CMU-Qatar Library

"Previous AI helped curate our world [...], but GPT AI will allow you to create your world" (Bowen & Watson, 2024, p. 26).

Created using GPT-4o

The Development of GenAI Models

ChatGPT is an AI model that falls under the category of deep learning, a subfield of machine learning and artificial intelligence. It is designed to generate text in a conversational style by learning complex patterns from input data, without being explicitly programmed for each task. ChatGPT's neural network consists of billions of parameters and mathematical operations that determine the probabilities of which words come next, rather than being a database of pre-written sentences.

This video by the University of Arizona Libraries provides a quick summary of the technology behind ChatGPT, and explains basic terms such as narrowAI, deep learning, and neural networks.

Training GenAI Models

Earlier language processing techniques had significant drawbacks that limited their effectiveness. The transformers architecture, introduced in 2017, overcame these challenges by introducing an attention mechanism, enabling AI to focus on the most important elements of text. This led to the development of large language models (LLMs), like ChatGPT, which are trained on vast amounts of data and can generalize to new domains. ChatGPT's training process involves reinforcement learning from human feedback (RLHF), allowing it to generate more conversational and human-like responses.

The next video by the University of Arizona Libraries provides an overview of how GenAI models are being trained and explains some of the terminology used when talking about GenAI such as transformers, large language models (LLMs), and reinforcement learning from human feedback (RLHF).

How can we Verify Information Generated by GenAI Models?

"AI driven misinformation is already flowing through social networks in ways that are difficult to detect and deal with" (Mollick, 2024, p. xviii). 

When using GenAI models, it is crucial to verify the accuracy of the generated content by following the links and ensuring that the AI-generated summary aligns with the original content and is relevant to the given task. To further mitigate the risk of misinformation, even when using websites as sources, employing the SIFT Method (Stop, Investigate the source, Find better coverage, and Trace claims to the original context) is recommended.

To What Extent Can you Use GenAI in your College Work?

When considering the use of GenAI in academic assignments, several factors come into play, including the assignment type, topic, required resources, and your professor's guidelines. Professors may have varying policies:

• Some may restrict GenAI use, requiring original work.
• Others might allow limited use, provided you acknowledge it.
• Some may design assignments that extensively incorporate GenAI.

Always refer to your course's academic integrity policy and consult your professor for clarity before starting your work. GenAI is a rapidly evolving tool that will likely be valuable in your future workplace, making AI literacy essential. However, it is crucial to continue to develop human cognitive, social, and emotional skills by engaging with diverse academic opportunities that still require human insight and interaction.

What AI Tools are Available?

Whether you need to generate some text, images, music, videos, or work on a research project, or transcribe and analyze some data, here are few examples of tools that you can try.

Exercise caution when AI tools. Always review your prompts and outputs carefully to ensure you do not share any private, sensitive, or copyrighted information.

*Gemini and Copilot subscriptions are available through CMU. Log in using your Andrew ID and password.

**Scite is a database available to you through CMU Libraries. You can access it through our Databases, eResources, & Tools page.