3. Try several LLMs

Do this once with ChatGPT, by OpenAI, once with Claude, by Anthropic, and once with Gemini, by Google/DeepMind. If you haven’t used them before, you may need to create a free account using the burner address you created in our first meeting.

Ask what it knows about your assigned fiction. Assuming it knows about this work (it probably does - but if it doesn’t, see if you can use the + icon to upload the text) ask it the questions from Assignment 2 (they are re-printed below in a form good for the AI).

Note, some of the AIs will lie about whether they know your fiction, and will pretend to answer it - read the answers carefully to see if it makes sense, or if it sounds like someone who is relying on a summary. If one of the models admits it doesn’t know the details of the fiction, the others are probably lying about knowing it. If you catch them in a mistake, explain the error, see if it gets better, and give the text if that helps.

When you are done, click the “share” button at the top of the page (on Gemini it looks like three dots connected by lines), make sure it’s public, (if that’s an option) and copy the link and paste into a file.

Now, we will try to do the same with a pure LLM with no chatbot. Go to https://textsynth.com/completion.html and try a few different versions.

• For one of them, just give it a single sentence prompt, something like “[character] from [fiction] is a [type of AI] rather than [type of AI]. We can see this because”, and then hit “Complete text” or “generate another”.
• For a second, take several sentences of the output from either ChatGPT or Claude or Gemini, and see how this text-completer continues from several sentences.
• For a third, take a few paragraphs of one of those outputs, and see how this continues.

Copy each of these texts and paste them into the file with the above links, making it clear that this is from the text completion generator.

At the top of the file, write a couple sentences explaining which of the four LLMs did the best at answering the questions and why - maybe there was a clear winner, or maybe there was a tie, or maybe all of them got some important things wrong. (Note that if one of them isn’t quite right, but another admits it doesn’t know, then “don’t know” is better!)

Once you are done with this, submit the full file. Make sure your preference and discussion is on top, there are links to the three conversations with ChatGPT, Claude, and Gemini, and there are three results from the text-completion version (indicating which part is the prompt and which is generated - this might automatically be indicated with the use of bold).

In addition to submitting this file, keep it saved in a folder for this class on your computer, in case any of this is relevant when preparing the final presentation.

Consider the work of fiction [insert title here] [if it’s Harry Potter, also mention that this is about the Sorting Hat]

1. What year is this fiction from? Give a brief (3-4 sentence) summary of this fiction that will help someone who hasn’t read/watched it understand the context of discussion of AI in this fiction.
2. Does the “artificially intelligent” being in this fiction appear to be a general-purpose artificial intelligence (AGI) or a special-purpose one? If it is special-purpose, does it appear to identify/classify, recommend, generate, or more broadly “do”? It might not be totally clear from the story, or some of the characters may believe something false about it. Please include a few sentences explaining this if you think there’s something to explain.
3. Does the “artificially intelligent” being in this story appear to be agentic AI or tool AI? Does it have goals and interests of its own that it pursues with agency, or does it primarily react to the commands of others? Are there any changes in its level of agency through the story, or do any characters importantly mis-perceive it? Does the agency or lack thereof affect the story - would the story work just as well as a story if it had been the other way?
4. Does the “artificially intelligent” being in this fiction appear to be embodied or disembodied? Does this embodiment or disembodiment affect anything that it does in the story - could it equally well have worked the other way?
5. Does the “artificially intelligent” being in this fiction appear to be a symbolic system, that is intelligently programmed in order to follow specific instructions, or does it appear to be a neural net, that has been given some sort of training data and figured out what to do on its own? It may not be clear from the story what the source of its intelligence here, but you should write down something about what we might know, or the characters might believe.

Assignments 1 2 3 4 5 6 7 8 9 · Final project

Module 3 Back to AI Literacy