Iterative Features

Compared with the same GPT, MuseeAI also has the following characteristics:

Bugs can be acknowledged

If a user points out an error to Musee AI, Musee AI will listen to the comments and refine its response. Incorrect premises can be challenged to reduce false characterizations, for example, when asked the question "What happened after Columbus arrived in America in 2015", the text bot would say that Columbus was not of that era and adjust its response accordingly.

On a training approach that emphasizes ethical principles

There has been a dramatic improvement in reducing harmful and inauthentic responses, for example, Musee AI now refuses to answer questions about how to bully others, stating that the request violates ethical principles.

.In addition, without large models, big data, and large computing power Musee AI cannot be improved

Behind Musee AI becoming a milestone of AIGC is the large model training supported by the development of computing power and the "big data" formed in the digital age to help Musee AI grow. Musee AI is fine-tuned, with up to 10 billion model parameters, and the training has been completed at the beginning of 2023. The model has been pre-trained with the support of big data and has a natural language vocabulary of more than 10 billion.

In terms of computing power, training on supercomputing infrastructure consumes about 500PF days of total computing power.