I'm 10 Years Ahead of Chatgpt

This year, there were only 3 results for "deep learning" on CNKI. It is written in black and white in the textbook that "Neural networks have been proven to be inferior to SVM". Python is called the "glue language" and no one publishes papers on it. GPU is still synonymous with games, and Theano's documentation is as bad as a draft. And in his mind, he holds a complete roadmap for the AI ​​revolution in the next thirteen years. From AlexNet to Transformer, from ImageNet to GPT, he clearly remembers every node that changed the world. The problem is: between knowing the answer and making the answer, there is a year of 2012 when there are no graphics cards, no mentors, and no one believes in neural networks. The deputy dean said he was doing alchemy. His classmates thought he was crazy. The consensus in academic circles is that this path has been proven to be unworkable. But Shen Mingkuang knew that eight months later, a paper would blow these people's worldview to pieces. What he wants to do is to explode before that paper.