Return to Video

Behind the Mic: The Science of Talking with Computers

  • Not Synced
    ♪ (Music fades in) ♪
  • Not Synced
    (Chirping)
  • Not Synced
    (Vocalizations, different languages)
  • Not Synced
    (Talking overlaps in background)
  • Not Synced
    (Computerized beeping)
  • Not Synced
    (Man) We come into this world with the
    innate ability to learn to interact
  • Not Synced
    with other sentient beings.
  • Not Synced
    (Child vocalizing)
  • Not Synced
    (Man) Suppose you are to interact with
    other people by writing little messages.
  • Not Synced
    (Man) It'd be a real pain.
  • Not Synced
    (Man) And that's how we interact
    with computers.
  • Not Synced
    It's much easier just to talk to them...
    just so much easier...
  • Not Synced
    (Man) If the computers could understand
    what we're saying.
  • Not Synced
    For that, you need really
    good speech recognition.
  • Not Synced
    (Narrator) The first speech recognition
    system was developed by Bell Laboratories
  • Not Synced
    in 1952. It could only recognize
    numbers spoken by one person.
  • Not Synced
    In the 1970s, Carnegie-Mellon
    came out with the Harpy System.
  • Not Synced
    This was able to recognize over
    1,000 words and different pronunciations
  • Not Synced
    (Narrator) of the same word.
    - (Man) Tomato - (Woman) Tomato
  • Not Synced
    (Narrator) Speech recognition continued
    in the 80s with the introduction of the
  • Not Synced
    Hidden Markov Model, which
    used a more mathematical approach
  • Not Synced
    to analyzing sound waves that led to
    many breakthroughs we have today.
  • Not Synced
    You're taking in very raw audio wave forms
  • Not Synced
    like you get through a microphone
  • Not Synced
    on your phone
  • Not Synced
    or (unintelligible)
  • Not Synced
    (Woman) We chop it into small pieces
    and it tries to identify which phoneme
  • Not Synced
    was spoken in that piece of speech.
  • Not Synced
    - Phoneme is a primitive unit for
    expressing words.
  • Not Synced
    (voicing phonemes shown above)
  • Not Synced
    And then you stitch those together
    into likely words like Palo Alto.
  • Not Synced
    - Speech recognition today is good at
    transcribing what you've said...
  • Not Synced
    (Man, to phone) What's the weather
    like in Topeka?
  • Not Synced
    (Man) You can talk about travels, your
    contacts, like, "Where can I get pizza?"
  • Not Synced
    (Phone) Here are the listings for Pizza.
  • Not Synced
    (Man) "Where's the Eiffel Tower?"
    (Phone) The Eiffel Tower is ...
  • Not Synced
    (Woman) We've made tremendous
    improvements very quickly.
  • Not Synced
    (Man, to phone) Who is the 21st
    President of the United States?
  • Not Synced
    (Phone beeps)
    (Phone) Chester A. Arthur was the 21st...
  • Not Synced
    (Man, to phone) Okay, Google,
    where is he from?
  • Not Synced
    (Man) Years ago, you had to be an engineer
    to interact with computers.
  • Not Synced
    Today, everybody can interact.
  • Not Synced
    - One thing still in its
    infancy is understanding.
  • Not Synced
    - We need a far more sophisticated
    language understanding model
  • Not Synced
    that understands what the sentence means.
  • Not Synced
    We're still a very long way from that.
  • Not Synced
    (Beeping)
  • Not Synced
    ♪ (Soft background music) ♪
  • Not Synced
    (Woman) Our ability to use language is one
    of the things that helps us have culture.
  • Not Synced
    It's one of the things that helps
    us pass on traditions across generations.
  • Not Synced
    Figuring out how the system of language
    works, even though it seems easy,
  • Not Synced
    turns out to be very hard, but is one that
    every baby understands by 2 years old.
  • Not Synced
    (Girl) There's two of them.
    (Woman) There's two Ls, yeah (spells word)
  • Not Synced
    - Language is extremely complex
    and sophisticated...
  • Not Synced
    - From the semantics
  • Not Synced
    - (Man in chair) Ironies...
    - (Woman) Strong accents...
  • Not Synced
    - (Man) Facial expressions, human emotions...
  • Not Synced
    - Because that's part of
    how we communicate.
  • Not Synced
    - Humor...
  • Not Synced
    (Aside) Do I have to careful
    not to offend the dinosaur?
  • Not Synced
    - Language has so many different
    layers and that's why it's
  • Not Synced
    such a difficult problem.
  • Not Synced
    (Man) The (unknown) human brain
    and the learning algorithms in it
  • Not Synced
    are far, far better at things like
    language understanding
  • Not Synced
    and they're still a lot better
    at pun recognition.
  • Not Synced
    - Whether or not we replicate exactly
    what the brain does, to understand
  • Not Synced
    language and speech, is still a question.
  • Not Synced
    (Beeping)
  • Not Synced
    (Man) For many years, we believed that
    neural networks should work better than
  • Not Synced
    the dull existing technology
Title:
Behind the Mic: The Science of Talking with Computers
Description:

more » « less
Video Language:
English
Team:
Captions Requested
Duration:
07:19
  • Thank you so much, Michael. You descriptions of non-verbal sounds are great!

    Claude

  • You're welcome - thank you very much. :)

  • Thank you, Michael and Claude.

English subtitles

Revisions Compare revisions