4 ms·
Python's a lingua franca in AI/NN because it was already a dominant language in scientific computing. Its dominance in scientific computing grew steadily throu
by auntienomen 5y ago
Python's a lingua franca in AI/NN because it was already a dominant language in scientific computing. Its dominance in scientific computing grew steadily through the 1990s and 2000s, for a few reasons:
1) Python -- specifically CPython -- made it easy to wrap existing, thoroughly tested high performance libraries in Python APIs. So, you got easy access to things like GSL and BLAS and LAPACK, but you get to call numpy.linalg.svd instead of GESDD.
2) Python was a general purpose language, unlike R or MATLAB, so you could extend existing systems to do more without running into a wall.
3) Python was a heck of a lot less effort to use than C++.
- z3phyr 5y agoCan you link any source on the python scientific usage in the 90s and early 2000s? I think the dominant language in science at that time was a mishmash of MATLAB, Java, C++, FORTRAN and Perl (In Biology at least, perl was the goto glue language due to its excellent string processing capabilities)
- auntienomen 5y agoI'm not sure I could find sources on the web any more easily than you. (Maybe start by looking at references in David Beazley's old talks?) I was in physics at the time, and what happened there was that Python basically enveloped Fortran & C++, letting people use the existing code without getting bogged down in complicated invocations.
- eesmith 5y agohttps://www.youtube.com/watch?v=4RSht_aV7AU https://www.youtube.com/watch?v=4RSht_aV7AU is a recent Beazley talk about that era.
- _ugfj 5y agohttps://aip.scitation.org/doi/pdf/10.1063/1.4822400 https://aip.scitation.org/doi/pdf/10.1063/1.4822400 I can attest to Python usage in physics exploding after this. What Livermore says, you listen to. They were considered the best of best, after all. An unknown student making an unknown library, no one cares. But when Livermore says, hey guys, Numeric is interesting, you listen. Things rolled from there.
- pjmlp 5y agoHere CERN usage of Python during those early 2000s. https://lhcb-comp.web.cern.ch/support/CMT/cmt.htm https://lhcb-comp.web.cern.ch/support/CMT/cmt.htm https://home.cern/news/press-release/cern/lhc-computing-grid-goes-online https://home.cern/news/press-release/cern/lhc-computing-grid... https://cds.cern.ch/record/840543/files/lhcc-2005-024.pdf https://cds.cern.ch/record/840543/files/lhcc-2005-024.pdf
- auntienomen 5y agoExtending the comment a little: I'm crediting CPython for the nice APIs, but really I should be crediting David Beazley's SWIG. That got the ball rolling on a lot of projects.
- Qem 5y agoHe gave a talk on this very question in the past week. I strongly recommend. See https://m.youtube.com/watch?v=4RSht_aV7AU https://m.youtube.com/watch?v=4RSht_aV7AU
- radmuzom 5y agoI was a student in the early 2000s and part of the academic community and have personally seen some of the world's best scientists working on physics or computer science. Not one of them used Python - in fact our professor chose to teach us Haskell and C in the first year of under-graduation. C, C++ or Fortran was quite common for scientific programming.
- pjmlp 5y agoIndeed, Python was already the main scripting language at CERN when I was there during the early 2000's.