This week's AI Nexus newsletter briefly highlights AI ethics in education, recent AI policy initiatives, and AI speech applications. AI Nexus is a resource from the Office of Academic Affairs focused on artificial intelligence at Regent University and beyond.
Perspective: AI Ethics in Education
During last weekend's Intellect Horizon conference, Dr. Bill Brown of the School of Communication & the Arts delivered an insightful presentation entitled "Navigating the Treacherous Waters of AI Ethics in Education." Dr. Brown approached the use of AI in education from an applied ethics perspective and identified challenges using research into formal & informal learning, mediatization, and stories. He then proposed a set of ethical principles for the use of AI in education and learning with the context of promoting human flourishing.
Watch: Navigating the Treacherous Waters of AI Ethics in Education
Policy: State & Federal Artificial Intelligence Declarations
Earlier this week the Biden administration released an Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence. It contains a wide-ranging set of initiatives including regulatory oversight, standards development, and priority statements. Regarding education, it includes the following call: "Shape AI’s potential to transform education by creating resources to support educators deploying AI-enabled educational tools, such as personalized tutoring in schools."
Last month the Youngkin administration released an Executive Directive on Recognizing the Risks and Seizing the Opportunities of Artificial Intelligence. This is focused around four areas: legal & regulatory environment, education & workforce development, modernization of state government, and economic development & job creation. Education received significant attention including examining the potential use of AI tools for personalized tutoring, embedding AI topics in higher education coursework, and supporting collaborative AI research and training.
Prototype: AI Speech Applications
Based on last week's multilingual custom neural voice prototype, numerous people reached out to learn more about the underlying technology. There's a growing industry of tools that enable transcription and translation of audio and video content. These include open source applications like OpenAI's Whisper automatic speech recognition system and commercial tools such as Google Cloud's Speech-to-Text and Text-to-Speech APIs. Last week's prototype was developed using a combination of tools including Whisper, Microsoft's Custom Neural Voice, and GPT-4.
If you're interested in experimenting with AI speech capabilities without writing code, consider test-driving the online tools from ElevenLabs including speech synthesis, voice cloning, and video dubbing. Please note that this is not a university-supported product; ElevenLabs is a VC-funded company that offers a popular tool in the AI speech space. ElevenLabs offers a free plan for hobbyists who wish to try out their services and provides an accessible means of learning how AI can be used with audio and video content.
Stay tuned for additional AI learning and development opportunities. Please feel free to reach out if you're ready to engage now.
Let’s use technology for good!
Jason D. Baker, Ph.D.
AVP & Senior Technology Strategist
Office of Academic Affairs
AI Nexus — Artificial Intelligence for Academic Innovation
This week's AI Nexus newsletter spotlights a video introduction to ChatGPT for faculty, highlights AI plagiarism tools, and relights last week's AI and transformational coaching presentation. AI Nexus is a resource from the Office of Academic Affairs focused on artificial intelligence at Regent University and beyond.
Spotlight: How ChatGPT Can Help You
Drs. Katie Goldman and Mitzi Fehl-Seward from the School of Education have created an excellent faculty-focused video entitled How ChatGPT Can Help You. In this practical workshop, Dr. Fehl-Seward engages with ChatGPT and demonstrates how faculty can prompt the leading generative AI tool to summarize articles, create quizzes, draft feedback, and more.
Watch: How ChatGPT Can Help You
Highlight: AI Plagiarism Tools
Students have been active users of ChatGPT since its release a year ago, raising concerns about how to flag content that has been written by AI. At present, there is no definitive way of identifying AI-written content; however, there are tools that estimate the probability that the content was generated by AI. As part of our transition to Canvas, Regent adopted Turnitin as our plagiarism checker, which includes AI writing detection. If you enable plagiarism detection in a Canvas assignment, and the submission is at least 300 words long, Turnitin will generate an AI writing detection report alongside the traditional plagiarism review. For shorter text, the GPTZero website hosts a free online AI detection form for submissions up to 5000 characters long, along with a Chrome extension. Please remember that AI detection tools can be helpful but cannot be considered authoritative.
Learn: Using Turnitin with Canvas
Relight: AI and Transformational Coaching
Last week's AI Nexus newsletter showcased Dr. Joseph Umidi's video presentation entitled AI Meets Transformational Coaching & Consulting. Since the associated slides don't appear in the video, we wanted to share them here.
Jason D. Baker, Ph.D.
AVP & Senior Technology Strategist
Office of Academic Affairs
AI Nexus — Artificial Intelligence for Academic Innovation
AI Nexus is a resource from the Office of Academic Affairs focused on artificial intelligence at Regent University and beyond. With the organizational turmoil that enveloped OpenAI over the past few weeks, and the associated media coverage that accompanied the firing and rehiring of its CEO, it's been easy to miss recent advances in the field. Therefore, this week's AI Nexus newsletter highlights four initiatives from across the industry of interest to the Regent community: video generation, African language models, materials discovery, and academic peer review.
Generating Video with AI
AI image generators such as Stable Diffusion, DALL-E, and Midjourney have received a lot of attention in the past year. These tools use generative AI to convert text prompts into images, from sketches and cartoons to paintings and photo-realistic pictures. Recently Stability.AI, the company behind Stable Diffusion, built upon these efforts and released a new video generation tool called Stable Video Diffusion. Users provide a starter image, then the AI system automatically generates a corresponding video clip, with corresponding text-to-video capability coming soon.
Building AI for African Languages
Since large language models are central to the latest developments in generative AI, it's important to train models for languages other than English. Two organizations that have responded to this challenge include Lelapa AI and Lesan. Lelapa AI has developed Vulavula, which can transcribe, analyze, and converse in Afrikaans, isiZulu, Sesotho, and English, four languages spoken in South Africa. Similarly, Lesan has developed translation tools focused on Ethiopia, with their online translation page currently supporting Amharic, Tigrinya, and English. MIT Technology Review has more on these and other efforts to develop AI models for African languages.
AI-Powered Materials Discovery
Google DeepMind has developed a machine-learning AI tool called Graph Networks for Materials Exploration (GNoME) that predicts new chemical structures. Using knowledge of existing crystals, DeepMind's GNoME generated over two million previously unknown crystal structures, with researchers at the University of California, Berkeley and elsewhere subsequently synthesizing over 700 of them to-date. The Economist summarizes the discovery, and the associated research article is available in the scientific journal Nature.
Researching AI Peer Review
Researchers at Stanford leveraged a database of peer reviewed scientific manuscripts, featuring nearly 5,000 articles submitted to Nature, its sub-journals, and the International Conference on Learning Representations, along with associated editorial comments, to compare human and AI generated peer review feedback using GPT-4. They found similar levels of overlapping comments between the human and AI feedback systems compared to multiple human reviewers, as well as generally positive responses from authors who evaluated the quality of the feedback. The researchers conclude that such applications could be a helpful means for authors to pre-review their manuscripts prior to journal submission. The research article can be read at the arXiv repository.
Jason D. Baker, Ph.D.
AVP & Senior Technology Strategist
Office of Academic Affairs
AI Nexus — Artificial Intelligence for Academic Innovation
Where is artificial intelligence heading in 2024? This edition of the AI Nexus newsletter highlights four expectations for the year ahead from leading industry watchers: AI for everything, multimodal chatbots, autonomous agents, and small language models.
AI for Everything
MIT Technology Review effectively summarizes overall industry sentiment by naming "AI for Everything" as one of their 10 Breakthrough Technologies 2024. Tech companies are rushing to embed AI in popular consumer devices and enterprise software, from Samsung Galaxy smartphones to Microsoft Windows and Office applications. To quote from their commentary, "Never has such radical new technology gone from experimental prototype to consumer product so fast and at such scale."
Multimodal Chatbots
The New York Times reports executives from OpenAI and Google agreeing that we'll see a huge leap forward in chatbot capabilities in the year to come. One key advance is the rise of multimodal chatbot capabilities. While previous chatbots were primarily text-based, more recent efforts have trained models with text, images, audio, and video, enabling the AI to handle multiple media simultaneously. This enables an AI system to analyze relationships between multiple types of content and generate more robust output as well.
Autonomous Agents
Boston Consulting Group highlights the emergence of autonomous agents, which are intelligent task completion systems (and the subject of an interesting discussion at last week's CES 2024). When given an objective, autonomous agents can identify, prioritize, and complete tasks independently in order to achieve a desired outcome. BCG offers the following metaphor: "Autonomous agents are, in effect, dynamic systems that can both sense and act on their environment. In other words, with stand-alone LLMs, you have access to a powerful brain; autonomous agents add arms and legs."
Small Language Models
Forbes includes small language models in its list of five generative AI trends in 2024. Whereas ChatGPT, Bard, and other popular tools were built with large language models (LLMs) trained on large swaths of the internet, small language models (SLMs) are developed with much less data. One approach is to build small language models using curated datasets containing in-depth information from a particular topic or domain, in part to reduce hallucinations and other errors often found in LLMs. As noted in Forbes, "Enterprises will be able to fine-tune SLMs so that they can be tailored to specific tasks and domain-specific functions. This will meet the legal and regulatory requirements, which will accelerate the adoption of language models."
Jason D. Baker, Ph.D.
AVP & Senior Technology Strategist
Office of Academic Affairs