Library Guides: Generative Artificial Intelligence : Strengths & Weaknesses

Strengths & Weaknesses

Generative AI is evolving quickly, and it’s crucial to remember that it has both strengths and weaknesses:

Strengths

Academic Support: Generative AI can assist with brainstorming essay topics, explaining complex subjects in simpler terms (e.g., coding), generating research terms, and creating essay outlines.
Consumer Use: These tools are excellent for creating travel itineraries, health and fitness plans, recipes, and more.
Design Purposes: Generative AI can be used for image creation and other design tasks.

Weaknesses

Outdated Information: The information provided may not be up-to-date with the latest events.
Fabrication: Generative AI often "hallucinates" or fabricates information, particularly with academic references or citations.
Prompt Knowledge: Producing the best results requires knowledge of effective prompting, which many users lack.
Bias and Language: The information presented can contain inherent biases and non-inclusive language. There is concern that the material used to train AI has inherent biases and the AI propagates this biased and sometimes harmful information in its output.
Topic Coverage: Generative AI has a better understanding of STEM topics than humanities.
Citation Issues: Some platforms do not include citations, leading to doubts about source validity. Even tools like Microsoft Copilot, which include citations, sometimes provide incorrect sources.
Environmental Impact: The energy consumption of AI servers is extremely high, which is harmful to the environment.
Academic Integrity: There are concerns about students passing off AI-generated work as their own, leading to academic integrity issues. As well, often the output from generative AI does not include citations or the citations are incorrect which means the generative AI is producing content without proper attribution. Generative AI output can also be difficult to cite as it is transitory.
Copyright Issues: There are copyright concerns such as using other artists' work without proper credit. The generative AI tool may be trained on copyright-protected content and then output that content to the user without the user knowing the content is copyright-protected.
Privacy and Intellectual Property: There are privacy concerns associated with using generative AI. For example, if you input private information, the generative AI tool can use that information to "learn" and the information becomes part of its knowledge-base. Similarly, if you input information that is protected by copyright, the generative AI tool can incorporate that intellectual property into its knowledge-base with no attribution to the creator.
Internet Dependency: An internet connection is required, which can be problematic in areas with connectivity issues. If the server goes down, the tool becomes inaccessible.
New Technology: Generative AI is still a new technology, and there are fears that it may become sentient in the future.
Harmful use: Generative AI can be used easily for creation of deepfake images or videos to spread misinformation.

Adapted from UCSanDiego Libraries:Generative Artificial Intelligence: Challenges and Possibilities of Generative AI

Artificial Intelligence Has a Problem With Gender and Racial Bias. Here’s How to Solve It
From Time, this article discusses the implicit bias issue with AI
Ethics of Artificial Intelligence
From UNESCO, recommendations on the ethics of artificial intelligence
Ethical Considerations for using Generative AI
From the University of Alberta's Using Generative AI Guide
Ethics in the Age of Generative AI
From LinkedInLearning, this course introduces the ethical concerns around generative AI, and then teaches how to use a framework of ethical analysis to address these concerns.

Synthetic images and media are created exclusively through the use of generative AI tools and do not require actual humans or any actual photography or video equipment.

A deepfake is a synthetic media, often a video or audio, that uses artificial intelligence (AI) to manipulate or generate realistic images, audio, or video content. The name deepfake comes from a portmanteau of “deep”, referring to deep learning, and “fake,” referring to the fact that the images generated are not genuine

Synthetic and Deepfake images and media pose as both potential threats and opportunities

Lyon, B., & Tora, M. (2023). Exploring deepfakes : deploy powerful AI techniques for face replacement and more with this comprehensive guide (1st ed.). Packt Publishing Ltd.

Exploring Deepfakes by Bryan Lyon; Matt Tora
Publication Date: 2023

This resource covers what deepfakes are and what makes them different from other machine-learning techniques.

Deepfakes: Can you spot them?
In this "Detect Fakes" experiment, by Northwestern's Kellogg School of Management, the user is presented with a series of images. These images may either be real photographs or AI-generated images.