ISSIP Blog Post: https://issip.org/2024-collab-ai_digital_twins/
Size: 905.94 KB
Language: en
Added: Apr 27, 2024
Slides: 3 pages
Slide Content
BACKGROUND The International Society of Service Innovation Professionals, ISSIP, partnered with the Pennsylvania State University to research and explore the creation of digital twins to help their volunteers be more productive and able to have high quality interactions with more ISSIP community members. ISSIP is a non-profit that focuses on enhancing the global service economy for both businesses and individuals. A Digital Twin is a digital representation of a physical object, person, or process. PROBLEM STATEMENT A recent challenge has surfaced where ISSIP board members occasionally find themselves unable to participate in scheduled panel discussions or meetings due to conflicting obligations. How can AI technology be harnessed to create a digital twin of a potential panel speaker than can take the place of the speaker if they are longer able to attend? OBJECTIVES Develop an AI-driven digital twin capable of generating and delivering a panel talk presentation based on a research paper or notes. Produce a white paper aimed at educating ISSIP members on leveraging AI technologies in their daily lives, while addressing associated ethical considerations. KEY REQUIREMENTS The white paper must analyze three professions from the O*NET and explore which specific tasks can be aided or completed by a digital twin. The white paper and simulation must be accessible to the ISSIP community with respect to language and content. The digital twins shall be able to give talks, summarize a document, or generate a paper based on background information from the subject. DIGITAL TWINS FOR PEOPLE IN BUSINESS AND SOCIETAL ROLES Sponsor: Jim Spohrer Advisors: Vittaldas Prabhu, Anshul Balamwar Alumni Coach: Toni Rae Bryn Goldman, Natalie Grim, Gonzalo Rambla, Clare Relihan, Shakeb Siddiqui
APPROACH An iterative process to determine the optimal simulation deliverable. Initially, we used a weighted decision matrix to compare options between a Simio simulation and a panel talk simulation. Subsequently, utilizing a morphological matrix, we brainstormed potential technologies for each component of the simulation: audio, video, and text. The resulting simulation structure is illustrated to the right. Upon thorough examination of the matrix and further research, the team discovered additional technologies. They concluded that consolidating audio and video functionalities into one platform might be more feasible. We used a weighted decision matrix when considering which APIs and technologies to incorporate into the simulation. PROTOTYPES HeyGen: Voice and Video Generation The Alpha prototype aimed to assess HeyGen's suitability for image and voice generation. HeyGen is a tool generating lifelike avatars from video samples and text. This prototype validated HeyGen’s functionality for the final simulation . GPT-3: Presentation Script Generation The Beta prototype focused on evaluating the GPT-3 API's efficacy in converting research papers into presentation scripts. Developed on a Next.js website, the prototype integrated ChatGPT-3 API for text input and prompt engineering. This prototype offered valuable insights into GPT-3 API's viability for research paper to speech functionality and Next.js platform suitability for hosting the final product. CHALLENGES A significant challenge arose when the technology tested in the Alpha prototype, HeyGen, became too expensive. This necessitated a substitution, leading to the incorporation of an alternative platform, D-ID, into the final prototype. Another challenge arose during the development of the simulation as each API has its own unique set of technical intricacies, causing the final result to be a work in progress demo that, in the future, could be developed into a deployable solution. FINAL SIMULATION The panel member simulation employs GPT-3, D-ID, and Next.Js to create a demonstration of how ISSIP member can use AI technologies to create a panel talk from a research paper.
BENEFITS Increased accessibility to AI technology for ISSIP members Ability to hold panel talk despite extenuating circumstances and changing availability. An analysis of how all types of careers that ISSIP members have can benefit from AI, FUTURE WORK While the project has achieved success, it faces limitations including reliance on external AI technologies and potential accessibility barriers. Ethical considerations surrounding AI usage, particularly in creating lifelike avatars, require ongoing evaluation. Future work includes integrating emerging AI technologies, addressing accessibility concerns, and developing ethical frameworks. User feedback will drive iterative refinement to meet evolving needs. ETHICAL CONSIDERATIONS The integration of digital twins in the ISSIP community raises ethical concerns regarding privacy, bias, and ownership. Obtaining consent for data collection is crucial to uphold ethical standards and prevent legal issues. ISSIP must establish protocols for data management and security to safeguard sensitive information. Clear guidelines on rights, ownership, and responsible usage are essential to ensure ethical deployment of digital twins while balancing innovation with individual rights protection. FINAL WHITE PAPER Through O*NET, the team was able to investigate some of the key activities that a wide variety of jobs entail and areas AI digital twins could assist in. When researching the different occupations, it was important to take a wide range of professions that being technical, non-technical, entry-level, and more. While many ISSIP community members hold many technical, managerial, and senior roles, it is important to still investigate how digital twins can help many entry level jobs. COST The main goal was to provide research and an understanding that does not require any outside costs. This is because the company and sponsor we are supporting throughout the semester is a non-profit. The project was successful in this regard and spent none of the allotted budget . VERIFICATION A nalyze three professions from the O*NET The white paper and simulation must be accessible to the ISSIP community The digital twins shall be able to give talks, summarize a document, or generate a paper based on background information from the subject.