Python: как самый медленный язык программирования стал королём нейросетей

Python, created by Guido van Rossum back in February 1991, has by 2026 established itself as the primary language for machine learning and data science — despite its reputation as one of the slowest languages. A breakdown on Habr explains that the issue isn't the speed of the language itself, but its architecture and ecosystem.

Why Python is considered slow

Python runs slower than compiled languages because it is interpreted, uses dynamic typing, and has a Global Interpreter Lock (GIL), which prevents threads from executing bytecode in parallel. The interpreter parses each line at runtime rather than in advance — hence the loss in speed.

That said, the language is far from young. The first version of Python was released in 1991 and was named after the British comedy show "Monty Python's Flying Circus." Python 2.0 appeared in 2000, and the incompatible Python 3.0 — in 2008, meaning years before today's AI boom and most modern web and ML frameworks.

How Python took over machine learning

Python became the standard for data science thanks to an ecosystem of libraries that offload heavy computation to fast C/C++ code and to GPUs. The developer writes a simple, readable script, while under the hood matrix operations are computed not by the interpreter, but by optimized backends.

Key milestones of this ecosystem:

• 1991 — Guido van Rossum releases the first version of Python
• 2006 — NumPy appears, a library for fast array computations in C
• 2008 — pandas is released for working with tabular data
• 2015 — Google open-sources the TensorFlow framework
• 2016 — Facebook (now Meta) releases PyTorch, which became the standard in research

It is precisely this combination — "convenient syntax on top, fast C and CUDA underneath" — that removed the main objection to Python: users don't need to sacrifice speed for simplicity, because performance-critical parts are already written in low-level languages.

What Python gives developers

Python's main advantage is its low barrier to entry and code readability, which is why it's chosen by both beginners and researchers, for whom quickly testing a hypothesis matters more than squeezing the maximum out of the hardware. Under the language's philosophy, simple and clear code is valued more highly than clever and "fast."

"Simple is better than complex.

Readability counts." — Tim Peters, "The Zen of Python" (PEP 20)

Popularity is also confirmed by rankings: in the TIOBE index, which tracks demand for programming languages, Python has held first place for several years now, overtaking traditional leaders like C and Java.

What this means

The history of Python shows that what determines a language's dominance in AI isn't "raw" speed, but ecosystem and convenience: heavy math can be delegated to C and GPUs, while people are left with simple syntax. That's precisely why a language from 1991, rather than faster newcomers, has become the face of today's wave of neural networks.

Frequently asked questions

Who created Python and in what year?

Python was created by Dutch programmer Guido van Rossum; the first version was released in February 1991. The language's name refers not to the snake, but to the comedy show "Monty Python's Flying Circus."

Why is Python slow, yet used in machine learning?

The Python interpreter itself is slow because of dynamic typing and the GIL, but in ML the heavy computations are performed by the NumPy, TensorFlow, and PyTorch libraries, which are written in fast C/C++ and run on GPUs. The developer writes simple code, while a low-level backend delivers the speed.

Is Python the most popular programming language?

According to the TIOBE index, Python has held first place among programming languages for several years now, largely thanks to the boom in data science and machine learning.

*Meta has been designated an extremist organization and is banned in Russia.