Key Points from the Introduction:

1. History of Language Models:

Language models are not new; they've been used in various applications,
from autofill in texting to generating fake news in elections.
2. ChatGPT Launch:
OpenAI released ChatGPT to the public on November 30, 2022, causing a
significant stir in schools and libraries.
3. Reasons for ChatGPT's Impact:
o Ease of Use: ChatGPT responds instantly to natural language
prompts.
o Human-like Output: It generates coherent and believable content,
making it seem like human-crafted text.
4. Concerns in Education:
Schools worry about undetectable cheating, and libraries fear AI could
replace essential services.
5. Practical Applications:
Real estate agents quickly adopted ChatGPT, using it to save time on tasks
like writing property listings.
6. Growing Body of Knowledge:
Although there’s limited academic literature on ChatGPT, preprints, news
articles, and discussions are rapidly expanding knowledge of this technology.
7. AI's Future in Academia:
ChatGPT and similar tools are improving and becoming more widespread. The
academic community and libraries must adapt by understanding their uses
and limitations, acknowledging ethical concerns, and recognizing that these
technologies cannot fully replace human expertise, at least for now.

Key Points for Presentation - The Basics of ChatGPT:

1. Chatbot Functionality:
ChatGPT is an AI-driven chatbot that generates responses by predicting
words sequentially, forming coherent answers to user prompts.
2. Underlying Technology - GPT-3:
o GPT-3 is the model behind ChatGPT, trained on an enormous dataset,
including the entire internet, 19 billion web texts, and 67 billion books,
creating a corpus of around 500 billion words.
 o It utilizes 175 billion parameters, enhancing the accuracy of its
language predictions and understanding.
3. Scalability and Versatility:
The scale of GPT-3 allows ChatGPT to handle diverse tasks, such as:
o Writing complex paragraphs in multiple languages.

o Generating computer code in unfamiliar libraries.

o Solving algebraic problems, creating recipes, and writing essays on

various topics.
4. Real-world Applications:
o In education, students used ChatGPT to develop startup prototypes
faster than usual.
o It is efficient in producing structured responses in areas such as coding,
academic writing, and even practical everyday tasks like cooking
recipes.
5. Creative Uses:
ChatGPT can generate creative outputs, such as scripts, including a playful
example where Dolly Parton introduces "information literacy" in a humorous
and accessible way.
6. Broad Impact:
Its ability to adapt to various languages and domains makes ChatGPT a
valuable tool for both technical tasks and creative storytelling.

The Drawbacks of ChatGPT

1. Accuracy Issues:
o ChatGPT generates inaccurate information around 20% of the time
("hallucinations").
o It cannot cite sources, making verification difficult.
o Longer responses often lack coherence or contain contradictions.
2. Limitations in Coding:
o Effective with simple code but unreliable with complex coding tasks.
o Banned from Stack Overflow due to inaccuracies in code generation.
3. Outdated Knowledge:
o ChatGPT's training data only goes up to 2021, making it unable to address current
events.
4. Bias Concerns:
o Reflects biases from its training dataset, potentially underrepresenting
marginalized views.
5. Computational Costs:
o ChatGPT is expensive to operate, costing about $100,000 daily during the free
research preview phase.
 o A subscription plan ($20/month) has been introduced, with future business
models likely.
6. Privacy Concerns:
o As a free version may not be sustainable, there are concerns about monetizing
users' private data.

Slide Title: Copyright Concerns with AI-generated Content

1. Human Authorship Requirement:

o Copyright typically applies only to works created by humans.
o U.S. Copyright Office rules: AI-generated content without human creative input is
not eligible for copyright.
2. Notable Case:
o Naruto v. Slater (2011): Monkey selfies were ruled non-copyrightable due to non-
human authorship.
3. AI-generated Works:
o AI content is often created automatically, raising questions about its copyright
eligibility.
o Human creative input (e.g., framing prompts) introduces a gray area for
copyrightability.
4. Global Copyright Variations:
o Laws vary by country; European and UK laws may offer more flexibility for AI
works.
o International conventions like the Berne Convention lack clear definitions of
authorship for AI-generated content.
5. Ownership Uncertainty:
o No clear rules exist on who would own the copyright of AI-generated works.
6. Future Legal Challenges:
o As AI usage grows, these issues will need to be addressed in law and academia.

Slide Title: ChatGPT in Academia

1. Authorship and Academic Integrity:

o Concerns about students using ChatGPT for cheating.
o Uncertainty about AI's role as a co-author or collaborator.
o Questions about the value of learning if AI replaces writing processes.
2. Challenges in Detecting AI-generated Text:
o Traditional plagiarism detectors can't reliably identify AI-generated text.
o New AI text detectors (e.g., GPTZero, AI Text Classifier) can detect AI-generated
content with up to 99% accuracy.
3. Innovative Use of ChatGPT:
o Can assist with brainstorming, outlining, revising, and proofreading.
o Encourages critical thinking and information literacy by evaluating AI-generated
responses.
 4. Educational Implications:
o Educators are adapting assignment design to incorporate AI responsibly (e.g.,
scaffolding, peer reviews).
o Instructors and librarians can teach students to engage critically with AI for
enhanced learning outcomes.
5. Balancing AI and Pedagogy:
o Engaging with ChatGPT strategically can enhance learning, like using a
calculator.
o Banning AI use may be impractical but leveraging it can enrich student education.

Slide Title: ChatGPT in Libraries

1. Impact on Information Literacy:

o ChatGPT supports digital fluency and information literacy, key roles for
librarians.
o Useful for answering frequently asked questions and providing basic information.
2. Potential Benefits:
o Frees librarians for complex tasks.
o Can help with metadata creation, book recommendations, and answering
reference questions.
o Assists non-expert coders with tasks like writing code for system modifications.
3. Limitations:
o Inconsistent accuracy and inability to cite credible scholarly sources.
o Lacks personalized, nuanced responses like human librarians.
o Cannot access specific library holdings for book recommendations.
4. Risks:
o Employment, financial, and privacy risks associated with reliance on AI.
o Risk of providing biased or incomplete information.
5. Conclusion:
o While ChatGPT offers valuable assistance in certain areas, human oversight is
crucial to ensure accuracy and relevance in libraries.