Slashdot Mirror
Microsoft Releases Its Deep Learning Toolkit On GitHub (microsoft.com)
An anonymous reader writes: Microsoft is moving its machine learning Computational Network Toolkit (CNTK) from its own hosting site, CodePlex, to GitHub. They're also putting it under the MIT open source license. The move marks an effort to make it easier for developers to collaborate on building their own deep learning applications using the CNTK. Under the CodePlex license, access was restricted to academics only, and it was wholly targeted to that audience. Now that it's opening the project to everyone, Microsoft hopes to attract a greater number of developers, and a wider variety as well.
This follows similar releases from Google and Baidu.
This is inevitable. Our society needs to figure out how to deal with it, instead of inciting fear that the world is doomed because of it.
All my liberal friends think I'm a conservative, all my conservative friends think I'm a liberal.
This is inevitable. Our society needs to figure out how to deal with it, instead of inciting fear that the world is doomed because of it.
This is why the universal basic income is inevitable at some point. It either that or massive riots to protest the organised poverty.
Video of some good progressive thrash music
..., Microsoft hopes to attract a greater number of developers, and a wider variety as well.
Good luck. I've worked – professionally, both directly and indirectly – on a couple different open source projects. They all had (and still have) lots of users.
But growing the list of active developers from outside the core cadre of devs? Next to impossible.
If they're open sourcing it to get more active development, I expect they've got a tough row to hoe.
I'm more inclined to think they're just not funding development and this is a desperation move.
I though their Deep Learning Toolkit was named Windows 10 - wait, who's learning what?
It must have been something you assimilated. . . .
Call me when software can dig ditches and build roads all on it's own.
Do not look at laser with remaining good eye.
Skynet?
You know Google did the same thing four months ago, right?
He's a troll notice the AC lol
Not necessarily, for 2 reasons. First, math is math, and most of the building blocks of machine learning are pretty well known. What is valuable is how the building blocks are assembled and set up for a given problem. Second, without large amounts of collected data, even knowing the exact setup of somebody's model isn't particularly valuable. So a company can open source a lot of good general optimizers and other functions without giving away any "secret sauce."
To some extent this relates to the regime change that happened after Ballmer left. Microsoft previously shunned open source and the Linux ecosystem. Now they are much more open to it, and are shifting to Linux for (e.g.) their GPU computing. Apparently they are also rewriting a lot of the CNTK code to work better with Linux. (Technically, C++ should work well with both Windows and Linux, but with things like threads there can be differences in the way you do things).
They likely shifted this from the standard MSR license (which only permits research use) to the MIT license mostly in order to increase adoption. Even many academic users do not like research-only licenses because if you become dependent on that type of software it closes the door to later commercialization and some types of collaboration with industry. I myself am an example of this-- previously I did not consider using CNTK because of the license. (In the meantime I have built my own toolkit with similar capabilities, and will likely stick with it now because I understand it).
The group wanted to convert to a more open license for a while but were probably waiting on management and lawyers. They are probably upset that it didn't happen sooner, because in the meantime Google released TensorFlow, which will have taken away a lot of the potential market.
The
Maybe never "all on its own" but with a massively reduced number of people required.
It used to require a lot of people with shovels.
Then it required fewer people with bull dozers.
The big earth movers are already moving to semi-autonomous. Next step is a single 'drone' operator in AC working on the stuff that needs a human to do while a few other dozen in the fleet manage themselves automatically.
http://www.cat.com/en_US/artic...
https://www.youtube.com/watch?...
Newsflash: there's not enough need of that to occupy that much of the population, and with a missing middle class current capitalism doesn't work anyway.
Did it? While one on the main researchers on NN keeps working on Theano, not to mention everything else that exists, I have my doubts.
The point is that TensorFlow addresses basically the same problems that CNTK does well-- e.g. recurrent architectures like LSTMs, and good GPU support.
The
One always needs to beware of that, especially from MS. I'd need to check, does the MIT license protect you against patent infringement suits based around the released code? If not, then I'd suspect a trap, even if not a rootkit.
I think we've pushed this "anyone can grow up to be president" thing too far.
You are leaving out the millions that starved to death in the wake of automation. (It didn't happen in one year, but try thousands over a hundred years.) Some times were worse than others, and it occasionally hit a peak. Check out the history of the Enclosure Acts. Or the Luddites. The modern weavers (in US/Europe) do much better than the weavers of that time, but they aren't their descendants. Too many of them died.
I think we've pushed this "anyone can grow up to be president" thing too far.
P.S.: The Luddites are also significant to attend to if you want to observe the way history is managed by the wealthy to make them look good...or at least not totally evil. It is rightly said "History is written *for* the winners". And much of how it's managed isn't through acts of war as such, but through economic pillaging. (And if you don't think that's happening wherever you live today, you've got blinders on.)
I think we've pushed this "anyone can grow up to be president" thing too far.
I think you missed the <sarcasm> tag.
I think we've pushed this "anyone can grow up to be president" thing too far.
AI isn't going to be close the Hollywood AI for a long time. Say 20-30 years. It requires too much computer power.
OTOH, there are specializations that can totally upset society that will probably be showing up a lot sooner. People can dis self-driving cars all they want, but they are going to cause social upheaval. Probably as much as did the switch from horses to cars, and over a shorter time-frame. And that's just *one* of the applications on a near future track. And while mobile AI is heavy, interconnected AI isn't. So Siri, etc., will be getting a lot better, and showing up in a lot more places. And there will be other *specialized* applications that we just don't see yet.
I think we've pushed this "anyone can grow up to be president" thing too far.
If MS wants to target a non-academic audience, this will live or die on the quality of the toolkit samples.
Every machine need less people working on it than the number of people needed to perform the task in the first place. That's automation: allowing fewer people to perform the same amount of work, including the guys needed to maintain the machines.
If it were different, there would be no automation going on. But there is, and it's steadily moving massive amounts of people to unemployment.
Video of some good progressive thrash music
I just bumped into Soren Kierkegaard in the elevator. He told me machines will never be able to make choices or value judgments since both are predicated on agency/subjectivity.
Call me when a bot wins a Turing competition without shenanigans.