This Architecture Powers ChatGPT | Transformers Explained

Key Takeaways

Hey folks, I'm back with new short. In this video, we'll discuss about transformers architecture. Uh let me show you transformers architecture how exactly it looks like. Well, don't worry about this massive architecture. It looks super fun and interesting. Diving forward, transformers architecture has two components. Encoder and decoder. Encoder is mainly used for understanding the text whereas decoder is used for generating the text. Uh let's look into history of transformers architecture. Well, transformers architecture is published in the paper called the attention is all you need. This paper is published by Google researchers in 2017. Uh Google want to build a language translator just like Google translation using this architecture. They don't have any intentions to build large language models or chatbot. that in 2018 OpenAI saw the potential of this architecture and using just the decoder part they built um uh GPT1 the first large language model built on top of transformers they train millions of available internet data at that particular time from there the evolution happened rapidly uh you can see the large language models in today's world like uh GPT cla Gemini gro many models uh But after a few days um Google also built other language model uh using just the encoder uh component but this model is good at understanding but it can't generate like uh GPT since it is used only the encoder part. Uh I hope you found this uh video more interesting. In coming videos we'll discuss about lot of uh transformers components. Uh until then signing off Mitt Dama. Thank you.