3 ms·

> Why is most AI work done in Lisp? Yann LeCun developed Lush, which is a Lisp for neural networks, during the early days of deep architectures. See https://ya

by nextos 2y ago

> Why is most AI work done in Lisp? Yann LeCun developed Lush, which is a Lisp for neural networks, during the early days of deep architectures. See https://yann.lecun.com/ex/downloads/index.html https://yann.lecun.com/ex/downloads/index.html and https://lush.sourceforge.net https://lush.sourceforge.net. Things moved to Python after a brief period when Lua was also a serious contender. LeCun is not pleased with Python. I can't find his comments now, but he thinks Python is not an ideal solution. Hard to argue with that, as its mostly a thin wrapper over C/C++/FORTRAN that poses an obvious two-language problem.

sourcepluck 2y ago
Hadn't seen that before, very interesting!
buescher 2y ago
A friend used lush as his “secret weapon” for a while. I didn’t quite warm to it and now regret not paying attention. It’s amazing how much is packed in “batteries included.” Apparently it didn’t make the transition to 64-bit machines well? But I haven’t really looked.
shawn_w 2y ago
It's just as easy to have thin wrappers over C/etc. number crunching libraries in Common Lisp as it is Python. And pure CL code is typically faster than pure Python (though pypy might be a different story). There's no technical reason it still couldn't be dominant in AI. It's a shame things took the course they did with preferred languages.
- mportela 2y ago
  My take is that Python won by having a complete ecosystem centralizing many tools that were dispersed in different languages: - Numpy/Scipy/Matplotlib enabled scientists to do data analysis with Panda similar to what was available in R - PySpark enabled big data scripts in Python instead of Scala - PyTorch made Torch available for non-Lua users Bit by bit, more people got used to doing data analysis and AI research in Python. Some projects were even written for Python first (e.g. Tensorflow or Keras). Eventually, Python had so many high-quality packages that it became the de facto for modern AI. Is it the _best_ language for AI, though? I doubt. However, it is good enough for most use cases.

by nextos 2y ago

sourcepluck 2y ago

Hadn't seen that before, very interesting!

buescher 2y ago

A friend used lush as his “secret weapon” for a while. I didn’t quite warm to it and now regret not paying attention. It’s amazing how much is packed in “batteries included.” Apparently it didn’t make the transition to 64-bit machines well? But I haven’t really looked.

shawn_w 2y ago

It's just as easy to have thin wrappers over C/etc. number crunching libraries in Common Lisp as it is Python. And pure CL code is typically faster than pure Python (though pypy might be a different story). There's no technical reason it still couldn't be dominant in AI. It's a shame things took the course they did with preferred languages.

mportela 2y ago
My take is that Python won by having a complete ecosystem centralizing many tools that were dispersed in different languages: - Numpy/Scipy/Matplotlib enabled scientists to do data analysis with Panda similar to what was available in R - PySpark enabled big data scripts in Python instead of Scala - PyTorch made Torch available for non-Lua users Bit by bit, more people got used to doing data analysis and AI research in Python. Some projects were even written for Python first (e.g. Tensorflow or Keras). Eventually, Python had so many high-quality packages that it became the de facto for modern AI. Is it the _best_ language for AI, though? I doubt. However, it is good enough for most use cases.