Google Moves Ahead With Contributing The MLIR Machine Learning IR To LLVM

[bridgman]

Too many acronyms.

You mean "TMA", right ?

[coder]

I can't say there's any real benefit to anyone other than conmen and some major corporations' bottom line. The current state of the art is still not real AI and it doesn't benefit anyone outside of research laboratories, yet.

Deep learning already performs a lot better, on certain pattern-matching problems, than humans.

But, it does more, besides - ever heard of "deep fakes"?

[coder]

Their statement is about AI integration and AI running in "products that you use every day and run smoothly on all the devices you have". That's squarely consumer hardware.

I used the word "limit". I didn't say they didn't address the scope of their work - the issue is that they went beyond that, with the part of the quote that you omitted.

Are you trying to troll me, or just having a momentary lapse of brain function?

This problem will solve itself like most others, by decreasing the total population, which will happen one way or another in the next decades anyway.

Okay, so it'll solve itself with even bigger problems? Gee, that's a good plan.

[stormcrow]

Too many acronyms. This is almost as bad as working for the government where everything ends up an acronym. If you're going to use "IR" as anything other than the normal use (infrared) please expand it in first reference.

IMO, the jury is still out on whether more AI is a net benefit to humanity.


The jury called back, they said that AI is a net benefit to humanity (i.e. a population), but it's not a net benefit for the single individuals that are discarded and left out as "obsolete".

Naw as it currently stands "AI" is nothing more than the current marketing language for "proprietary black box" and snake oil. It's quite literally lost its original meaning because it's being applied to anything from simple algorithmic filtering and snake oil in some rather public drama to simple polymorphic programming techniques and deep learning. Of them, only the last is even remotely artificial intelligence.

I can't say there's any real benefit to anyone other than conmen and some major corporations' bottom line. The current state of the art is still not real AI and it doesn't benefit anyone outside of research laboratories, yet.

[nuetzel]

Push 'standards' due to market share...
Times repeat?

[starshipeleven]

I think they should really limit their statements to that, and not try to make broader statements that are clearly beyond the scope of their work.

Their statement is about AI integration and AI running in "products that you use every day and run smoothly on all the devices you have". That's squarely consumer hardware.

The stuff that is killing jobs is in the industrial or at the very least business sector, and in most cases is not using AI.

AI has yet to destroy any significant amount of jobs, and it will probably take down most "servant" roles like secretaries, front desk roles, and drivers. Neither is consumer hardware.

IMO, the jury is still out on whether more AI is a net benefit to humanity.

The jury called back, they said that AI is a net benefit to humanity (i.e. a population), but it's not a net benefit for the single individuals that are discarded and left out as "obsolete".

This problem will solve itself like most others, by decreasing the total population, which will happen one way or another in the next decades anyway.

[coder]

These innovations can then quickly make their way into products that you use every day and run smoothly on all the devices you have—ultimately leading to AI being more helpful and more useful to everyone on the planet.

Well, that's sure optimistic. To the extent AI destroys jobs, it'll sure do that faster. IMO, the jury is still out on whether more AI is a net benefit to humanity.

Obviously, that's not their problem to solve. Understandably, they're just trying to do AI more efficiently. I think they should really limit their statements to that, and not try to make broader statements that are clearly beyond the scope of their work.

[tildearrow]

Latner or Lattner?

[phoronix]

Phoronix: Google Moves Ahead With Contributing The MLIR Machine Learning IR To LLVM

Back in April we wrote about MLIR as Google's new IR designed for machine learning. This intermediate representation was designed for use by any machine learning framework and now this common format is being contributed to LLVM...

http://www.phoronix.com/scan.php?page=news_item&px=Google-MLIR-To-LLVM