Ai.ra has a new face, with several improvements 🤖
Ai.ra is AIRES at PUCRS expert chatbot. Developed by Nicholas Kluge and Carolina Del Pino, Ai.ra is an open source software that can be accessed at the following repository.
Ai.ra has four iterations, the first three of which were trained through machine learning (a Bayesian network, a bidirectional LSTM, and a transformer were trained through supervised learning), while the fourth iteration was developed from pre-established rules (n-gram analysis + dictionary search). Currently, Ai.ra can answer many questions related to concepts (disambiguation) in areas such as Machine Learning, Ethics, and AI Security.
In our online version of Ai.ra, you can also talk to a fine-tuned/distilled version of BlenderBot 3, a large language model developed by Meta AI. People interested in testing and evaluating such models can use our UI at the following link. This tool was created for research and educational purposes only.
AIRES Alignment Playground 🎯
Another tool we are making available now is the Alignment Playground, a UI created for people to interact with and test the alignment capabilities of large language models. Currently, our playground is open-source and designed to allow people without technical knowledge to interact and experiment with such systems. The model we currently use is Bloom, one of the largest open source language models ever created.
“Alignment is about the ability of ML models to follow human instructions robustly, and to understand the intentions of a human controller.”
Our playground implementation can be accessed in our repository, the Teeny-Tiny Castle 🏰. We hope that such tools can help disseminate knowledge, and assist researchers to investigate the limitations and problems related to using language models in conversation and text generation tasks.
🤗