1
00:00:00,000 --> 00:00:05,360
Moore's Law sounds like a comforting, or perhaps disturbing, promise that computer

2
00:00:05,360 --> 00:00:08,920
processing power will double about every two years.

3
00:00:08,920 --> 00:00:13,360
But despite the fact we call it a law, there have been plenty of folks in the tech center

4
00:00:13,360 --> 00:00:17,040
opining that Moore's Law is in fact dead.

5
00:00:17,040 --> 00:00:20,760
And we're hitting some kind of wall when it comes to just how powerful we can make

6
00:00:20,760 --> 00:00:25,800
our technology. But how much truth is there to the rumor, and who the heck is Moore anyway?

7
00:00:25,800 --> 00:00:30,680
It turns out Gordon Moore wasn't just some guy that made an observation about CPUs.

8
00:00:30,680 --> 00:00:38,360
He was one of the founders of Intel. His story begins back in 1957, a few years after he finished his PhD at Caltech.

9
00:00:38,360 --> 00:00:44,080
At the time he was working for Nobel Prize winner William Shockley on semiconductor development.

10
00:00:44,080 --> 00:00:49,400
But Shockley wasn't exactly the best boss, with his management style characterized by

11
00:00:49,400 --> 00:00:52,520
paranoia, aggressiveness, and intrusiveness.

12
00:00:52,520 --> 00:01:00,040
The trinity of bad boss qualities. So maybe it's not surprising that he was later disgraced for espousing views that could

13
00:01:00,040 --> 00:01:04,040
only be described as, uh, alarmingly racist.

14
00:01:04,040 --> 00:01:09,600
Moore and seven of his colleagues left Shockley's company and started a new firm called Fairchild

15
00:01:09,600 --> 00:01:15,240
Semiconductor. This walkout became so famous that Moore's group became known in tech circles as The

16
00:01:15,240 --> 00:01:18,600
Traderous Eight.

17
00:01:18,600 --> 00:01:22,400
Ravaging towns across Silicon Valley.

18
00:01:22,400 --> 00:01:26,920
Fairchild went on to develop the first commercially viable integrated circuit, and although the

19
00:01:26,920 --> 00:01:32,080
company is no longer around, its employees went on to found other influential tech firms,

20
00:01:32,080 --> 00:01:39,320
including both AMD and Intel. It was also at Fairchild that Moore thought up the first formulation of his famous law

21
00:01:39,320 --> 00:01:44,640
in 1965, that the number of components in an integrated circuit would double yearly.

22
00:01:44,640 --> 00:01:49,200
Along with fellow Traderous Eightster, Robert...nice.

23
00:01:49,200 --> 00:01:54,880
Moore left Fairchild in 1968 to found a new company they originally thought about naming,

24
00:01:54,880 --> 00:02:01,200
Moore Noise. But since that sounded too much like Moore Noise, they decided that wasn't the best brand

25
00:02:01,200 --> 00:02:06,120
for an electronics firm. And I wish I could go back in time to tell them they're making a huge mistake.

26
00:02:06,120 --> 00:02:09,560
Instead, they went with NM Electronics.

27
00:02:09,560 --> 00:02:15,120
But quickly changed the name to Intel, less than a month after the company was founded.

28
00:02:15,120 --> 00:02:23,120
Sounded a little more...nice. While Moore's law prophesies exponential growth in transistor density, Moore himself

29
00:02:23,120 --> 00:02:27,360
later predicted that the rate of increase would eventually slow down.

30
00:02:27,360 --> 00:02:32,520
In 2005, Moore pointed out that we'd eventually hit some pretty hard physical limits, such

31
00:02:32,520 --> 00:02:38,720
as the size of an atom. And indeed, while we've seen experimental transistors that work by changing the position

32
00:02:38,720 --> 00:02:44,080
of only a single atom, there's still a debate raging as to whether this is the right way

33
00:02:44,080 --> 00:02:51,440
to look at Moore's law. One consideration is that there isn't a standardized way to determine how big transistors are.

34
00:02:51,440 --> 00:02:56,080
There already isn't much consistency between Intel and AMD, and with newer transistor types

35
00:02:56,080 --> 00:03:01,520
that involve three-dimensional designs and alternative materials, focusing on transistors

36
00:03:01,520 --> 00:03:05,120
per unit area might actually be missing the point.

37
00:03:05,120 --> 00:03:09,000
Instead, current discussions on Moore's law tend to be more about whether we're going

38
00:03:09,000 --> 00:03:13,800
to continue to see a certain amount of processing power gradually become cheaper.

39
00:03:13,800 --> 00:03:19,400
Jensen Wong, CEO of NVIDIA, famously said last year that he believes Moore's law to

40
00:03:19,400 --> 00:03:24,920
be truly dead, as silicon wafer costs had risen to the point where consumers could no

41
00:03:24,920 --> 00:03:29,720
longer count on getting significantly more performance at a lower price if they waited

42
00:03:29,720 --> 00:03:32,760
a year or two. Jensen was simply making an observation.

43
00:03:32,760 --> 00:03:38,840
There's nothing he can do about it. But of course, that doesn't mean the price of a wafer is the only variable to think

44
00:03:38,840 --> 00:03:45,080
about. Between chiplet-based processors that save money over monolithic chips that get lower

45
00:03:45,080 --> 00:03:49,280
yields, new transistor designs like Gate All Around that can push through more electrons

46
00:03:49,280 --> 00:03:54,520
at once, and the rise of artificial intelligence, which aims to provide more performance through

47
00:03:54,520 --> 00:03:59,400
machine learning than we'd otherwise see just from doubling the number of transistors alone.

48
00:03:59,400 --> 00:04:04,840
So while we might not be able to keep making our chips impossibly chock-full of transistors,

49
00:04:04,840 --> 00:04:09,560
the tech industry still has plenty of tricks up its sleeve to make a future full of self-driving

50
00:04:09,600 --> 00:04:14,440
cars, hyper-realistic games, and disturbing deep fakes a reality.

51
00:04:14,440 --> 00:04:18,680
Somehow, I don't think computers will ever be powerful enough to make Twitter a more

52
00:04:18,680 --> 00:04:22,280
civil place. But you were powerful enough to make it to the end of this video.

53
00:04:22,280 --> 00:04:25,920
Hey, thanks for watching! Like the video if you liked it, dislike it if you disliked it.

54
00:04:25,920 --> 00:04:29,640
Check out our other videos, comment below with video suggestions, and don't forget to

55
00:04:29,640 --> 00:04:35,760
subscribe and follow TechWiki, where we will suggest things to you that maybe you hadn't

56
00:04:35,760 --> 00:04:36,280
thought of before.