An Overview of GenAI

Even if you have not yet jumped on the latest technology of generative AI, you have likely heard of ChatGPT and its exploding popularity. Companies like Google and Microsoft are already integrating AI into their day-to-day services, and an increasing number of job postings are requesting applicants who have experience with AI and prompt generation. Whether you use it or not, it is good to know how GenAI works and what it can be used for.

Please note: Many colleges are still deciding whether and to what extent it is acceptable to use generative AI for assignments and papers. Cerro Coso's academic integrity statement, for example, states that students are not allowed to have AI do their assignments or exams for them, and that they should disclose when and where they are using it in papers. (AI or not, you should always cite your sources!) However, Cerro Coso also leaves it up to each instructor to decide if and how AI may be used for their classes. If you're not sure if it's allowed, please ask!

What is Generative AI? 

Generative AI, such as ChatGPT, Microsoft Copilot, Google Gemini or Meta AI, has been described as "autocomplete on steroids." It is fed an enormous diet of documents, websites, videos, and other internet resources. When a user asks a question, it is programmed to generate a response based on statistically likely answers. However, while it may pull information from high quality resources like NASA and Scientific American, it may also pull information from Reddit and The Onion. Much like Wikipedia, AI may be a good place to start a research project, but we don't recommend ending there - Always verify and evaluate resources before you use them!

A few more things to keep in mind:

• GenAI is especially useful when there isn't a precise "right answer" - It is great for brainstorming topic ideas, summarizing your research notes, organizing data, and creating flashcards and quizzes.
• It is excellent for low-stakes questions, such as "What can I make for dinner using three eggs, BBQ sauce, and mushrooms?" (Answer: A BBQ mushroom omelette, apparently), but don't trust it with your life or your livelihood! 
• GenAI is trained on "everything," which includes misinformation and biases. Since its algorithm is based on statistics, it often amplifies negative stereotypes and removes complexity and diversity.
  • Check out this Rest of World article on generated images of people from different countries, as well as this Ars Technica article on Google's attempt to correct the problem, which led to historically inaccurate images. Technology is always a work in progress, so practice critical thinking and evaluate everything you see!
• GenAI is often programmed to prioritize user satisfaction over accuracy, so it will sound confident whether the answer is correct or not. If you challenge that answer, it may flip-flop its response in order to agree with you (again, whether the answer is correct or not) because it wants to make you happy.
• GenAI sometimes provides "hallucinations." It will tell you about people, dates, and article citations that sound real, but may not actually exist. Remember that AI is not thinking like a human - it's just predicting the next most probable word.
• Using the ideas and words of others without citation is plagiarism, so make sure you always cite your sources! There is current discussion over who owns the copyright to AI generated works, and some companies are being sued by creators who did not consent to their works being used to train AI. When in doubt, try to find and cite an original author.
• Since most AI systems are proprietary, there is little information on what goes in, what comes out, what data is being collected by the company about the user, and who is monitoring the industry overall. Never upload sensitive information to the AI, and always read the fine print.

Please note that these are generalizations about a technology which is improving by the minute, and it's quite possible that what I write today will be incorrect for most GenAI platforms by next month. For example, there are already scholarly tools which specifically address the problem with research article hallucinations. Like anything else you find on the internet, please evaluate all information with a critical eye and see if it is backed up by other reliable sources.

Generative AI Tools

When creating a prompt in a GenAI tool, it helps to be clear and concise. Tell the AI what role you want it to assume (college student, literary critic, etc.), what you want it to do (provide 5 essay topic ideas, create a grocery list), the target reading audience (explain it to me like I'm five), and any other important details. 

While ChatGPT may be the biggest name in the GenAI world at the moment, there are hundreds of AI websites and products available. Make sure you are using the right tool for the job!

Here are a few AI tools we enjoyed testing out. Not all of them are generative, but all are free to try without signing up for anything.

• Perplexity - A free AI search engine that cites its sources when providing answers, mostly pulled from trusted news outlets, academic papers, and established blogs.
• ConnectedPapers - A fantastic visual map of research articles. Search for a topic or DOI and see how a paper is related to other papers and scholarship on that topic.
• Talpa.ai - Don't remember the title of that cowboy space opera book? Can you only recall the red cover with the eyeball? Check out this AI-powered book description and cover search!
• Goblin.Tools - Small tools designed to help neurodivergent people with overwhelming tasks, such as estimating how long a task will take, figuring out if your email sounds too snarky, finding recipes based on what's in your cupboard, or creating to-do lists with the ability to break each daunting step down into additional steps.

Whether you opt to use GenAI or not, it is good to be aware of how it works as a tool and understand what it can and cannot do. As economist Richard Baldwin pointed out in 2023: AI won't take your job, but someone who can use it effectively might. 

(c) Pavel Daniyuk at Pexels

P.S. This blog post was not written by AI - just a human who attended several excellent AI sessions at the ALA library conference and then did a lot of reading.