---
tags: ai
---

# character.ai

* [link](https://beta.character.ai/)

## description
* Generic description:
* Character AI is an artificial intelligence app that allows you to access and interact with thousands of AI-generated characters. Similar to OpenAI’s ChatGPT, you can chat with these AI bots freely online.

* User notes:
* Similar to chatGPT, but feels more stiff
* As the AI is limited to acting “in persona”
* trained to masquerade as various characters or people
* I am curious how they did this– is the character limited to canon?
* You can create your own character
* Elements required:
* Short description
* Long description
* Example responses
* “Example conversations and information to define your Character. The first 15-30 messages are the most important.”
* However, it can also simulate conversations, plan travel, etc.
* Yet, these applications seem better on chatGPT or other AI services specialized to those uses

## tutorials

* [how to create a character](https://book.character.ai/character-book/welcome-to-character-book)
* [youtube review and tutorial](https://www.youtube.com/watch?v=zbwx0e1cx-c)

## prompting tips
* [“Learn Prompting”](https://learnprompting.org/)

## examples of creations
* [eventually link a character we make?]

## pedagogical use-case

* Most useful for pedagogy by:
* Putting students in conversation with dead thinkers/historical figures
* You could run an “interview” with someone
* Prepare questions beforehand by doing research on this person
* OR have a debate on a specific topic
* Though, the AI struggles to “change its mind”
* So it would be quite impressive to convince AI Thomas Aquinas that god isn’t real
* And anyone who can do that should get an a+
* But only if the bot is well-trained
* Some characters made by users are rather weak, as they haven’t been trained on enough examples/texts from the “character”
* VERY good way of testing students on their working knowledge in the humanities
* As long as the chosen “character” relates to the course material and is well-trained.
* Also an effective way to study and practice working knowledge as a student. Especially in a context like a political philosophy class where the whole aim is to be able to have genuine debates between differing idologies, it is very powerful to be able to take one side of an argument against an ai that embodies a certain thinker.
* This can even work in debates where the thinkers were never in direct opposition, [for example](https://hackmd.io/@lluf-22-23/HJJWWaLEn) I tried to argue for Marxism against Thomas Jefferson which was very interesting.
* It is able to "update" certain views of historical figures, like in this bot Jefferson came out against slavery and Native American genocide. This can be interesting in terms of keeping debates relevant to the time so you can really focus on the solid parts of their ideologies which could still stand up today. But, of course one has to be cognizant of the historical inaccuracies and understand the real views of these thinkers at the time.
* It could be a very interesting practice to try training one of these characters yourself and create as close to a real thing as possible. This would require active thought and synthesization of which sources are relevant to this "character's" opinions and understandings