1
00:00:00,06 --> 00:00:04,05
- [Narrator] Today's car comes with voice assist features.

2
00:00:04,05 --> 00:00:07,08
You can add an Alexa voice assistant to some cars,

3
00:00:07,08 --> 00:00:11,01
or you can use the voice assistant available in the car.

4
00:00:11,01 --> 00:00:12,09
You can talk to the car.

5
00:00:12,09 --> 00:00:14,06
Talk to the car?

6
00:00:14,06 --> 00:00:16,00
Can you think of some examples

7
00:00:16,00 --> 00:00:18,06
where you can talk to the car?

8
00:00:18,06 --> 00:00:20,09
Those are great ideas.

9
00:00:20,09 --> 00:00:23,04
You sure know how to talk to your car.

10
00:00:23,04 --> 00:00:26,07
We can use voice to talk to any voice assistant.

11
00:00:26,07 --> 00:00:30,07
They all have an invocation term like "Hey, Siri"

12
00:00:30,07 --> 00:00:32,09
or "Okay, Ford"

13
00:00:32,09 --> 00:00:34,04
or "Hey, Mercedes".

14
00:00:34,04 --> 00:00:36,06
Can you think why we need an invocation word

15
00:00:36,06 --> 00:00:39,05
when interacting with the voice interface?

16
00:00:39,05 --> 00:00:43,07
The invocation word wakes up the AI to listen.

17
00:00:43,07 --> 00:00:46,05
This ensures that the voice AI is not listening

18
00:00:46,05 --> 00:00:51,04
to us all the time simply to protect our privacy voice.

19
00:00:51,04 --> 00:00:55,08
AI also uses data and comes under machine learning

20
00:00:55,08 --> 00:01:00,04
and it works with sound and audio.

21
00:01:00,04 --> 00:01:02,04
Some of the things you can ask are looking

22
00:01:02,04 --> 00:01:04,00
for driving directions.

23
00:01:04,00 --> 00:01:05,07
Play your favorite music

24
00:01:05,07 --> 00:01:09,00
or ask the car to change the inside temperature.

25
00:01:09,00 --> 00:01:12,00
Mercedes allows you to start the car

26
00:01:12,00 --> 00:01:15,04
and warm up the car remotely in cold weather.

27
00:01:15,04 --> 00:01:16,09
Voice is an interface

28
00:01:16,09 --> 00:01:18,07
to do something you can do with a click

29
00:01:18,07 --> 00:01:20,01
or turning a knob,

30
00:01:20,01 --> 00:01:22,02
something that lends itself

31
00:01:22,02 --> 00:01:24,09
to a human-centric interface.

32
00:01:24,09 --> 00:01:29,01
So think of voice as data being utilized.

33
00:01:29,01 --> 00:01:31,02
Then think about the AI.

34
00:01:31,02 --> 00:01:35,02
The voice AI, what does it do for the users?

35
00:01:35,02 --> 00:01:37,02
And how to make this experience

36
00:01:37,02 --> 00:01:40,01
delight the user in the product design

37
00:01:40,01 --> 00:01:42,05
and to earn the user's trust?

38
00:01:42,05 --> 00:01:45,01
Finally, it's important to know that AI has

39
00:01:45,01 --> 00:01:47,03
to reside somewhere.

40
00:01:47,03 --> 00:01:50,04
This is where it becomes edge AI.

41
00:01:50,04 --> 00:01:55,01
AI can run like software from a web or mobile app,

42
00:01:55,01 --> 00:01:57,05
or a device or a car.

43
00:01:57,05 --> 00:02:02,04
Voice AI running on a car interface

44
00:02:02,04 --> 00:02:08,00
or a device interface becomes edge AI.

45
00:02:08,00 --> 00:02:12,00
Imagine this scenario, if the user wants to go

46
00:02:12,00 --> 00:02:16,02
to Rambla Square in Spain and they ask the car

47
00:02:16,02 --> 00:02:18,03
and it doesn't understand it,

48
00:02:18,03 --> 00:02:23,07
and it wonders if it is Rambla Square in Timbuktu.

49
00:02:23,07 --> 00:02:29,01
It is better for the AI application to ask respectfully

50
00:02:29,01 --> 00:02:31,08
and be aware of the user's time urgency,

51
00:02:31,08 --> 00:02:34,04
being respectful and clarifying

52
00:02:34,04 --> 00:02:37,04
their ask might be a better decision instead

53
00:02:37,04 --> 00:02:39,08
of driving them to Timbuktu.

54
00:02:39,08 --> 00:02:45,04
This is why the data input being filtered correctly matters.

55
00:02:45,04 --> 00:02:49,09
There may be a need for edge computing to process the data

56
00:02:49,09 --> 00:02:52,01
to understand different accents

57
00:02:52,01 --> 00:02:54,07
or to filter out background sounds

58
00:02:54,07 --> 00:02:57,09
of multiple people talking all at once.

59
00:02:57,09 --> 00:03:01,04
Machine learning models can do this very easily.

60
00:03:01,04 --> 00:03:02,09
I gave you voice AI as an example

61
00:03:02,09 --> 00:03:05,06
of edge AI that uses sound.

62
00:03:05,06 --> 00:03:09,06
In the next lesson, we will learn to build an edge AI

63
00:03:09,06 --> 00:03:13,07
that can be trained using voice or any audio sound,

64
00:03:13,07 --> 00:03:18,06
and you can expand it past the car to more edge AI devices.

65
00:03:18,06 --> 00:03:20,00
Let's go.