Academic Technology gave several presentations during the AI symposium. Thank you to everyone who attended. We also understand it is a difficult time for faculty and students to attend. Below is a brief synopsis of our presentations and notes. James and Todd are both happy to meet with anyone if you have questions or see something of interest.
AI as a Partner: Prompt Engineering
The presentation gave a brief overview of how a large language model works which then lead to the importance of building a prompt. Most users only prompt with one or two sentences, but the output can be greatly improved by providing more context and detail for the desired output. The key parts to remember:
- Your role – who are you?
- Context – what is the background for this query?
- Information – what other information do you have that may be helpful? Upload additional files, provide examples, specific details about your situation, etc.
- Output – be as clear and specific as possible about the output you’re seeking. Note, AI models will understand ###Markup Language###, so you can provide a template for the format of your desired output.
The presentation concluded with examples of using AI in ways you may not have considered including brainstorming, critical feedback and simulations.
James’ presentation slides for Prompt Engineering
NotebookLM
We started the presentation with the definition of a RAG, Retrieval Augmented Generation. A RAG is a method to tell an AI that you want the answer to your query to come from your own set of documents that you have provided. In business, they may use a RAG to query their own financials, ad campaigns, etc. In academia it may be a collection of articles, primary resources, textbooks, or a combination of all the above. Google’s NotebookLM was given as the example for everyone to try. It provides linked citations within its content to your uploaded texts, and allows you to query all of your texts or just a sub-section. Other additional functions include the creation of study guides, audio summaries, and visual mind maps.
Todd’s Presentation on RAGs and NotebookLM
Deep Research
Agents are new capability provided in recent AI releases from OpenAI, Google, and Anthropic. Agents allow the AI to perform a multistep process. Before we would query an AI, and it could only provide a response based on the text within its model. With “Deep Research” the AI may first ask you follow-up questions about your research question, create a plan that you can edit for finding and summarizing research, search the open web for sources, and then compile a report. We used Google’s Deep Research, which is free at the moment, and gave examples for producing a literature review, help in choosing a thesis topic, and a summary overview of a question in an unfamiliar field.
Todd’s Presentation on Deep Research and other AI research tools
Foreign Language Chatbots
We are getting close to the point where we can provide personalized one-on-one authentic communication to our students in the target language. We currently have the ability for our beginning and intermediate students to chat with a bot via text as part of a scenario or as a more structured Q&A. We have used them in German, Spanish, Japanese and now Chinese for students to review conversational activities from class and prepare for oral exams. We hope to be able to provide speaking and listening practice in the fall.