Charitable Trust formed the Artificial Intelligence Design Team
TOKYO GLOBAL ENGINEERING CORPORATION CHARITABLE TRUST
ARTIFICIAL INTELLIGENCE DESIGN TEAM
Introduction and purpose
WHEREAS the potential role of artificial intelligence in eradicating trade in firearms, munitions, and explosives has gained increased value following recent attacks by Boko Haram and around the world; and
WHEREAS artificial intelligence has increasingly been the subject of weaponization instead of disarmament; and
WHEREAS no systematic investigation has been undertaken of the possibility to use artificial intelligence to stop arms trafficking; and
THE TOKYO GLOBAL ENGINEERING CORPORATION CHARITABLE TRUST THUS CALLS ON all concerned persons, anywhere, to form teams, including for academic credit (which can be assigned to university students by the Charitable Trust), to design such an artificial intelligence system.
The targeted problem, exactly
Law-enforcement activities are unable to identify ammunition-shipment paths that end with Boko Haram, using current forensics. Unlike chicken-egg distributors, ammunition manufacturers do not want bullets to be easily traceable. Although ammunition tracing systems have been designed, none have met implementation efforts, without opposition. Indeed, no bullet maker wants a piece of lead to be pulled from an infant’s skull and linked to a maker’s mold. Unless a user’s weapon has be examined by a ballistics investigator, the act of putting a bullet in someone is, presently, an anonymous act. However, in the absence of ammunition, firearms are mere bludgeons. TGECCT proposes to design an artificial intelligence system to help law-enforcement activities identify ammunition shipments, for seizure, before shipments can be obtained by Boko Haram.
TGECCT’s three-step path to a possible solution, using artificial intelligence
1.) Make a model (abstract, enter-numbers-in-the-blank) of how weapons and munitions are transferred and used, both legally and illegally.
2.) Enter extant data about manufacture and transfer, plus observational data (whatever is intercepted along the pipeline, &c).
3.) Use Bayesian inference methodology to back trace and distinguish, as much as possible, the unknown parameters from the known parameters.
This would not be a panacea. No system could conclude that “Shipment X was used in Massacre Y.” However, there would be elicited probabilistic inferences, ranges of possibility, and suggestive patterns, sometimes strong enough to justify field-investigation.
This methodology is already employed by scientists, elsewhere, to measure unobservable data. For example, astrophysicists cannot examine most things they study; however, by having a partial understanding of how things like light and gravitational waves move between places, and by having probable models of stellar composition, galactic interstellar gas, &c, they can learn much, even at a distance, even though most of what they want to measure is not accessible via direct observation. Again, not a panacea, but what TGECCT proposes is possible, and, in two-dimensional, terrestrial contexts, it could be extremely effective in pinpointing specific links that could then be explored, via more direct methodology.
Though there are many easily-accessible instructional materials regarding Bayesian statistics and Bayesian inference, an informal example (not meant to be an ignorant example that would result in disregard for a tool, itself) would be thus. Suppose the Nigerian Army were to discover 100 different weapons during raids of a given week. This would not mean that there were 100 weapons at the beginning of the week and at the end of the week there were none. However, were one to have a good mathematical model of how likely a member of Boko Haram were to be apprehended (which could be further modeled based on data such as geographic coordinates, time of day, weather conditions, &c), then one could work backwards and derive a probability distribution that would deduce something to the effect that, “With 100% certainty there were at least 100 guns; with 90% certainty there were at least 187; &c; and there is a 5% chance that there are more than 2,000.” In practice, this is done either through bespoke code or leveraging Stan statistical software, often through R language and environment. One could code such a model using Stan language, and supply both the model and the observations to the system, which would generate a set of probability distributions for each unobserved variable of interest.
After that, one would then need to extend a model to include observable variables. For example, in the previous example, one might have begun attempting to model the number of guns in a village without thinking about apprehensions, but then extend the model to include apprehensions because apprehensions are measurable, and thus use apprehensions as a first link in a chain backwards to hidden parameters. Next, one would need to construct data sets for all observable variables. Finally, one must then translate a model into something statistical software could compute, run the inference (which is often an iterative process, working with domain experts) and, then, analyze the results. Obtaining the knowledge of domain experts, to input into a formal model, would likely be the hardest step; but, achieve this, and measurably save lives.
University student design team
The Tokyo Global Engineering Corporation Charitable Trust (TGECCT) introduces its artificial intelligence design team (AIDT).
Participants enrolled in TGECCT’s AIDT are given the opportunity to earn academic credit while enrolled in an unpaid internship, and while stopping Boko Haram. This internship experience serves as an experiential learning activity that is designed to help learners apply academically learned concepts, theories, principles, and skills in an artificial intelligence laboratory setting. Participants will have opportunities to gain additional knowledge, expertise, and experience, while allowing TGECCT the opportunity to develop potential future projects.
The program’s weekly “field trips,” from beanbag laboratories of Tokyo’s most profitable video-game designers, to tours of the hottest technology conventions in the Tokyo area, ensure that any participant will learn how to use artificial intelligence tools, while meeting top industry movers that constantly design and implement revolutionary AI systems.
Whether you’d like to know whether AI can solve all anthropogenic problems, or whether you’d like to use the latest, cutting-edge technology to find new and improved ways of eliminating violence, this program may be exactly what you’re seeking.
To participate in the Tokyo Global Engineering Corporation Charitable Trust (TGECCT) artificial intelligence design team (AIDT), the following must be fully understood:
1.) The applicant has read and fully understands the following information, the AIDT syllabus, and all expectations.
2.) A TGECCT Form 15, Internship Placement Agreement, must be completed prior to enrolling in the class and sent to the AIDT coordinator. Forms may be requested from TGECCT Public Affairs.
3.) Members must have emergency health-insurance coverage. Questions regarding student insurance must be directed to academic institutions awarding academic credit.
4.) Academic grades will be determined via the following:
Logbook entries: A brief log of internship-participant activities for each day of the design period is required. All logbook entries for each week must be submitted to the AIDT coordinator before the end of operation hours on the last operation day of the week, in portable digital format (PDF), using the following naming scheme: TGECCTnumber_Name_Name_Logbook Entry ## MONTH YEAR. Naming example: A12345_Godzilla_Logbook Entry 06 JAN 2020.pdf. It is the team member’s responsibility to maintain electronic backup copies of all logbook entries, and to present all entries as a single PDF document at the end of the design period. Failure to maintain timely logbook entries can result in lower recommended final grades. A sample logbook entry is located here, and a blank form is here.
Laboratory supervisor evaluations: A separate evaluation is required at the end of each calendar month of the design period. The evaluation form will be completed by the team member’s assigned supervisor (as indicated on the Form 15) and returned to the AIDT coordinator, when complete. It is the team member’s responsibility to ensure that the supervisor receives the evaluation form for completion. A blank evaluation form is available here. Design-team members should communicate with their supervisors when the end of each calendar month approaches and present evaluation forms. Again, it is the team member’s responsibility to give an evaluation form to one’s supervisor. Please, notify the AIDT coordinator of any additional evaluation criteria not provided on the form. Laboratory supervisors are free to implement any additional control necessary, to ensure maximum efficiency, and participants must be evaluated, in writing, for each criterion.
Final report: A three- to four-page, single-spaced report, replete with citations, covering the nature of the team member’s assignment and how the experience helped in obtaining one’s learning objectives, is required toward the conclusion of the design period. Such reports should focus on one or two major points of knowledge gained during the design period. The final report is to be e-mailed to the AIDT coordinator prior to the beginning of the last week of the design period. It will be graded using the final report grading rubric, which is located here.
The complete logbook (all entries as a single document) and the final report must be e-mailed to the AIDT coordinator prior to the beginning of the last week of the academic period, to be included in any grade calculations.
5.) The AIDT coordinator will contact the student and the student’s academic supervisor during the design period, to assess the student’s progress. This requires the student to keep the AIDT coordinator informed of any changes as the period progresses, such as changes in supervisors, additional laboratory duties, new telephone number, et cetera.
Design team member responsibilities:
∙ Request and complete an application package.
∙ Ensure health-insurance coverage during design period.
∙ Keep AIDT coordinator informed of progress and changes.
∙ Present evaluation form to supervisor for completion at the end of each month.
∙ Send each week’s daily logbook entries to AIDT coordinator at the end of each week.
∙ Send final report to AIDT coordinator before beginning the final week of the semester.
∙ Save lives.
NOTE 1: The TGECCT artificial intelligence design team is distinctly different from positions with the TGECCT artificial intelligence internship program. The TGECCT artificial intelligence design team involves learning how to design an artificial intelligence system–and doing it, while saving lives. TGECCT’s AI internship program, on the other hand, creates theoretical artificial intelligence systems, but does not implement them, and presents the designs in an academic publication, much like the differences between a university automotive technology program and a university department of mechanical engineering. In either case, academic credit can be awarded, by TGECCT.
NOTE 2: Traveling to (and around) Tokyo is not a requirement. Participants can contribute, from any location on the planet. Advisors, too, are welcome (and needed), such as to answer questions, via e-mail.