Yes it is possible. Bloom is based on the Megatron GPT model which is also designed to be a "causal" language model. Causal here means that the text the model generates is based on the sequence of words that preceded it (this is called "unidirectional").

So depending on what you want your chatbot to say, you may want to do some fine-tuning to the model on your own dataset. You can also pre-prompt the model with text to give it context. Example pre-prompt: "I am a discord chatbot, how can I help?"

Once you have the output text of your model, you can use the Discord API to feed it into your discord chatbot.

Keep in mind, your chatbot may be unpredictable and likely won't say the same thing every time. If you want a more predictable chatbot that can dynamically choose from a pre-set number of phrases, you can look into tutorials like this - https://www.youtube.com/watch?v=8HifpykuTI4