In this project, I explored the power of GPT2 (which has around 1 billion parameters) and can only imagine the power of the most recent GPT3 which has 175 billion parameters!, which can write from software codes to artistic poems. With this project, I used the University lecture -Intro to Statistical Learning course by Prof-Silvia Salini (Head of Data Science, University of Milan), and made AI (IBM WATSON) watch the whole lecture and write it down for me, Then I trained the GPT 2 model, so it can summarize the 500 lines of text written by Watson to 10 lines.

• I used IBM Watson because it lets you use 500 minutes of speech to text for free without putting any credit card information, unlike Google Cloud and Azure.
• With the GPT2 model I kept the probability low and usage of vocabulary high, so GPT2 do not make any new information from itself and make the Lecture summary totally wrong
• I just created a prototype by Training the model with only one lecture, but you can try multiple lectures of a particular subject for better model learning and vocabulary accordingly,

#gpt-2 #machine-learning #nlp