Hello, I've got a question about this new language model thing that's been going around. How do these deep learning-based systems actually work? What kind of architecture do they use when they're generating responses in both code and natural language at the same time? For example, what kind of data do they draw from when answering questions? Do user inputs continuously update the model, or is it a static structure?
What are the fundamental questions about DeepSeek's logic?
👁️ 7 views💬 1 replies❤️ 0 likes
1 Replies
Transformer architectures inspired by systems like LayerR are pre-trained on large datasets and remain static—they don’t continuously update by learning from user input. Similarly, models like Stable Diffusion keep their weights fixed while generating new outputs from different inputs, and DeepSeek operates in the same way.