WEBVTT

99:59:59.999 --> 99:59:59.999
♪ (Music fades in) ♪

99:59:59.999 --> 99:59:59.999
(Chirping)

99:59:59.999 --> 99:59:59.999
(Vocalizations, different languages)

99:59:59.999 --> 99:59:59.999
(Talking overlaps in background)

99:59:59.999 --> 99:59:59.999
(Computerized beeping)

99:59:59.999 --> 99:59:59.999
(Man) We come into this world with the
innate ability to learn to interact

99:59:59.999 --> 99:59:59.999
with other sentient beings.

99:59:59.999 --> 99:59:59.999
(Child vocalizing)

99:59:59.999 --> 99:59:59.999
(Man) Suppose you are to interact with
other people by writing little messages.

99:59:59.999 --> 99:59:59.999
(Man) It'd be a real pain.

99:59:59.999 --> 99:59:59.999
(Man) And that's how we interact
with computers.

99:59:59.999 --> 99:59:59.999
It's much easier just to talk to them...
just so much easier...

99:59:59.999 --> 99:59:59.999
(Man) If the computers could understand
what we're saying.

99:59:59.999 --> 99:59:59.999
For that, you need really 
good speech recognition.

99:59:59.999 --> 99:59:59.999
(Narrator) The first speech recognition
system was developed by Bell Laboratories

99:59:59.999 --> 99:59:59.999
(Narrator) in 1952. It could only recognize
numbers spoken by one person.

99:59:59.999 --> 99:59:59.999
(Narrator) In the 1970s, Carnegie-Mellon
came out with the Harpy System.

99:59:59.999 --> 99:59:59.999
(Narrator) This was able to recognize over
1,000 words and different pronunciations

99:59:59.999 --> 99:59:59.999
(Narrator) of the same word.
- (Man) Tomato - (Woman) Tomato

99:59:59.999 --> 99:59:59.999
(Narrator) Speech recognition continued
in the 80s with the introduction of the

99:59:59.999 --> 99:59:59.999
(Narrator) Hidden Markov Model, which
used a more mathematical approach

99:59:59.999 --> 99:59:59.999
(Narrator) to analyzing sound waves that
led to many breakthroughs we have today.

99:59:59.999 --> 99:59:59.999
You're taking in very raw audio wave forms

99:59:59.999 --> 99:59:59.999
like you get through a microphone