Chatbots and Beyond: A Guide about AI Technologies

Hello Everyone! I'm Alexis Pavenick, PhD, MLIS. I am an Associate Librarian for CSULB, and I currently cover Literatures and Languages. I have written this guide through interest and research about ChatGPT - I'm a little bit of a tech geek now that I am a librarian. I share with you here information my colleagues and I have found about the topic.

While I am by no means an expert in AI Chatbots, I do have 16+ years experience as an English adjunct lecturer for Cal Poly Pomona, Los Angeles City College, Glendale Community College in the LA Area and Bronx Community College, The Art Institute of New York City, and ASA College in New York City. I am quite familiar with assigning and grading long and short essays and dealing with plagiarism. My PhD is in English Literature from UC Riverside, so I am also familiar with writing essays in literature, as well as in social science, as I have a MPhil in Anthropology from Cambridge. 

Please take my suggestions of ways to respond to concerns about ChatGPT and other chatbot usage as my own inspiration of approaches I might take in teaching and assigning tasks to assess writing and critical thinking skills.

Select HERE to find my other library guides, and Select HERE to see all the guides of our CSULB Librarians.

The technology company Oracle defines a chatbot as follows, "At the most basic level, a chatbot is a computer program that simulates and processes human conversation (either written or spoken), allowing humans to interact with digital devices as if they were communicating with a real person. Chatbots can be as simple as rudimentary programs that answer a simple query with a single-line response, or as sophisticated as digital assistants that learn and evolve to deliver increasing levels of personalization as they gather and process information."

• Chatbots use Natural Language Processing (NLP) and other machine-learning artificial intelligence algorithms to create their output. IBM defines NLP as "the branch of computer science—and more specifically, the branch of artificial intelligence or AI—concerned with giving computers the ability to understand text and spoken words in much the same way human beings can."


• The purpose of many chatbots we encounter is to save an institution or company time and money by answering common questions. The widespread use of chatbots suggests that marketing strategies believe interacting with a human-like text or voice experience appeals to people more than reading a Frequently Asked Questions page.


• Chatbots are also available as subscription-based text creators; these are often used by people who write for blogs and sales. If you search the Internet with phrases such as: "subscription chatbots for blogging," you will find many companies who offer text-creation services. This has been going on for some time. 

ChatGPT works in a similar fashion to all chatbots. It uses NLP and other algorithms to examine user inputs and respond with what its software engineers consider useful and apt responses. ChatGPT stands out for two main reasons: 1) The currently free access to using the bot and 2) Its billion-word corpus, which increases with each of the bot's iterations. The enormous scope of Internet data ChatGPT pulls from is part of its "large language model" programming. It is this scope that, in large part, gives ChatGPT's results such a natural tone and potential for accuracy.

• ChatGPT is based on OpenAI's GPT (Generative Pre-trained Transformer) language model, which has been trained and will continue to train on an ever-increasing corpus of text data in order to generate human-like responses to text-based prompts.


• How it works:
  • When a user types a query into ChatGPT, the program analyzes the words by examining their order and content meaning.
  • It then searches its data for similar words and content meaning.
  • To answer the query, ChatGPT predicts the most typical words that have followed each other in the past, in relation to the text and content of the query.
    • It also chooses intermittent stages to insert a less-typical word, which gives the tone of the writing its more human-like, if bland, quality.
  • If you want to see a consideration of how this is developed through code, Stephen Wolfram's article "What is ChatGPT Doing . . . and Why Does It Work?" has a good explanation.


• OpenAI is aware of the concerns educators have about the use of ChatGPT by students. They have created an Educator Considerations page which explains their philosophy and suggestions for use.

Microsoft Copilot is an advanced AI assistant designed to enhance productivity and creativity across various Microsoft products. It leverages the power of large language models (LLMs) to assist users with a wide range of tasks, including:

• Answering Questions: Copilot can provide detailed answers to queries, helping users find information quickly and efficiently.
• Generating Text: Whether it’s drafting emails, writing reports, or creating content, Copilot can generate coherent and contextually appropriate text.
• Creating Images: Users can describe the image they need, and Copilot can generate it based on the description.
• Summarizing Content: Copilot can condense long documents or articles into concise summaries, making it easier to grasp key points.
• Web Search Capabilities: Copilot can perform web searches to provide up-to-date information and answers. By leveraging Bing’s search engine, Copilot can access a vast array of public data, ensuring users receive accurate and current information. This feature enhances Copilot’s ability to assist with research, fact-checking, and discovering new insights.

Copilot is integrated into Microsoft 365 apps like Word, Excel, PowerPoint, and Outlook, as well as in Windows and Microsoft Edge. This integration allows users to access Copilot’s capabilities seamlessly within the tools they use daily.

For the version available to CSULB students, faculty, staff, it’s important to note that their interactions with Copilot are kept private. The questions they ask and the answers they receive are not included in the training data or corpus used to improve the AI. This ensures that their data remains confidential and is not used to refine or develop the AI further.

In other words, CSULB's contract with Microsoft includes the structure that questions and answers made by CSULB students, faculty, and staff in CSULB's version of Copilot are not added to the Copilot corpus nor are they remembered in the person's account. So, when using Copilot, it will not remember what you have asked or what it has answered once you have closed the session.

MS Copilot also exists in various levels for corporate industry.

On 3/17/23, a popular independent editorial book site called Book Riot, "ASKED CHATGPT WHY BOOKS SHOULD BE BANNED."

They found ChatGPT's answer, "a better defender of the First Amendment rights of all than Moms For Liberty, No Left Turn in Education, Utah Parents United, and all of the other book banning groups."

For liberal-minded educators, this is good news. 

The thing to consider about this, however, is that ChatGPT is a product. It and AiChatbots like it will eventually be available for sale, making their language and content slant customizable. Currently, OpenAi/Microsoft are controlling ChatGPT's answers in a liberally-minded way. This will change depending on the interests of the people using the tool. See The New York Times* March 22, 2023 article, "Conservatives Aim to Build a Chatbot of Their Own."

Once again, this "AI" is still controlled by humans. Please remember this as it moves forward in development.

State of the Art - Basic information about Chatbots as of 9/09/2024.

Currently Popular AI Chatbots

Copilot (in your CSULB Tiles and when you signing CSULB's MS Edge) | Functionally owned by Microsoft, CSULB’s access to Copilot is termed “secure” in that our Copilot will not record questions or answers. So inputs are not collected for the corpus of information drawn on for answers.

Grammarly (https://www.grammarly.com/) | Independent Company – MS Office & Google now offer similar support in their office products; all of them will rewrite entire sections if asked

Chat GPT (https://chatgpt.com/) | Open AI – partnered with Microsoft; Chat GPT is currently the most successful chatbot (though the ranks are always changing). It’s the backbone of MS Copilot.

Claude (https://claude.ai/) | Anthropic – meant to be a more ethical chatbot

Gemini (https://gemini.google.com/app) | Google – used in Google Searches     

Perplexity (https://www.perplexity.ai/) | Privately owned: Aravind Srinivas & Denis Yarats are the primary owners. Yarats, the CTO, was an AI research scientist at Meta; Srinivas, the CEO, worked at OpenAI as an AI researcher. Perplexity works like a refined, AI-assisted Google Scholar.

Elict (https://elicit.com/) and Scite (https://scite.ai/) | Independent Companies; used to analyze research papers/create bibliographies. Perplexity will do some of this work as well.

Popular Chatbots that focus on writing that sounds less robotic: Claude and QuillBot (https://quillbot.com/)

Please note that CSULB University Library has access to The New York Times and many other newspapers, journals, and periodicals. To find and access what we have, on the homepage of the Library, underneath the OneSearch box, select Advanced Search. Then, at the top black ribbon, select Journal Search. Enter the title of the periodical, e.g. The New York Times, and you will see whether we have the item and the length of the run we have access to. If we do not have what you are looking for, please use Interlibrary Loan to request it. 

Here are some common concerns educators have about ChatGPT and other LLM AI ChatBots:

1. Cheating: People may use AI chatbots to generate written content for assignments, potentially engaging in academic dishonesty and plagiarism that can be challenging to detect.
2. Quality of information & bias: ChatGPT's current billion-word corpus of data is compiled from text it finds by searching the Internet. We know that information on the Internet can be incorrect, biased, discriminatory, outdated, or fake, among its problems. While ChatGPT does have some safeguards and restrictions on what kinds of answers it will return, it is not infallible, nor can it be relied upon to offer correct information; it is often wrong.
3. Undermines critical thinking: ChatGPT's language reads and sounds like human speech. Thus, users may be more prone to accepting its results without critical analysis. 
4. Undermines problem-solving: As ChatGPT gives full and detailed answers, users may become less interested in learning and practicing the skills of research and problem-solving. 
5. Threatens privacy: As with any interaction on the Internet, users may input personal information during conversations. Any data put into any chatbot is likely collected, and how this information is stored or used is difficult to discern. 
6. Career concerns: People are also concerned that their jobs and careers may be at risk if software like ChatGPT can effectively impersonate human thinking in writing and speaking.

AI Technology at CSULB Homepage

• Includes the CSULB AI Steering Committee resource pages for Faculty, including sample Syllabus statements
• Includes Student Guidelines for the use of AI at CSULB

Introduction to AI Tools for Teaching and Learning

AI for Teaching and Learning

Exploring AI Pedagogy: A Community Collection of Teaching Reflections

An initiative of the MLA-CCCC Joint Task Force on AI and Writing

Modern Language Association (MLA) | Conference on College Composition and Communication (CCCC, aka The 4 Cs)

Here is a summary of actions educators may take to engage with AI Chatbot and/or mitigate concerns about it:

1. Use AI Chatbots for critical thinking activities: Ask an AI Chatbot a question (or have it write a short essay) and then have students evaluate and try to verify the results. This is a good activity to develop skills of evaluation, especially of a result they would like to trust. It also gives students a chance to "peer-review" an artificial student, rather than their classmates. 


2. Use AI Chatbots as an outlining tool: One of an AI Chatbot's greatest abilities is to make sense of disorganized notes.
   • Students can discuss ideas in a group or write them down in class. Together, they can input their ideas into an AI Chatbot and ask it to create a bullet-point outline for them. In this way, students can see AI Chatbots as a tool to help them clarify their thoughts, not just a source of answers. 
   
   
   • As well, educators can ask AI Chatbots to make lesson plans or create class activities this way. Remember, AI Chatbots pull information from the Internet, and there are huge numbers of shared lessons and activities out there.


3. Use AI Chatbots as a discussion topic: Explore an AI Chatbot's capabilities in class in relation to class content and discuss the ethical issues that its use creates. This is a way into topics such as ethics, responsible use, and privacy.
   • Remind students that, like Social Media, commercial AI Chatbots collect their ideas and information. For example, anything anyone enters into ChatGPT can become part of its dataset. This is NOT the case when we use the CSULB AI Chatbot, Copilot (see above: What is CSULB's Copilot?)


4. Stay informed: Stay informed about the latest developments in ChatGPT and related technologies, to be aware of the potential impact on student learning and society. See below for news articles and links to AI tech blogs.


5. Ask your CSULB Librarian to give you resources or speak to your class about Information Literacy and Digital Literacy to help students explore critical thinking and evaluation techniques when using Internet and library resources.

Consider the tried-and-true method of the composition educator: 

• Writing in class throughout the semester
  • Begin the semester with a diagnostic essay to get a baseline of student writing ability and basic knowledge of your topics.
    • Diagnostic essays are typically general/big idea questions about what students may or may not know about the topics to be covered in class, what they expect or hope for from the class, what barriers they may be concerned about.
  • Short assignments, written in class, add to that initial baseline
    • Any class time used to write and revise will help educators (and students) understand the students' current and improving skill levels.  
    • Testing in short bursts in between discussion and in-class work can have a similar effect.
• Online synchronous classes can require short oral presentations without notes or slides. These can be live or recorded.
• Asynchronous classes can request alternative assignment structures, such as (but not limited to) image collages, mind-maps, personal reflections as well as process reflections. peer-interviews and group assignments.

To keep aware and updated about the developments of ChatGPT and other Artificial Intelligence technologies, consider subscribing to the Tech blogs of CNET, MIT News, Stanford's Machine Learning, and UC Berkeley's AI Blog and their AI Department.

For information about the dynamic changes in leadership at OpenAI and other tech industry topics, I suggest The New York Times podcast Hard Fork.

Below are some articles I have collected since OpenAI's Chat GPT was launched in November 2022. Many of them are outdated now because AI development moves so quickly. Please use the resources above to keep aware and updated.

Here is a Glossary of Terms related to Generative AI offered by a prompt writing company called AI Prompt Marketplace & Prompt Engineering Community.

A big thank you to Amelia, a student researcher, along with her mom, for suggesting this list, which they found while investigating the benefits and concerns of using AI technology.