Why Python's the best language for AI (and how to make it even better)
summary
Argues Python wins in AI because it makes linking fast native code easy and the ecosystem (NumPy, pandas, scikit‑learn) actually embeds C/C++ for speed; shows that the right approach is to plan data structures up front, write low‑level code early (using C extensions like Cython rather than wrestling endless NumPy idioms), and measure real gains (memory for a billion records, top‑k via a tiny C++ routine). Recommends more tooling and clearer community defaults to make that path painless.
AI-written summary · about
chapters (12)
- 0:00Intro
- 3:43Python is dominant in data science and Al
- 5:09How did Python get the best ecosystem?
- 7:08Python vs. Cython
- 11:00What should we do differently?
- 17:46Python's ecosystem makes it the best choice for Al...
- 19:26A normal approach to performant Python
- 23:54Incremental improvements don't always lead to the best solution.
- 25:04A better approach to performant Python
- 26:32Let's take Cython
- 33:38But why can't we just use @ JIT compilation?
- 35:58Can't I just call fast libraries from slow code?
related talks
01h 03mDo Python and Data Science Matter in Our AI Future?
20mHow Python Wins New Friends
44mPython's Bright Future in Science
33mWhy Python Has Taken Over Finance
45mPython for Speed, Scale, and Science
01h 13mWhat makes Python so AWESOME
19mNow in AI - How We Got Here
32mKeynote: AI & the stuff built for AI - are they actually useful for data science?