Will Mojo replace PyTorch and TensorFlow? | Chris Lattner and Lex Fridman

Please support this podcast by checking out our sponsors:

GUEST BIO:
Chris Lattner is a legendary software and hardware engineer, leading projects at Apple, Tesla, Google, SiFive, and Modular AI, including the development of Swift, LLVM, Clang, MLIR, CIRCT, TPUs, and Mojo.

PODCAST INFO:

SOCIAL:

[LexClips]

Guest bio: Chris Lattner is a legendary software and hardware engineer, leading projects at Apple, Tesla, Google, SiFive, and Modular AI, including the development of Swift, LLVM, Clang, MLIR, CIRCT, TPUs, and Mojo.

[isuzu]

What exactly is the business case for Mojo? Can Chris commit to making it open source? Julia is academia-driven. Python has huge open-source backers. Neither of them is a product of a startup like Mojo is for Modular.
If Mojo is made free and publicly available, who's paying the bills for Modular? Will it have a licensing tier structure for academia/research/commercial use like other software vendors (Mathematica/Matlab) do?

The entire "closed beta with a wait list" approach that Modular is taking with Mojo has made me skeptical about it's open source viability. Any clarifications would be welcome!

[codelabspro]

Mojo is a language. PyTorch and TensorFlow are libraries written in Python. Asking if a language can replace libraries is a non sequitur. Sorry Lex. But you have to ask Lattner questions that make logical sense.

[freedom_aint_free]

In a not so distant future coding AI will be so incredible that you will be able to rewrite the whole libraries on one language into the other allowing any language to be broadly useful, paradoxically by that time, it won't matter anymore, AI will be able to write in machine code anything for anything getting insane speeds far beyond of what is humanly possible.