0
1
00:00:00,720 --> 00:00:08,190
Our sense organs interact with the outer world and send the visual and sound information to the neurons,
1

2
00:00:08,670 --> 00:00:13,740
let's say you are looking at this picture and your neurons want to decide.
2

3
00:00:13,770 --> 00:00:15,900
if your reaction would be laugh or not.
3

4
00:00:16,050 --> 00:00:26,130
By looking at this picture, each neuron get fired or activated only when its respective criteria are
4

5
00:00:26,130 --> 00:00:26,700
met.
5

6
00:00:26,820 --> 00:00:30,640
Or we can say each of these neurons have a specific tolerance.
6

7
00:00:30,660 --> 00:00:34,820
So if the tolerance is met, then it would be fired or activated.
7

8
00:00:35,100 --> 00:00:38,610
And then at the result we can decide to laugh or not.
8

9
00:00:39,100 --> 00:00:46,020
But actually, in reality, it's not only a few neurons who decide to laugh to show a reaction to one
9

10
00:00:46,020 --> 00:00:46,540
event.
10

11
00:00:46,560 --> 00:00:54,030
Of course, there are millions of neurons interconnected to Chander and they are making a hierarchical
11

12
00:00:54,030 --> 00:00:56,990
decision based on what they receive.
12

13
00:00:57,000 --> 00:01:03,030
It can be any signal like a visual signal, look, a sound signal, or like other sense signals.
13

14
00:01:03,360 --> 00:01:11,440
Now the sense organs pass the information to the first or lowest layer of neurons to process it, and
14

15
00:01:11,440 --> 00:01:18,960
the outputs of these processes are passed on the next layers in a hierarchical manner.
15

16
00:01:19,410 --> 00:01:22,420
Some of the neurons will fire and some won't.
16

17
00:01:22,440 --> 00:01:29,960
And this process goes on until the result is finally the response, which in this case is.
17

18
00:01:30,840 --> 00:01:35,700
This massive parallel network also ensure that there is a division of work.
18

19
00:01:36,090 --> 00:01:42,660
Each neuron only fired and activated when its specific criteria are met.
19

20
00:01:43,770 --> 00:01:50,670
It is believed that neurons are our age in a hierarchical fashion and each layer has its own role and
20

21
00:01:50,670 --> 00:01:53,360
responsibility. To detect the face,
21

22
00:01:53,370 --> 00:02:00,870
it's not only one neuron to decide whether this is a face or not, but there are hierarchical and different
22

23
00:02:00,870 --> 00:02:04,920
layers of neurons to decide whether it is a face or not.
23

24
00:02:05,340 --> 00:02:12,380
In this example, the first layer, let's say, to take the edge and corners if it criterium met, it
24

25
00:02:12,380 --> 00:02:20,700
will pass to the second layer, which will decide on the form feature of this group's nose, mouth and
25

26
00:02:20,700 --> 00:02:28,020
eyes and the third layer detect the high-level objects, like if this is a face or not and puts everything
26

27
00:02:28,020 --> 00:02:28,560
together.
27

28
00:02:29,010 --> 00:02:31,080
We are using the same criteria.
28

29
00:02:31,080 --> 00:02:38,640
we're using the same method for machines as well to teach them how to detect a specific animal, a specific
29

30
00:02:38,640 --> 00:02:40,420
object or a face.
30

31
00:02:40,740 --> 00:02:47,300
This is an example of deep learning to detect either this animal is an elephant or not.
31

32
00:02:47,330 --> 00:02:50,490
There are several layers. suppose,
32

33
00:02:50,490 --> 00:02:55,340
one layer is for detecting the tail, another one is for detecting the ears.
33

34
00:02:55,350 --> 00:02:58,830
and finally, we can have the full shape of an elephant.