - G06F 18/2413: based on distances to training or reference patterns - G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] - G06N 3/0464: Convolutional networks [CNN, ConvNet] - G06N 3/047: Probabilistic or stochastic networks - G06N 3/09: Supervised learning - G06N 20/00: Machine learning - G06N 3/044: Recurrent networks, e.g. Hopfield networks - G06N 3/045: Combinations of networks - G06N 3/126: Evolutionary algorithms, e.g. genetic algorithms or genetic programming - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract The present invention concern systems and methods for maintaining diversity in AI and ML environments through the cooperation of various AI and ML systems such that they optimize for social and cultural diversity. The examination of behavior, infrastructure, and governance, mimicking of genetic biodiversity, and application of the foregoing to machine reasoning mitigates the tendency of systems to find optimized or single best solutions. AI and ML environments may thus derive multiple diverse solutions that contribute to richer ecosystems in which human beings may function and thrive. ## Claims ### Claim 1 A system for dynamic reconfiguration of machine learning datasets, the system comprising: memory that maintains information including one or more different datasets concerning behaviors, infrastructure, and governance; and a software engine maintained in a non-transitory storage medium and executable by a processing device to: receive a selection of a training set from among the different datasets, wherein the training set is selected as a starting point; identify one or more groupings including at least one niche grouping within the training set and one or more relationships between the groupings, wherein two or more of the groupings share data in common, and wherein the data is maintained separately for each of the groupings; continuously monitor one or more social networks using artificial intelligence to identify one or more social trends and to extrapolate the social trends into one or more factors; apply the extrapolated factors to the training set and one or more commonalities; and generate a prediction regarding behaviors, infrastructure, and governance at another point in time based on a trend using one or more neural networks, wherein the trend includes one or more changes in the one or more groupings and the relationships based on the extrapolated factors. ### Claim 2 (depends on 1) The system of claim 1, wherein each dataset of the one or more different datasets represents different points in time. ### Claim 3 (depends on 1) The system of claim 1, wherein the software engine is further executable to generate one or more trend lines based on the one or more different datasets representing different points in time. ### Claim 4 (depends on 3) The system of claim 3, wherein the software engine is further executable to map commonality scores based on the one or more trend lines. ### Claim 5 (depends on 1) The system of claim 1, wherein the software engine is further executable to identify a factor of the one or more factors to be removed from the training set. ### Claim 6 (depends on 1) The system of claim 1, wherein the software engine is further executable to identify a factor of the one or more factors to be added to the training set. ### Claim 7 (depends on 1) The system of claim 1, wherein the software engine is further executable to introduce cross-pollination of different groupings into the training set and to update the training set further based on the introduction of cross-pollination. ### Claim 8 (depends on 7) The system of claim 7, wherein the software engine is further executable to track the relationships between the groupings after the introduction of cross-pollination. ### Claim 9 (depends on 7) The system of claim 7, wherein the software engine is further executable to monitor the niche grouping for changes in niche population after the introduction of cross-pollination and to identify impacts to the niche grouping from inadvertent cross-pollination. ### Claim 10 (depends on 1) The system of claim 1, wherein the identified one or more relationships include a correlation and causation. ### Claim 11 (depends on 1) The system of claim 1, wherein the software engine is further executable to forensically track a provenance for one or more machine decisions. ### Claim 12 (depends on 11) The system of claim 11, wherein the software engine is further executable to restricting access to structured data regarding the forensically tracked provenance in accordance with one or more territories. ### Claim 13 A method for dynamic reconfiguration of machine learning datasets, the method comprising: maintaining information including one or more different datasets concerning behaviors, infrastructure, and governance; receiving a selection of a training set from among the different datasets, wherein the training set is selected as a starting point; identifying one or more groupings including at least one niche grouping within the training set and one or more relationships between the groupings, wherein two or more of the groupings share data in common, and wherein the data is maintained separately for each of the groupings; continuously monitoring one or more social networks using artificial intelligence to identify one or more social trends and to extrapolate the social trends into one or more factors; applying the extrapolated factors to the training set and one or more commonalities; and generating a prediction regarding behaviors, infrastructure, and governance at another point in time based on a trend using one or more neural networks, wherein the trend includes one or more changes in the one or more groupings and the relationships based on the extrapolated factors. ### Claim 14 (depends on 13) The method of claim 13, wherein each dataset of the one or more different datasets represents a different point in time. ### Claim 15 (depends on 13) The method of claim 13, further comprising generating one or more trend lines based on the one or more different datasets representing different points in time. ### Claim 16 (depends on 15) The method of claim 15, further comprising mapping commonality scores based on the one or more trend lines. ### Claim 17 (depends on 13) The method of claim 13, further comprising identifying a factor of the one or more factors to be removed from the training set. ### Claim 18 (depends on 13) The method of claim 13, further comprising identifying a factor of the one or more factors to be added to the training set. ### Claim 19 (depends on 13) The method of claim 13, further comprising introducing cross-pollination of different groupings into the training set, and updating the training set based on the introduction of cross-pollination. ### Claim 20 (depends on 19) The method of claim 19, further comprising tracking the relationships between the groupings after the introduction of cross-pollination. ### Claim 21 (depends on 13) The method of claim 13, further comprising monitoring the niche grouping after the introduction of cross-pollination, and identifying impacts to the niche grouping from inadvertent cross-pollination. ### Claim 22 (depends on 13) The method of claim 13, wherein the identified one or more relationships include a correlation and causation. ### Claim 23 (depends on 13) The method of claim 13, further comprising forensically tracking a provenance for one or more machine decisions. ### Claim 24 (depends on 23) The method of claim 23, further comprising restricting access to structured data regarding the forensically tracked provenance in accordance with one or more territories. ### Claim 25 A non-transitory, computer-readable storage medium, having embodied there on a program executable by a processor to perform a method for dynamic reconfiguration of machine learning datasets, the method comprising: maintaining information including one or more different datasets concerning behaviors, infrastructure, and governance; receiving a selection of a training set from among the different datasets, wherein the training set is selected as a starting point; identifying one or more groupings including at least one niche grouping within the training set and one or more relationships between the groupings, wherein two or more of the groupings share data in common, and wherein the data is maintained separately for each of the groupings; continuously monitoring one or more social networks using artificial intelligence to identify one or more social trends and to extrapolate the social trends into one or more factors; apply the extrapolated factors to the training set and one or more commonalities; and generating a prediction regarding behaviors, infrastructure, and governance at another point in time based on a trend using one or more neural networks, wherein the trend includes one or more changes in the one or more groupings and the relationships based on the extrapolated factors.