Week 19: Defining the technologies of the project

In this week, the team started to define the details of the project, such as the technologies to be used. From the research carried out, both Bert and GPT-3 are the best pre-trained models available for natural language processing (NLP) and auxiliate in many difficult tasks in this field. Despite GPT-3 looking like the better option in performance, the team identified a few problems with the tool when compared with Bert. For starters, the GPT-3 is not open source which can cause a few drawbacks when trying to develop a full project without using financial resources. Another problem is that so far looks like the GPT-3 is harder to apply to specific business domains due to the GPT-3 model not being able to be fined tuned.

Now, the idea is to keep exploring both the tools and to find out which one serves the propose of this project better.

The classes of this week focused on language modelling in the Natural Language course while in the Machine Learning course the focus was on the inner workings of neural networks and perceptrons.