0:36

Exactly kind of what we're talking about is how that

0:38

shit is going to take my job . So

0:40

tell me how it's different

0:42

to train a neural network . It sounds

0:44

like we're talking about is training a neural network

0:46

and what's different about

0:49

training a neural network versus training

0:51

the human brain , for instance ? Just as a fun

0:53

comparison .

0:55

Well , I would say they are very

0:57

similar approaches because

0:59

neural networks was invented back in

1:01

the middle of 20th century

1:03

, so it was invented in 1940s

1:06

that the concept of neural networks and

1:09

I would say that to

1:11

train a human and train artificial

1:13

neural network is something similar . So

1:16

the overall idea is that you need

1:18

to show different examples and

1:20

say something like hey , these are good

1:23

examples , these are bad examples . And

1:25

if you multiply it by millions

1:27

or trillions examples , then

1:30

any human can understand

1:32

what to do . And the same

1:34

goes with neural networks networks

1:45

. So neural neural networks was invented as a kind of way how we can uh simulate

1:47

uh our brain inside the computers . So they are pretty much uh similar to

1:50

what we have in our brain . So in fact

1:52

I can show you the picture here which

1:55

is like quite common . So

1:57

here how neural networks

1:59

looks in our brain . Uh

2:01

, it's uh contains uh synapses

2:04

and neurons . So these big guys are

2:06

neurons and these uh like threads

2:08

or like connections , they are synapses

2:11

. And in like mathematical

2:13

world or in computer world , it's

2:16

uh becomes a kind of a

2:18

complex equation where we

2:20

just sum

2:22

and multiply some numbers and

2:24

get some results . So all in all

2:27

, to answer your question

2:29

, how , like , human training is different

2:31

from neural network training

2:33

. They are quite similar

2:36

to each other . Similar to each

2:38

other . The question is because

2:40

human brains at

2:42

least right now they

2:45

are much more advanced than existing neural networks . Neural

2:47

networks , they just need yeah , they just

2:49

need more like examples , more

2:52

training . That's maybe the

2:54

main difference , I would say .

2:58

That's fascinating the way you said that . So may

3:01

I ask a follow up question to

3:04

how you just compare them ? What

3:06

is the metric that you would use to

3:08

compare them ? So if we're going to

3:10

compare ,