The largest streaming music platform in the United States, Spotify, has a pure Python backend, handling extremely high concurrency without any slowdown, and it continues to use Python to this day.
Both Reddit and Instagram have a significant amount of Python backend, which does not affect user experience at all.
Python is nicknamed the “glue language.”This nickname may not sound appealing, but it is particularly apt. Its role is to bond various high-performance modules written in different languages together like glue, allowing you to call them conveniently.
Consider the workflow of an AI project: data cleaning (which may involve connecting to databases and reading/writing files), feature engineering, model training (which may require calling GPUs), and model deployment (which may need to be packaged as a web service). In each of these stages, Python has ready-to-use, excellent libraries. It can easily communicate with the file system, network, databases, and C++ libraries.
This “jack-of-all-trades” characteristic makes it the best choice throughout the entire AI workflow.
So, stop worrying about Python’s “slowness.” It is a trade-off, a form of wisdom. It sacrifices its relatively insignificant native execution speed in exchange for a significant increase in research and development efficiency across the entire AI industry.
For newcomers looking to enter the field, my advice is: diligently learn Python and make good use of libraries like Pandas, NumPy, Scikit-learn, and PyTorch/TensorFlow, and you will be able to solve 99% of the problems. When you become an expert and need to optimize low-level operators or write CUDA kernels, you will naturally learn C++. But for the vast majority of people, Python is the best bridge to the world of artificial intelligence.
There is too much information online, and everyone is desperately bookmarking, resulting in too little reading. Many people’s bookmarks eventually become “knowledge graveyards.”
As the old saying goes: learn to drive first, then worry about how the engine is built.