Ali Hussein Lazem

Generative AI using Compression-based Language Models:

Generative AI such as ChatGPT for generating natural language text using large language models and deep learning have been extremely effective recently. However, these models require enormous amounts of training data, and the methods used are opaque – i.e. they are black-box and are not easily explainable.

Recently, compression-based language models have been shown to be very competitive, usually producing state-of-the-art results for tasks such as classification and prediction when compared to traditional machine-learning and generative AI methods. They also require significantly less resources to train and are explainable.

However, effective techniques for generative AI have yet to be developed for compression-based approaches. The purpose of this research project will be to develop new techniques so that these models can rival ChatGPT-based solutions at automatically generating natural language text.