1
00:00:00,000 --> 00:00:09,153
[MUSIC]

2
00:00:09,153 --> 00:00:10,498
Hi I'm Nick.

3
00:00:10,498 --> 00:00:16,300
In this course we're going to learn about
the broad topic area of machine learning.

4
00:00:16,300 --> 00:00:19,437
We'll explore some of the big ideas and,
toward the end,

5
00:00:19,437 --> 00:00:22,074
we'll even write a little
bit of code in Python.

6
00:00:22,074 --> 00:00:24,720
I can make some intelligent predictions.

7
00:00:24,720 --> 00:00:26,739
If you don't know Python already,

8
00:00:26,739 --> 00:00:31,131
I highly recommend you start with some
of the prerequisites for this course.

9
00:00:31,131 --> 00:00:33,485
You'll need to install Anaconda,

10
00:00:33,485 --> 00:00:39,320
which is a Python based platform focused
on data science and machine learning.

11
00:00:39,320 --> 00:00:42,160
We won't install Anaconda in these videos.

12
00:00:42,160 --> 00:00:44,740
So if you haven't installed it already,

13
00:00:44,740 --> 00:00:47,838
check out the notes associated
with this video for

14
00:00:47,838 --> 00:00:53,540
help on how to do that, along with some
other Python resources on Treehouse.

15
00:00:53,540 --> 00:00:58,352
Machine learning encompasses many
different ideas, programming languages,

16
00:00:58,352 --> 00:01:00,978
frameworks and approaches to the subject.

17
00:01:00,978 --> 00:01:07,640
So the term machine learning is difficult
to define in just a sentence or two.

18
00:01:07,640 --> 00:01:13,603
But essentially, machine learning
is giving a computer the ability

19
00:01:13,603 --> 00:01:18,760
to write its own rules and
learn about new things on its own.

20
00:01:18,760 --> 00:01:22,358
In this course, you're going
to hear a lot of new terms and

21
00:01:22,358 --> 00:01:25,050
ideas you may not be familiar with.

22
00:01:25,050 --> 00:01:29,010
It's a lot to take in and you might
not understand it all right away.

23
00:01:29,010 --> 00:01:30,150
That's normal.

24
00:01:30,150 --> 00:01:33,354
I was confused the first time I
learned about these concepts.

25
00:01:33,354 --> 00:01:37,919
I encourage you to take a look at
the teacher notes after each video and

26
00:01:37,919 --> 00:01:41,630
watch videos more than once
if you feel like you need to.

27
00:01:42,890 --> 00:01:47,251
That said, machine learning actually is
not as difficult as you might believe.

28
00:01:47,251 --> 00:01:50,640
And its applications are far reaching.

29
00:01:50,640 --> 00:01:55,242
You might be surprised where machine
learning could show up in your life and

30
00:01:55,242 --> 00:01:58,688
how it might be useful to you now and
in the coming years.

31
00:01:58,688 --> 00:02:00,753
Just to name a few applications,

32
00:02:00,753 --> 00:02:06,070
machine learning is already unlocking new
possibilities like self-driving cars,

33
00:02:06,070 --> 00:02:11,383
digital assistants and chat bots, and new
approaches to science and health care that

34
00:02:11,383 --> 00:02:17,340
can help us dig through vast libraries of
complicated information in an instance.

35
00:02:17,340 --> 00:02:22,757
Machine learning can be used to create
spam filters, search engines that can find

36
00:02:22,757 --> 00:02:28,340
your photos based a text description, and
even create original fine art and music.

37
00:02:28,340 --> 00:02:32,950
It's putting the power of human
reasoning into a computer.

38
00:02:32,950 --> 00:02:38,022
In the past, this topic has had other
labels like artificial intelligence,

39
00:02:38,022 --> 00:02:40,292
and that term still applies here.

40
00:02:40,292 --> 00:02:45,266
What's different, however, is that
past approaches to machine learning

41
00:02:45,266 --> 00:02:49,580
involved a human being writing most or
all of the rules.

42
00:02:49,580 --> 00:02:54,023
In the mid 90s, the IBM Deep Blue super
computer famously won a series of

43
00:02:54,023 --> 00:02:56,460
chess matches against Garry Kasparov,

44
00:02:56,460 --> 00:03:00,421
who many considered to be one of
the greatest chess players ever.

45
00:03:00,421 --> 00:03:04,860
This is an impressive accomplishment
even by today's standards.

46
00:03:04,860 --> 00:03:07,408
However, older approaches to artificial

47
00:03:07,408 --> 00:03:10,813
intelligence often involved
millions of lines of code.

48
00:03:10,813 --> 00:03:15,433
Writing if statements and conditional
logic to guide the computer through

49
00:03:15,433 --> 00:03:18,770
an incredible number of small decisions.

50
00:03:18,770 --> 00:03:23,143
I'm oversimplifying, but
in essence Deep Blue would look at many

51
00:03:23,143 --> 00:03:27,687
branching chess moves into the future,
far more than a human can do.

52
00:03:27,687 --> 00:03:34,080
And it would pick the best one based on
a complicated set of predefined criteria.

53
00:03:34,080 --> 00:03:37,782
This type of approach only had
the appearance of intelligence.

54
00:03:37,782 --> 00:03:38,997
When in reality,

55
00:03:38,997 --> 00:03:44,640
most AIs could be distilled down to
a giant flow chart created by people.

56
00:03:44,640 --> 00:03:48,464
And more importantly,
Deep Blue could only play chess.

57
00:03:48,464 --> 00:03:53,318
Ultimately, it's more desirable to
create a generic machine intelligence

58
00:03:53,318 --> 00:03:55,640
that can perform a variety of tasks.

59
00:03:56,790 --> 00:04:00,910
Put simply, machine learning is
still artificial intelligence but

60
00:04:00,910 --> 00:04:05,455
with a strong focus on the computer's
ability to write its own set of rules as

61
00:04:05,455 --> 00:04:06,960
it processes more data.

62
00:04:06,960 --> 00:04:10,310
Rather than humans teaching
it everything upfront.

63
00:04:10,310 --> 00:04:13,040
Using machine learning tools and
algorithms,

64
00:04:13,040 --> 00:04:17,625
we can explore data in new ways that would
otherwise take a ton of work by humans.

65
00:04:17,625 --> 00:04:22,380
Or even by traditional programming or
data processing approaches.

66
00:04:22,380 --> 00:04:24,735
There's one more thing I need to mention.

67
00:04:24,735 --> 00:04:29,030
While this technology has the potential
to be more impactful than even

68
00:04:29,030 --> 00:04:31,990
the microprocessor or the Internet.

69
00:04:31,990 --> 00:04:36,918
It also introduces many ethical questions
that, over the next few years and

70
00:04:36,918 --> 00:04:40,239
decades, will need answers
from people like you.

71
00:04:40,239 --> 00:04:43,169
This is a huge ongoing
discussion on its own and

72
00:04:43,169 --> 00:04:45,733
it's outside the scope of this course.

73
00:04:45,733 --> 00:04:51,006
But I highly recommend you check
the notes associated with this video for

74
00:04:51,006 --> 00:04:54,466
more resources and
insight into the ethical and

75
00:04:54,466 --> 00:04:57,856
societal implications of machine learning.

76
00:04:57,856 --> 00:05:02,329
Now let's continue with this broad
overview by taking a closer look at some

77
00:05:02,329 --> 00:05:05,390
of the different approaches
to machine learning.