Murdoch University Research Repository

Welcome to the Murdoch University Research Repository

The Murdoch University Research Repository is an open access digital collection of research
created by Murdoch University staff, researchers and postgraduate students.

Learn more

Data-driven, Mediterranean, A.I. simulations and the ethical implications of embedded qualitative bias in digital twin deepfake games

Svatos, Karl Benton WestORCID: 0000-0003-1468-0601 (2021) Data-driven, Mediterranean, A.I. simulations and the ethical implications of embedded qualitative bias in digital twin deepfake games. PhD thesis, Murdoch University.

[img]
Preview
PDF - Whole Thesis
Download (15MB) | Preview

Abstract

Put simply, the question that has intrigued me and led me to do a PhD is to undertake scientific research which gave me the opportunity to study the underlying ethical and philosophical bases of the research process itself. The thesis takes the reader through real-life studies which combine developing computational power with the handling of large datasets and the ethics of data collection and ownership. This journey resulted in the study and representation of real-life and in-silico or digitalised copies of research data (called digital twins) and compares their physical and in-silico representation from an ethical viewpoint.

Digital twins and deepfake imagery (computer simulations based on mathematical principles) may be used to verify significance of statistical outcomes during artificial simulations’ via computational processes. Deepfakes are a possible occidental progression due to increasingly complex environmental variation being used to explain adaptation, or an evolutionary process because variation is oft unable to be replicated in experiments, e.g. for political reasons.

A renaissance of informatics ‘big-data science;’ implementing adaptive real-time A.I. via digital twin construction, to predicted outcomes concerning simulations’ predictability themselves’, digital twins can be used to infer without disturbing system balance. Extensive implementations of this technology with adaptive frequency hopping based on Gaussian ‘pseudo-random signalling’ now exist. Several model species’ were selected including hairy marron, dairy farm cattle, and a unique Australian Tibetan barley. Using environmental data from the Mediterranean environment; environmental growth, habit, and sensory behaviour; digital twins’ predictions based on complementary statistical heuristics included incorporating genome sequences, and phenotype data through embedded kernalisation telematic software, for automated species informatics simulation. ‘Murdoch Twins’ was created to test whether TCP/IP latency data supporting machine learning programs, and visualisation statistics could alter both quantitative and qualitative outcomes (in a simulated game carried out in 2020 in Japan).

This thesis makes use of a series of quantitative experiments that were conceived, proposed, carried out, and are written up as four separate manuscripts (two results manuscripts published in 2018 and 2019 respectively, one abstract presented in Osaka Japan in 2018, and one manuscript currently under peer review), with seven supporting appendices (including industry specific publications, data collection and analyses). The full contributions span two decades culminated in a move to Murdoch University in 2016. The final thesis product only possible by the award, of a fully funded, Murdoch University postgraduate research scholarship, administered by The Western Australian, State, Agricultural, and Biotechnology Center (SABC) under the tutelage of Professor Michael Jones. And has since generated significant research-related opportunities including, “PhD scholarship in digital connectivity and big-data analytics in agriculture”, administered through the newly coined ‘agtech group’ at Murdoch University. The research experiments were made possible through contributions including through the receipt of smaller financial bursaries (grant monies), in-kind contributions, corporate awards (purchased and/or borrowed equipment, including hired casual staff), or industry contributions. Some research was also personally funded, or earned through internships, further grants rewards, in-kind volunteering, and/or while consulting for Karl Corporation (including in Japan, through 2020 at Kazusa DNA Research Institute in Chiba Prefecture as a visiting professional informatics scientist).

The thesis, “Data-driven, Mediterranean, A.I. simulations and the ethical implications of embedded qualitative bias in digital twin deepfake games” contains an original, standalone, independently researched introduction, methodology, with the sequential and logical presentations of four manuscripts (results chapters), culminating in the general results titled, “Murdoch Twins: containerised .NET API backend data implementation for real-time deepfake games”, proceeded by the general discussion. As it relates to the thesis in its entirety; the combined output is the development of a new, computer-based, rapid, real-time, data visualisation strategy (and technique), that facilitated the creation of a digital twin called ‘Murdoch Twins.’ The purpose of Murdoch Twins is for rapidly measuring and quantifying, the digital authenticity of informatics analyses (via construction of a digital twin including for a confidential* Tibetan-Australian barley species; a type of doubled haploid population of almost identical genetic clones); based on complementary environmental, sequencer-based genotyping, phenotype data lakes (for the Mediterranean environment); through the creation of a trust enabled 5G/LTE, IoT, core to core, embedded kernel stack, to therefore test that it may be penetrated, remotely, (via VPN), with deepfake imagery, during a real-time A.I. game simulation (in a real-world scenario delivered through .NET APIs).

Implications deepfake games have regarding current real-world scenarios include real-time medicine, geophysical and A.I., predictions; but as it relates to the manuscripts and the central tenet of the thesis deepfake games concerns the ethical use of deepfake images to securely verify the digital authenticity of phenotype data, as it relates to environmental variation of raw ‘trusted’ data sources, and their use in qualitative descriptions (embedded bias).

The four manuscripts demonstrate that a methodological progression was followed, concerning the development of the aims and the hypothesis. Tables and figures throughout the thesis are sequential and are not broken up for individual manuscripts. Because of the strategic nature of the thesis, there were significant challenges along the way. To address this, each manuscript has a preamble, including its own declaration statement, that summarises the authors’ contributions’ at that time, as well a peroratory summation. Additional contributions, as they relate to the generation of each manuscript of the thesis (in terms of the overall aims and hypothesis) are addressed in 7 appendices. Contributions reflect the time required as follows.

1. Manuscript one: An independently researched, student-funded, sole-author, published manuscript (Karl Svatos (KS) contribution 100%),
•“Svatos KBW, 2018. Commercial silicate phosphate sequestration and desorption leads to a gradual decline of aquatic systems. Environmental Science and Pollution Research 25, 5386-92. doi: 10.1007/s11356-017-0846-9”

2. Manuscript two: an independently researched, partially industry-funded (Dairy Australia Grant; UWA 13344), first-author, published manuscript (KS contribution 95%, (UWA staff 5% includes Em. Prof. Abbott (LA) contribution 3%)),
• Svatos KBW, Abbott LK, 2019. Dairy soil bacterial responses to nitrogen application in simulated Italian ryegrass and white clover pasture. Journal of Dairy Science 102, 9495- 504. doi: 10.3168/jds.2018-16107 o Original research description (visualisation stats) (Appendix 3 preamble)

3. Manuscript three: an, independently researched, privately funded, first-author, published conference, journal abstract (Osaka Japan, 2018),
• “Svatos KBW, Diepeveen D, Abbott LK, Li C, 2018. Big data GPU/CPU kernalisation pipeline for API based quantitative genetic assessments in field-based drone research (Abstract Submitted). Journal of Plant Pathology and Microbiology, 9. doi: 10.4172/2157-7471-C2-011”
And two subsequent, corresponding, report-style projects with unpublished results (Project 1 and Project 2), and including one published, strategic white position paper associated with Project 1,
• Project 1; a student-led, partially industry-funded (GRDC Grant; UMU00049) collaboration between Murdoch University, The Western Australian Department of Primary Industries and Regional Development (DPIRD) (formerly Department of Agriculture), UWA, Scientific Aerospace, Karl Corporation, and The Western Crop Genetics Alliance (formerly Western Barley Genetics alliance and affiliated Australian institutions). (KS contribution 55%), (DPIRD staff 20% includes Dr. Diepeveen (DD) 10% and Prof. Li (CL) 5%), (Murdoch staff 10% includes DD 5% Prof. Jones (MJ) 2%, CL 1%, Dr. Murray (DM) 1%, Dr. Hill (CH) 1%), (UWA 3% includes LA 2%) (Scientific Aerospace includes Mr. Trowbridge RET. (GT) 2%) (Karl Corporation 10%).
• “Rapid downstream glasshouse field trial phenotype assessments for variability minimisation in GPU core processing and telematics data analyses”
• “Svatos K, Trowbridge G, 2018. Australian drone technology assisting a significant step in crop tolerance to heat and drought stress. Future Directions International. http: futuredirections.org.au/publication/australian-drone-technology-assisting-significantstep- crop-tolerance-heat-drought-stress/”
• Project 2; a student-led, independently organised, resourced, and partially industryfunded, project collaboration between Murdoch University, UWA, DPIRD, Pivotel, Nokia-Bell, Microsoft, Precision Ag, Edith Cowan University (ECU), Kazusa DNA Research Institute (KDRI), and Karl Corporation (and affiliated partners). (KS contribution 30%), (Murdoch staff 25% includes MJ 15%, DM 10%), (DPIRD staff 10% includes DD 5%), (Pivotel, Nokia-Bell, Microsoft, Precision Ag, ECU, UWA, and KDRI 30%), (Karl Corporation 5%).
• “A scaleable, private LTE/4G, Boolean GPU networking stack for automated, remote, IoT decision making”

4. Manuscript 4; an industry-led, partially industry-funded, jointly student industry conceived, run, and managed, first author (unpublished), data-science research collaboration between iPREP, Murdoch University, ECU, and the industry partner udrew. (KS contribution 50%), (udrew 50% includes AR (Angela Recaldes) and ZA (Zubair Ahmed) 10%).
“Heuristics encanced SAAS platform: remote geospatial machine learning of soil profiles from an ancient Mediterranean environment” Each manuscript in the thesis is complete and acknowledges all authors’ and contributions’. Additional research methods, results, and discussion generated during this research are addressed in the disclaimer at the beginning of the methodology, and in the privacy and confidentiality statement after the preamble of manuscript three. Individual appendices also contain declaration statements about the significance and relevance to the thesis aims and hypothesis, concerning co-authors’ contributions’ respectively (including Karl Corporation).

The research presented in this thesis shows that the rapid rise of data-driven ‘A.I., big-data science’, has an embedded, objective bias that quantitative computation cannot be used to solve in all real-time simulations. Predictions were supported through the creation of binary-tree data islands. Supporting technologies were connected through an embedded pythonic .NET API (AARCH64) and then utilised to create a digital twin to assess deepfake risk factors via the digital twin (concerning data, security, and ownership). The implications are substantial for this type of implementation due to the ever-expanding collection and use of said data to support qualitative interpretation for action by humans as it relates to A.I. ethics. This process may offer scientists, engineers, land managers, farmers and governments an advantage; knowing how a change (Δ) at any given time (t) might alter an organism’s behaviour, based on issued quantitative source-code trust certificates (.NET APIs, in LTE/5G, real-time). However, there are no ‘real’ solutions in non-binary calculations. Using deepfakes in digital twins to model game outcomes thus resulted in occidental natural latency ‘blips’. Trusted, quantitative, A.I., source-code program manifests, only support purely open-source hypotheses testing.

Item Type: Thesis (PhD)
Murdoch Affiliation(s): Western Australian State Agricultural Biotechnology Centre
College of Science, Health, Engineering and Education
Supervisor(s): Jones, Michael and Murray, David
URI: http://researchrepository.murdoch.edu.au/id/eprint/63264
Item Control Page Item Control Page

Downloads

Downloads per month over past year