March_AI Modeling|Operation, Limits and Breakthroughs of Big Speech Models(Up)
Everyone knows that artificial intelligence is called AI, but have you ever heard of AGI? Artificial General Intelligence (AGI) refers to artificial intelligence that can perform various cognitive tasks and has human-like learning and reasoning capabilities. Unlike specialized AI, AGI can adapt to different environments and solve problems autonomously, which is considered a key goal for realizing truly intelligent machines.
Currently, everyone's understanding of AI comes from the most accessible large-scale language models (LLMs), such as ChatGPT, BERT, T5, etc. It is a deep learning model with more than 100 billion parameters. It is a deep learning model with more than 100 billion parameters of Natural Language Processing (NLP) system, which has been trained with a large amount of text, so that it has the ability to recognize, summarize, translate, and generate text from a huge amount of knowledge. Most of the NLP models are based on the Transformer architecture, a groundbreaking concept that originated from Google's 2017 paper "Attention Is All You Need". The core principle is "self-attention mechanism", which recognizes keywords and calculates their relevance, and then deduces the overall semantic meaning. Compared to traditional RNN/LSTM, which needs to process sequential data step by step, Transformer can process all the data in one go.