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

Lex Fridman Podcast full episode: • Chris Lattner: Future ...
Please support this podcast by checking out our sponsors:
- iHerb: lexfridman.com/iherb and use code LEX to get 22% off your order
- Numerai: numer.ai/lex
- InsideTracker: insidetracker.com/lex to get 20% off
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:
Podcast website: lexfridman.com/podcast
Apple Podcasts: apple.co/2lwqZIr
Spotify: spoti.fi/2nEwCF8
RSS: lexfridman.com/feed/podcast/
Full episodes playlist: • Lex Fridman Podcast
Clips playlist: • Lex Fridman Podcast Clips
SOCIAL:
- Twitter: / lexfridman
- LinkedIn: / lexfridman
- Facebook: / lexfridman
- Instagram: / lexfridman
- Medium: / lexfridman
- Reddit: / lexfridman
- Support on Patreon: / lexfridman

[@LexClips]

Full podcast episode: kzbin.info/www/bejne/ppWtgmufi7mqoJo Lex Fridman podcast channel: kzbin.info 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.

[@alexandrebrownAI]

TLDR: No

[@isuzu343]

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!

[@MadlipzMarathi]

There are two part of modular businesses MOJO programing language free os Modular engine closed They are planning making money on modular engine

[@user-rh1sw6ji6t]

llama.mojo now is faster then llama.c

[@yldrmcs]

It's a profit-driven company. they don't care about open source. they care about profit like other companies. even though they open the source code of mojo later, I don't expect this to be "free" in the sense that community decides and drives the direction. as a matter of fact, I feel mojo is like their a "PR" tool

[@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.

[@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.

[@Jeppelelle]

Also, a big chunk of PyTorch & TensorFlow heavy lifting is done in C++ so it makes even less sense of replacing PT & TF with clones written in Mojo & how much speedup would replacing the Python parts with Mojo since it is mostly used in the non speed critical parts anyway

[@satwikshresth3048]

Yo take a step back listen to the question

[@asunj]

A programming language is pretty much a core library. And this one in particular has functions that overlap with pytorch and tensorflow, so…

[@user-wd8wx5md5z]

The implicit question is "Will Mojo have libraries that can make it replace python/Pytorch/Tensorflow ? ". The poorly phrasing made me uncomfortable too, and it's a reason why I clicked on it out of curiosity . So I dont think it is an actual mistake, but just a way to provoke reactions .

[@Ramkatt]

​@@JeppelelleI don't think it's so much a matter of speed as it is a matter of scalability and flexibility.