Slashdot Mirror

← Back to Stories (view on slashdot.org)

Baidu Releases Open Source Artificial Intelligence Code (thestack.com)

Posted by samzenpus on from the check-it-out dept.

An anonymous reader writes: Chinese web services company Baidu has released a new artificial intelligence software called WARP-CTC. The code is apparently capable of speech recognition, particularly for short segments, that exceeds human capability. The source code uses an approach called 'connectionist temporal classification' and has been released on GitHub.

34 comments

  1. WHY? by Anonymous Coward · · Score: 0

    Um why?

    1. Re: WHY? by Anonymous Coward · · Score: 0

      If you read the link you'd know

    2. Re:WHY? by Anonymous Coward · · Score: 0

      The GitHub picy explains E.G [O-RRR-G-AAA-S-I-M] translates to the mandarin word "Kung Pow".

  2. "better than human" was achieved in 1994, already by ffkom · · Score: 5, Informative

    I've been developing automatic speech recognition systems in the 1990s. Back then, the best performing recognizers were based on Hidden Markov Models, and for "out of context" tasks like "determine whether an individual spoken word from an unknown speaker is 'nine' or 'none'", the automatic recognizers already achieved better recognition rates than humans. However, the specific human strength when recognizing fluent speech is to (a) quickly adapt to different speakers and (b) to fill in all the uncertain words from the understanding of the context, requiring "world knowledge". And that strength makes a very big difference. So the claim in the article is not really anything special, it is to be expected that computers are better than humans in this special task, for at least the last 20 years.

  3. Good by penguinoid · · Score: 1

    Now, can they please put this code to use on those terrible phone menus? With single words or phrases, like the article says? They can even have the advantage of AI-understandable context -- only a few responses match the question -- and yet they still get it horribly, hilariously wrong.

    --
    Don't waste your vote! Vote for whoever you want, unless you live in a swing state it won't matter anyways
    1. Re:Good by Anonymous Coward · · Score: 0

      Plug it in to Asterisk some how and shebang ... hey sounds like a cool idea right?

    2. Re:Good by rtb61 · · Score: 1

      Artificial intelligence do not belong in portable devices not for any paranoid reason but simply energy efficiency. To be actual artificially intelligent it must be capable of learning and investigating new solutions. So right or wrong cease for all non core functions where better or worse answers dominate. So logical, when queried the artificial intelligence will produce a 'good enough' answer, which is either used or rejected whilst in search for a better answer. Even with the good enough answer is used, the artificial intelligence should use available spare capacity to produce a better answer if possible, until such time as new requests come in or other internal research takes precedence (greater possible benefit or better results being generated). So pretty much AI computers would run at near 100% capacity all of the time, a substantive portion of it being internal algorithm research and generation to provide better solutions. AI should never be blanket AI covering everything but specific appliance based AI covering a particular range of functions only, otherwise it will always end up doing nothing, much the same as overly intelligent people sometimes do ;D (lost in the pure thought zone).

      --
      Chaos - everything, everywhere, everywhen
  4. GitHub in China... by Anonymous Coward · · Score: 0

    Is it not blocked ATM?

  5. Using Chinese code? by Anonymous Coward · · Score: 0

    Nope! After the Great Cannon incident and all the industrial espionage they've conducted, I'm not putting any code the Chinese government approves on any of my machines. Baidu is too close to the Chinese government to be trusted. I don't care if this project is open source. We've seen too many instances of malicious code slipping past open source audits.

    The irony that this code is on GitHub is rich.

    1. Re:Using Chinese code? by jandersen · · Score: 1

      The irony that this code is on GitHub is rich.

      Not the only irony around here. When I grew up, everybody KNEW, they just KNEW, that those evil, Chinese Communists should open up their society, embrace Capitalism and take part in the international society, where we freedom-loving Westerners would welcome them and teach them how to think. So they did, and now they are being pummeled for that very thing, not least because they dare to make up teir own thinking. It's open source - you are supposed to be able to inspect the code, and in fact, one of the fundamental strong points of open source is that there will be hundreds of millions of eyes, all ogling at every last semicolon, right? And because this isn't just some hobby project, but a serious effort into something of commercial and academic interest, it will get a serious looking over by people who understand this kind of thing.

      But if you don't like it, because it was thought up by somebody in China, then suit yourself. No-one's going to force you.

  6. Re:"better than human" was achieved in 1994, alrea by Gravis+Zero · · Score: 4, Interesting

    I've been developing automatic speech recognition systems in the 1990s. Back then, the best performing recognizers were based on Hidden Markov Models, and for "out of context" tasks like "determine whether an individual spoken word from an unknown speaker is 'nine' or 'none'", the automatic recognizers already achieved better recognition rates than humans.

    Sure... but this time they did it with Mandarin instead of English.

    --
    Anons need not reply. Questions end with a question mark.
  7. communist vs capitalist by Anonymous Coward · · Score: 1

    Apparently this CTC technique was pioneered by the Swiss (IDSIA) who attempted to capitalize on it (Lifeware). Where the first impulse of the chinese researchers at Baidu was to post it on github.

    To be fair, the Baidu researchers are located in Silicon Valley. So maybe this is just comparing socialists to capitalists.

    1. Re:communist vs capitalist by Anonymous Coward · · Score: 0

      So there is a communist enclave in the Silicon Valley?

    2. Re:communist vs capitalist by Anonymous Coward · · Score: 0

      Machine Learning researchers have a pretty solid history of opening up knowledge for everyone. Like MIT's Journal of Machine Learning Research being one of the first publicly open journals. Maybe something about the field impresses the importance of sharing knowledge on the practitioners. I'm sure that these releases (Baidu, MS, Google) are more about platform adoption and "free" development from the open source community when it comes to the company leaders. But none of these releases would be possible without the researchers pushing and facilitating.

  8. Re:"better than human" was achieved in 1994, alrea by Anonymous Coward · · Score: 0

    For those who haven't studied Mandarin:
    https://www.youtube.com/watch?v=5TuVL3mlBR4

  9. Recognising certain phrases... by Catmeat · · Score: 2

    Is it smart enough to automatically call the secret police if it hears the words "Falun Gong".

    1. Re:Recognising certain phrases... by Anonymous Coward · · Score: 0

      but then there's no google to get the secret number of the secret police.

  10. Re:"better than human" was achieved in 1994, alrea by Anonymous Coward · · Score: 0

    So you're saying when they say "this call may be recorded for quality or training purposes" that's consent for training the voice recognition system?

  11. What... by liqu1d · · Score: 0

    How on earth can you exceed human capability for understanding speech? Foreign tongue ok fair enough otherwise huh.

    1. Re:What... by Anonymous Coward · · Score: 0

      Humans don't recognize speech perfectly. Why is it difficult to imagine a machine might do better?

      If you test human ability against a set of speech samples they'll get some percentage wrong. If a machine can then better the error rate of humans you may conclude the system has exceeded human capability. Seems entirely plausible to me.

    2. Re:What... by Anonymous Coward · · Score: 0

      Uh, how would you know which examples it got right if a human can't tell?

    3. Re: What... by Anonymous Coward · · Score: 0

      Because they wrote down the words for the various persons to speak. Then they played this back to other people too see if what they wrote down matched the sentences given to the speakers.

      How could you not understand that?

  12. Re:"better than human" was achieved in 1994, alrea by Rob+Lister · · Score: 1, Interesting

    I had no idea they have been that good for that long. In the wild however, they suck. I hope I'm not coming off as super-critical. But it is the fault of humans for trying to make the machine pretend to be human.

    I want to start this by saying that the following is decidedly NOT satire.

    It is most probably the manner in which they are used that sucks the most. But maybe it is my own personal bias that is really to blame. Which is to say: I am prejudice against AI programmers.

    When a robot answers the phone, I get miffed. I can't help it. When they ask [i]whatever[/i] I answer [i]curtly[/i] because I think that it is beneath me to answer politely. And to be honest, I'm right! I'm not trying to start a class war here (okay, I said no satire but there it is).

    My belabored point is that I DO.NOT.WANT a machine to simulate politeness; it takes a human to pretend correctly. A machine has no conception of remorse or any other emotion so to simulate a human is ... well, it is insulting. I don't want to be a hater. And the machine doesn't care that I'm a hater. And I don't hate the machine, I hate the humans behind it that make it pretend to be human. Best:

    Say the name of your product [beep] Say the issue with your product [beep]

    And the voice that says that should not try to be passin'.

    Where am I going with this? Let those that respond decide.

  13. Re: "better than human" was achieved in 1994, alre by Zalbik · · Score: 1

    Take your pointless rambling nonsense somewhere else please, this is Slashdot.

    Dang, this is Slashdot?!?! Never mind....carry on.

  14. You would be surprised how bad you are... by ffkom · · Score: 4, Informative
    ... understanding speech if you had somebody (not yourself, of course) make recordings of strangers on the street uttering just a single word, each, randomly picked from a dictionary book dictionary. Chances are your recognition rate would be somewhere around 90%.

    Automatic recognizers achieve better rates on this task - but they'll loose against you when it's complete, sensible sentences that are being spoken, even more so if you heard more sentences from the speaker, before.

    1. Re: You would be surprised how bad you are... by liqu1d · · Score: 1

      Ahhh ok. Clearly I didn't understand the process there. Thanks :).

    2. Re:You would be surprised how bad you are... by Anonymous Coward · · Score: 1

      The Baidu code has 110 percent success rate --- the benefits of totalitarian capitalism revealed.

    3. Re:You would be surprised how bad you are... by Anonymous Coward · · Score: 0

      loose lips sink ships

  15. Re:"better than human" was achieved in 1994, alrea by Anonymous Coward · · Score: 1

    For those who may be interested learning more about a certain subject, here's an unlinkified URL of a video that may or may not contain any interesting information. I'm not going to tell you anything about the contents, let alone an interesting summary.

  16. Re:"better than human" was achieved in 1994, alrea by Anne+Thwacks · · Score: 1

    Most of the current crap fails miserably with English - it may or may not work with American - I could not possibly comment.

    --
    Sent from my ASR33 using ASCII
  17. Re:"better than human" was achieved in 1994, alrea by Anonymous Coward · · Score: 0

    I've been developing automatic speech recognition systems in the 1990s. Back then, the best performing recognizers were based on Hidden Markov Models, and for "out of context" tasks like "determine whether an individual spoken word from an unknown speaker is 'nine' or 'none'", the automatic recognizers already achieved better recognition rates than humans.

    However, the specific human strength when recognizing fluent speech is to (a) quickly adapt to different speakers and (b) to fill in all the uncertain words from the understanding of the context, requiring "world knowledge". And that strength makes a very big difference.

    So the claim in the article is not really anything special, it is to be expected that computers are better than humans in this special task, for at least the last 20 years.

    Utter bullshit. HMM never achieved better than human performance. The new approach is based on DNNs.

  18. Re:"better than human" was achieved in 1994, alrea by Anonymous Coward · · Score: 0

    https://www.youtube.com/watch?v=6TPMEoM-cjc

  19. Re:"better than human" was achieved in 1994, alrea by nycsubway · · Score: 1

    References please? I'm interested in this.

    --
    http://github.com/gbook/nidb