According to the Center of Teaching Innovation at Cornell University (2023), generative artificial intelligence (AI) is "a subset of AI that utilizes learning models to create new, original content, such as images, text, or music, based on patterns and structures learned from existing data." These web tools use large language models (LLMs), algorithms, data, and statistical models to make inferences about what content should be generated based on prompts provided by end users. Generative AI predicts tokens (pieces of data) based on what it has been fed before. They are comparable to trained chatbots. For example, when an end user provides a prompt to an AI tool such as ChatGPT or Perplexity AI, it provides a text response based on the LLMs it has been trained on. The chatbot can then provide a text sequence of what could come next. 

The generative AI landscape is rapidly changing. Private and public industries, including state and local governments, are bringing to probe and test ways that generative AI can either improve or replace human-created work.  Generative AI tools can help users create content, gather background information on a topic, and synthesize information. However, due to how generative AI tools "learn" and lack of regulation, they can "hallucinate", or make up sources, facts, and information, and create biased and harmful content. There is some debate about whether or not the hallucination problem is fixable. While there are some ways to differentiate and spot AI-generated content from human-generated content, as generative AI becomes more sophisticated, it may become more difficult to differentiate real content from fabricated content, such as deepfakes.

There may be some benefits to incorporating text-generating AI tools. Below is a non-exhaustive list. What do you think are some other benefits of using generative AI?

• Use as a starting place for information, but it can't be relied on for factual information
• Simple text editing and help jumpstart some ideas to get you writing
• Prioritization and organization of work by creating schedules, outlines, and lists

While text-generating AI may have some benefits, it also has very serious limitations that users should be cautious of. Below are some of those limitations:

• Hallucinations: false or inaccurate statements output
• Biased and harmful output
• Use of unclear data sets: There is no mechanism to figure out what data sets AI tools pull from when making predictions. Some LLMs only have access to knowledge up to a given date. This may result in outdated and unchecked information. This issue is further compounded by some tools having a paywall. 
• It may make up or provide false citations. The works it cites may not exist.

Can you think of any other limitations that may be associated with generative AI?

While you can use generative AI in this class, you may be unable to in others. You should always check with your professor if they don't explicitly have an AI policy mentioned in their syllabus. 

When you use generative AI in your work, you must cite your AI usage. Failure to cite generative AI in your work would be considered plagiarism and violates the university's academic integrity code. There are also other ethical considerations such as copyright infringement and the collection of personal data. Can you think of any other ethical issues you should consider before using generative AI?

For this course, you should use Chicago-style citations. You can refer to this guide for more information on how to cite generative AI using the author-date (and notes) format or ask me for more help.

Below are some examples of text-generating AI tools. Generative AI is a rapidly growing industry so this list is not exhaustive. There may be additional tools you know of that work to varying degrees.

Andi

Bing Copilot

ChatAI ( requires login)

ChatGPT (OpenAI, requires login)

Claude (Anthropic, requires login)

Consensus

Gemini (Google, unavailable in UPS Google account as of 2/2024)

PerplexityAI

We'll practice using some generative AI tools for your upcoming writing assignment.

1. Pick a paper prompt.

2. Choose a Large Language Model (LLM) that you'd like to use, such as GPT (in any version), Claude, PerplexityAI, among others. If you have never used one, try this one.

3. Use the program to write a draft of your paper. HOW TO DO THAT IS UP TO YOU! You may have to play around with instructions. It might take you some time to get a successful outcome (by whatever standards you use to assess success). You can use these tips from Harvard University Information Technology as a guide on how to format your prompt in an AI tool. 

4. Paste the draft on Canvas and then comment on the process and the result: was it easy or hard to get a serviceable draft? How does this compare to the process of writing a paper without this type of tool? What is gained and what is lost?