What is ChatGPT? How does it work?

ChatGPT 3.5 is a highly sophisticated "large language model," a tool that can generate human-like text based on context and past conversations. ChatGPT is built on large amount of "training material" found on the internet from pre-2021. While it can answer questions, generate images, and write fluently, ChatGPT 3.5: 

• Is not an encyclopedia of knowledge or a database of information 

• Is not connected to the internet (unlike some other generative AI tools)

• Is not a discrete work by an author - your prompt can generate different information each time

• Is not unbiased or neutral, as bias is integrated into the training materials

ChatGPT can, however: 

• Assist with creative writing tasks, including brainstorming, composing an outline, grammar checks, generate research questions 

• Provide generic writing or knowledge that may lack specific details or accuracy

• Write code, analyze data

There are thousands of large language models, and while some are based on OpenAI, many use a smaller subset or entirely different training material. We will try to keep this page up to date with new tools that add additional features. 

How can you use LLMs?

Fall 2023 examples: Text-based Generative AI tools like ChatGPT can be useful for you when you are brainstorming topic ideas or trying to come up with keywords to use to search on a specific topic. Remember that these tools are not storehouses of knowledge: they are fluency-based text-language generators, and are using their training materials to deliver an answer.

ChatGPT cannot yet find specific peer-reviewed articles, but it can help you get ideas for how to search for scholarship. It can suggest keywords, search strings, and format citations. 

Example: I asked ChatGPT for some keywords to search for sources on the Conquest of Mexico. ChatGPT generated the following list, which is mostly fine until about the 15th option:

Here are some keywords that you could use to research the conquest of Mexico (notes on results in red):

Hernán Cortés, Aztecs, Montezuma II, Spanish Conquest, Tenochtitlan, Indigenous peoples of Mexico, Conquistadors, New Spain, Mesoamerica, Nahuatl language, Mexica, Tlaxcala, Bernal Diaz del Castillo, La Malinche, (less-than-accurate keywords begin here): Francisco Pizarro (best known for expeditions that led to conquest of Peru - not Mexico), Inca Empire (The primary indigenous empire of Peru), Conquest of Peru, Colonialism (too broad), Imperialism (too broad), Indigenous resistance (too broad)

You can prompt ChatGPT to be more specific in one or more areas, but note that the same will happen - a mix of good ideas, wrong ideas, and overly broad suggestions. 

In addition to ChatGPT, there are other AI tools that would be useful, and offer similar assistance with brainstorming.

I used the same prompt in Google Bard. At this time (fall 2023), its results were a little more accurate.

Here are some keyword phrases you can use to find scholarship about the conquest of Mexico:

• Spanish conquest of Mexico, Hernán Cortés, Aztec Empire, Mexica, Tenochtitlan, La Noche Triste, Malinche, Tlaxcala, Cuauhtemoc, Montezuma II, Bernal Díaz del Castillo, Francisco López de Gómara, Sahagún, Bernardino de

Additionally, Google Bard offered me the following advice: "You can also use more specific phrases, such as:

• The role of disease in the conquest of Mexico, The impact of the conquest of Mexico on the Aztec people, The legacy of the conquest of Mexico

When searching for scholarship, it is more helpful to use databases of work by academic publishers, such as JSTOR, Project MUSE, and EBSCOhost. These databases will allow you to search for articles and books that have been published in scholarly journals and by academic presses.

Here are some additional tips for finding scholarship:

• Read the introductions and conclusions of articles and books to get a sense of the main arguments being made.
• Pay attention to the footnotes and bibliographies to find other sources that you can consult.
• Be aware of the different perspectives - there is often no single "correct" interpretation of an event.