1
00:00:00,000 --> 00:00:06,000
Welcome to week three, where we discuss
listening.

2
00:00:07,040 --> 00:00:15,020
We all live in an ambient sea of data.
So whether or not we want to look for

3
00:00:15,020 --> 00:00:22,089
anything, we need to process this volume
of data continuously bombarding us.

4
00:00:24,032 --> 00:00:31,039
Nevertheless even as we walk around and
interact with our world we do get a sense

5
00:00:31,039 --> 00:00:36,044
of things.
When walking into a new building, we sort

6
00:00:36,044 --> 00:00:40,064
of get a smell of the place.
Is it a happy place?

7
00:00:40,064 --> 00:00:47,012
Is it a very strict environment?
And many other such impressions naturally

8
00:00:47,012 --> 00:00:51,093
form with our consciously wanting to
evaluate this.

9
00:00:54,034 --> 00:01:01,003
Similarly we are able to easily recognize
the familiar, as well as the rare and

10
00:01:01,003 --> 00:01:09,043
single it out for special detention.
What does all this have to do with the

11
00:01:09,043 --> 00:01:15,789
web?
Now if you think about it web properties,

12
00:01:15,789 --> 00:01:24,108
whether they are search engines, social
networks, auction sites, twitter,

13
00:01:24,108 --> 00:01:34,054
virtually anything live on advertising.
As a result, they want to process this

14
00:01:34,054 --> 00:01:41,511
large volume of data arising from our
interactions with them.

15
00:01:41,511 --> 00:01:49,194
To discern our intents so as to target us
with the right messages from their

16
00:01:49,194 --> 00:01:55,392
perspective and presumably from ours if we
find them useful.

17
00:01:55,392 --> 00:02:04,047
So they want to do things like recognize
who is actually shopping from those who

18
00:02:04,047 --> 00:02:10,941
are merely surfing, and treat the former
with a little more attention than the vast

19
00:02:10,941 --> 00:02:16,864
majority who might not actually be
shopping for anything.

20
00:02:16,864 --> 00:02:25,949
They want to figure out how to gauge our
opinions about products, services as well

21
00:02:25,949 --> 00:02:32,852
as almost anything under the sun.
In short they want to understand what

22
00:02:32,852 --> 00:02:40,593
people are saying or doing, much as in the
reverse tooling test we discussed last

23
00:02:40,593 --> 00:02:46,096
week, so that they can target the right
advertising message to them.

24
00:02:47,028 --> 00:02:55,090
So this week we'll study techniques that
allow these web properties, to learn

25
00:02:55,090 --> 00:03:04,064
information about our intentions from
whatever they are able to measure about

26
00:03:04,064 --> 00:03:09,022
us.
Learning information.

27
00:03:09,022 --> 00:03:17,084
So we will study machine learning from the
perspective of information theory.