WEBVTT
Kind: captions
Language: en

1
00:00:00.080 --> 00:00:02.640
The polling site for this&nbsp;
video was powered by Fasthosts.

2
00:00:02.640 --> 00:00:04.240
UK viewers can enter&nbsp;
their competition to win

3
00:00:04.240 --> 00:00:07.360
a tech bundle and dream PC&nbsp;
setup worth up to £5,000.

4
00:00:08.000 --> 00:00:09.040
There's a link in the description,

5
00:00:09.040 --> 00:00:10.960
and a question at the end of the video.

6
00:00:11.520 --> 00:00:14.160
This isn't the first time that&nbsp;
this question's been asked.

7
00:00:14.160 --> 00:00:16.000
Several web projects have&nbsp;
tried to answer it before,

8
00:00:16.000 --> 00:00:17.840
some are still running,&nbsp;
and some are defunct.

9
00:00:17.840 --> 00:00:20.400
And the question itself has&nbsp;
been used as a punchline,

10
00:00:20.400 --> 00:00:23.680
asked by an inept, self-obsessed&nbsp;
radio DJ character.

11
00:00:23.680 --> 00:00:26.000
"What is... the best thing?"

12
00:00:26.960 --> 00:00:30.400
Of course there isn't a meaningful&nbsp;
answer to that question. Of course.

13
00:00:30.400 --> 00:00:32.560
But I think there are&nbsp;
some really interesting&nbsp;&nbsp;

14
00:00:32.560 --> 00:00:34.880
challenges in trying to&nbsp;
find an answer anyway,

15
00:00:34.880 --> 00:00:37.920
and the results can reveal a&nbsp;
lot more than you might think.

16
00:00:38.720 --> 00:00:40.880
The first problem is&nbsp;
trying to list everything.

17
00:00:40.880 --> 00:00:42.640
I know it's an obvious&nbsp;
thing to say, but:

18
00:00:42.640 --> 00:00:44.560
there are a lot of&nbsp;
things in the universe.

19
00:00:44.560 --> 00:00:46.960
So let's reduce the scope to&nbsp;&nbsp;

20
00:00:46.960 --> 00:00:50.080
"everything that most people&nbsp;
could form an opinion about".

21
00:00:50.640 --> 00:00:53.040
How do you get a list&nbsp;
of everything like that?

22
00:00:53.600 --> 00:00:56.160
Well, the starting point is Wikidata.

23
00:00:56.160 --> 00:00:59.680
Everything that has a Wikipedia&nbsp;
article also has an entry in Wikidata,

24
00:00:59.680 --> 00:01:01.680
but so does every category of things,&nbsp;&nbsp;

25
00:01:01.680 --> 00:01:03.600
every property that&nbsp;
something might have,

26
00:01:03.600 --> 00:01:06.960
and every link and connection&nbsp;
between all of those.

27
00:01:06.960 --> 00:01:09.440
And it's all designed to&nbsp;
be processed by computers.

28
00:01:09.440 --> 00:01:12.560
So I figured I'd start&nbsp;
by downloading it.

29
00:01:12.560 --> 00:01:14.520
More than a terabyte of&nbsp;data,

30
00:01:14.520 --> 00:01:17.600
more than 88 million things
exhaustively described.

31
00:01:18.480 --> 00:01:21.200
And most of those things&nbsp;
are not interesting.

32
00:01:21.200 --> 00:01:24.080
More than that, they're going to&nbsp;
be a mystery to almost everyone.

33
00:01:24.080 --> 00:01:26.720
Every named location in the&nbsp;world
no matter how obscure,

34
00:01:26.720 --> 00:01:30.800
every species and genus of animal,&nbsp;
enormous numbers of scholarly articles.

35
00:01:30.800 --> 00:01:35.360
If you show most of those to people&nbsp;
and ask them to form an opinion,

36
00:01:35.360 --> 00:01:37.760
the answer isn't just&nbsp;
going to be "I don't know":

37
00:01:37.760 --> 00:01:39.920
it'll be "I don't care".

38
00:01:39.920 --> 00:01:42.640
So I had to filter&nbsp;
those 88 million things.

39
00:01:42.640 --> 00:01:45.200
And the first steps were&nbsp;
actually kind of easy.

40
00:01:45.200 --> 00:01:47.440
First, I removed all&nbsp;
listings for people.

41
00:01:47.440 --> 00:01:51.280
We're ranking things, so someone&nbsp;
else can do "who is the best person".

42
00:01:51.280 --> 00:01:52.000
You're welcome to that.

43
00:01:52.000 --> 00:01:55.680
But I just removed any item that&nbsp;
was tagged with Q5, "human".

44
00:01:55.680 --> 00:01:57.520
And that's good, that's a good start.

45
00:01:57.520 --> 00:02:00.640
Also, I removed all listings for&nbsp;
groups of people, because: yikes.

46
00:02:01.200 --> 00:02:03.920
Next, places. If you're doing&nbsp;
"what is the best thing",

47
00:02:03.920 --> 00:02:06.080
no country or river or&nbsp;building
is ever going to win,

48
00:02:06.080 --> 00:02:09.760
it'll get voted down by political&nbsp;rivalries
or people elsewhere&nbsp; who've never heard of it,

49
00:02:09.760 --> 00:02:12.320
so if the item was tagged&nbsp;
with a latitude and longitude,

50
00:02:12.960 --> 00:02:14.560
it also got thrown out.

51
00:02:14.560 --> 00:02:17.680
Also, anything tagged as&nbsp;
fictional got removed too:

52
00:02:17.680 --> 00:02:19.040
not works of art themselves,&nbsp;&nbsp;

53
00:02:19.040 --> 00:02:21.840
but characters and events&nbsp;
that aren't part of reality.

54
00:02:22.400 --> 00:02:25.440
That still left an&nbsp;
enormous number of items.

55
00:02:25.440 --> 00:02:28.640
But we're only looking for things&nbsp;
that most people will know about,

56
00:02:28.640 --> 00:02:30.480
and there's a really&nbsp;
good metric for that:

57
00:02:30.480 --> 00:02:35.920
I kept anything that had a&nbsp;Wikipedia article
in at least fifty different languages.

58
00:02:35.920 --> 00:02:37.520
I tried different thresholds for that,

59
00:02:37.520 --> 00:02:40.320
but fifty seemed to have&nbsp;
the right balance where

60
00:02:40.320 --> 00:02:44.240
almost everything that remained&nbsp;
would be recognisable to most people.

61
00:02:44.800 --> 00:02:51.440
And that brought it down to 8,850&nbsp;things.
Which is a managable number.

62
00:02:52.400 --> 00:02:55.520
But there was no way to&nbsp;
automate the last part.

63
00:02:55.520 --> 00:03:00.320
I had to manually check through all&nbsp;
those thousands of things to find&nbsp;&nbsp;

64
00:03:00.320 --> 00:03:01.120
the bad ones.

65
00:03:01.760 --> 00:03:04.080
Not just things that most people&nbsp;
would vote against because they're&nbsp;&nbsp;

66
00:03:04.080 --> 00:03:05.280
unpleasant or harmful,

67
00:03:05.280 --> 00:03:09.760
but things that no-one should be asking&nbsp;
about in a lighthearted web poll.

68
00:03:09.760 --> 00:03:10.991
Crimes against the person.

69
00:03:10.991 --> 00:03:14.640
A couple of disturbing things that were&nbsp;
just listed as "rituals".

70
00:03:14.640 --> 00:03:17.920
Anything to do with the Nazis.&nbsp;
Which it turns out is quite a lot.

71
00:03:17.920 --> 00:03:21.680
They kept showing up under&nbsp;
apparently-innocent categories?

72
00:03:21.680 --> 00:03:25.520
Like, eugenics was just&nbsp;
tagged as "social philosophy".

73
00:03:25.520 --> 00:03:28.000
Mein Kampf, just listed&nbsp;
as "written work".

74
00:03:28.000 --> 00:03:33.200
Unless you kept constantly vigilant&nbsp;
for them, they kept trying to sneak in.

75
00:03:34.160 --> 00:03:36.000
Then there were the dull&nbsp;
groups of things that could&nbsp;&nbsp;

76
00:03:36.000 --> 00:03:38.160
be summed up in a single entry instead.

77
00:03:38.160 --> 00:03:42.000
Every time zone. Every language,
every&nbsp; country's flag and national anthem.

78
00:03:42.000 --> 00:03:44.720
Every individual book&nbsp;
of every religious text.

79
00:03:44.720 --> 00:03:50.000
A lot of mythological figures who&nbsp;weren't
tagged as either "human"&nbsp;or "fictional".

80
00:03:50.640 --> 00:03:53.040
Hundreds of generic names of galaxies--

81
00:03:53.040 --> 00:03:55.920
"Okay, you know what? I talked&nbsp;
about that for far too long.

82
00:03:55.920 --> 00:03:58.320
"Let's just say I removed the boring&nbsp;
ones, okay? There were a lot of them.

83
00:03:58.320 --> 00:03:59.520
"Let's skip forward."

84
00:03:59.520 --> 00:04:01.360
And then, there was the vandalism.

85
00:04:01.360 --> 00:04:04.640
All of which has since been corrected,&nbsp;
but in the snapshot I downloaded,

86
00:04:04.640 --> 00:04:10.000
someone had replaced the title of&nbsp;
"graphics" with "Pro player de fifa"

87
00:04:10.000 --> 00:04:13.920
and the description of&nbsp;
"worm" with "dog go fishing".

88
00:04:13.920 --> 00:04:17.520
Also, "pipe organ" was described as&nbsp;
"wind instrument that causes cancer".

89
00:04:17.520 --> 00:04:21.600
So there's someone out there who&nbsp;
really, really doesn't like pipe organs.

90
00:04:22.320 --> 00:04:26.720
When all was done: 7,188 things.

91
00:04:26.720 --> 00:04:28.000
I knew it wasn't going to be perfect,

92
00:04:28.000 --> 00:04:30.000
people would still find&nbsp;mistakes,
and they did.

93
00:04:30.560 --> 00:04:34.880
But it was good enough. It was time&nbsp;
to ask the world which was best.

94
00:04:36.560 --> 00:04:39.120
One of the best approaches&nbsp;
for ranking items in a list

95
00:04:39.120 --> 00:04:40.880
is to show them to people two at a time,

96
00:04:40.880 --> 00:04:43.680
and then ask them to pick&nbsp;
the best of each random pair.

97
00:04:43.680 --> 00:04:46.000
The best ones will be&nbsp;
consistently voted for,

98
00:04:46.000 --> 00:04:47.600
and the worst ones voted against.

99
00:04:47.600 --> 00:04:50.000
And as long as you have&nbsp;
enough votes in total,

100
00:04:50.000 --> 00:04:52.320
you don't need to keep track&nbsp;
of all the different pairings:

101
00:04:52.320 --> 00:04:55.840
just the total number of wins&nbsp;
and losses for each item.

102
00:04:55.840 --> 00:04:57.440
Now, I've written code&nbsp;
to do that before,

103
00:04:57.440 --> 00:04:59.600
so I just reused it, put&nbsp;
a quick site together,&nbsp;&nbsp;

104
00:04:59.600 --> 00:05:00.960
and launched it out on Twitter.

105
00:05:01.920 --> 00:05:06.080
My code broke immediately&nbsp;
because I'd forgotten to change&nbsp;&nbsp;

106
00:05:06.080 --> 00:05:07.600
one line before going live,

107
00:05:07.600 --> 00:05:11.360
I fixed it within a minute or so while&nbsp;
a hundred people rushed to tweet me

108
00:05:11.360 --> 00:05:14.160
about something I obviously&nbsp;
already knew about. Anyway. So.

109
00:05:14.160 --> 00:05:18.880
Five hours and more than&nbsp;
1.2 million votes later,

110
00:05:18.880 --> 00:05:20.560
the order of items had settled down,

111
00:05:20.560 --> 00:05:24.480
and I closed the poll before anyone&nbsp;
wrote code to try and break it.

112
00:05:24.480 --> 00:05:27.680
Now, you'll remember that each&nbsp;
pairing was randomly chosen.

113
00:05:27.680 --> 00:05:31.840
That means some items had more match-ups&nbsp;
than others, just through sheer luck.

114
00:05:31.840 --> 00:05:35.200
The outliers were "mold",
which was in 125 match-ups,

115
00:05:35.200 --> 00:05:37.520
and "canal", which was in 236.

116
00:05:37.520 --> 00:05:40.800
There was the expected&nbsp;
distribution between those.

117
00:05:40.800 --> 00:05:44.080
So, at this point, we&nbsp;
had ranked everything.

118
00:05:45.920 --> 00:05:47.840
I don't want to spend too much&nbsp;
time on the bottom of the list.

119
00:05:47.840 --> 00:05:51.120
It's a lot of nasty diseases&nbsp;
and unpleasant concepts.

120
00:05:51.120 --> 00:05:53.040
Also one of the Twilight movies.

121
00:05:53.040 --> 00:05:56.400
I will say that The Worst&nbsp;
Thing... is Lyme disease.

122
00:05:56.400 --> 00:05:57.502
I've no idea why.

123
00:05:57.502 --> 00:06:00.400
It did significantly worse than
everything&nbsp;else, by a good margin.

124
00:06:00.400 --> 00:06:02.640
Maybe, statistically, out&nbsp;
of the thousands of items,

125
00:06:02.640 --> 00:06:04.640
one had to get a lot&nbsp;
of unlucky matchups?

126
00:06:04.640 --> 00:06:09.040
But, honestly, it is a really&nbsp;
long way below any other item.

127
00:06:09.040 --> 00:06:10.720
"Coronavirus", also fairly low.

128
00:06:10.720 --> 00:06:13.920
And anything religious did&nbsp;
quite poorly, which makes sense:

129
00:06:13.920 --> 00:06:15.520
if you're not religious&nbsp;
you're rarely going to&nbsp;&nbsp;

130
00:06:15.520 --> 00:06:18.240
vote for anything to do with faith,
and if you are religious&nbsp;&nbsp;

131
00:06:18.240 --> 00:06:22.000
you're hardly ever going to vote&nbsp;
anything other than your own faith.

132
00:06:22.000 --> 00:06:24.320
In hindsight, I should have&nbsp;
done something like consolidate

133
00:06:24.320 --> 00:06:28.240
all the entries for faiths into&nbsp;
one just called "religion".

134
00:06:28.240 --> 00:06:31.360
Which I'm sure wouldn't have&nbsp;
caused me any problems at all.

135
00:06:31.360 --> 00:06:33.840
Anyway. The best things.

136
00:06:33.840 --> 00:06:34.720
First, let's be clear:&nbsp;&nbsp;

137
00:06:34.720 --> 00:06:37.520
these are the results as voted by&nbsp;
the people who follow me on Twitter.

138
00:06:37.520 --> 00:06:41.520
This is about "the best thing"&nbsp;
as decided by, if we're honest,

139
00:06:41.520 --> 00:06:45.440
a group of English-speaking,&nbsp;
extremely-online nerds.

140
00:06:45.440 --> 00:06:47.920
However, that's also going to be a lot&nbsp;
of the people who watch this video,

141
00:06:47.920 --> 00:06:49.520
so, I think it's fair to say,&nbsp;&nbsp;

142
00:06:49.520 --> 00:06:53.440
as voted by you: here are&nbsp;
the top ten best things.

143
00:06:53.440 --> 00:06:55.280
At number 10, privacy.

144
00:06:55.280 --> 00:06:58.560
And ranked above privacy,&nbsp;
at number 9, pizza.

145
00:06:58.560 --> 00:07:02.800
Is pizza better or more important&nbsp;
than privacy? [indecisive noise]

146
00:07:02.800 --> 00:07:06.400
...but pizza is more likely to win a&nbsp;
match-up, and that's what counts here!

147
00:07:06.400 --> 00:07:09.520
By the way, the next highest&nbsp;
food was ice cream, at number 43,

148
00:07:09.520 --> 00:07:13.120
and while that could imply that
my audience have the palates of&nbsp;five-year-olds,

149
00:07:13.120 --> 00:07:16.640
I think it's more that, while those&nbsp;
may not be everyone's favourite foods,

150
00:07:16.640 --> 00:07:19.840
there are very few people&nbsp;
who actively dislike them,

151
00:07:19.840 --> 00:07:22.960
so they'll win a lot of generic&nbsp;
match-ups just because of that.

152
00:07:22.960 --> 00:07:26.160
The next items up: knowledge,&nbsp;
creativity and logic.

153
00:07:26.160 --> 00:07:30.800
The foundations of human thought.
Given&nbsp;my audience, that makes a lot of sense.

154
00:07:30.800 --> 00:07:32.800
At number 5: hugs.

155
00:07:32.800 --> 00:07:35.360
Which Wikidata clinically describes as&nbsp;&nbsp;

156
00:07:35.360 --> 00:07:38.800
"a form of endearment,&nbsp;
universal in human communities".

157
00:07:39.440 --> 00:07:41.440
Granted, it's 2020 as&nbsp;
this video goes out,

158
00:07:41.440 --> 00:07:43.760
so they're less universal than&nbsp;
they perhaps should be right now,

159
00:07:43.760 --> 00:07:45.280
but that's still lovely.

160
00:07:45.840 --> 00:07:50.000
Then we get to three items that I&nbsp;
honestly wasn't expecting to be so high.

161
00:07:50.000 --> 00:07:52.000
At number 4: gravity.

162
00:07:52.960 --> 00:07:55.200
Sure, it's essential for&nbsp;
the entire universe to work,

163
00:07:55.200 --> 00:07:58.400
I just didn't expect it to beat&nbsp;
"hugs". And then, at number 3...

164
00:07:59.040 --> 00:08:00.560
...the Earth's magnetic field.

165
00:08:02.400 --> 00:08:04.400
Like I said, extremely-online nerds.

166
00:08:04.400 --> 00:08:06.560
Because, again, yes,&nbsp;
essential for life to exist,

167
00:08:06.560 --> 00:08:11.280
but just to be clear, "air" and "fresh&nbsp;
water" only just made the top 25,

168
00:08:11.280 --> 00:08:14.080
and somehow the Earth's&nbsp;
magnetic field is at number 3.

169
00:08:14.080 --> 00:08:17.360
And it's at this point that I really&nbsp;
start to doubt my own methodology.

170
00:08:17.360 --> 00:08:19.200
Because at number two is electricity.

171
00:08:19.760 --> 00:08:23.280
I do realise that using an&nbsp;
electronic device to run this poll

172
00:08:23.280 --> 00:08:26.720
does give that a certain&nbsp;
advantage, but again,

173
00:08:26.720 --> 00:08:28.240
should that really be higher than air?

174
00:08:29.520 --> 00:08:31.040
Before we get to the best thing, though,

175
00:08:31.040 --> 00:08:35.120
here are some other interesting&nbsp;
results in specific categories:

176
00:08:35.120 --> 00:08:37.360
the best part of the body is the brain.

177
00:08:37.360 --> 00:08:41.120
"Space" and "time" both fought and won&nbsp;
exactly the same number of match-ups,

178
00:08:41.120 --> 00:08:43.920
they landed in joint 36th place.

179
00:08:43.920 --> 00:08:47.040
Despite there being quite a few&nbsp;
things about sex in the list,

180
00:08:47.040 --> 00:08:49.200
none of them got near the top 50.

181
00:08:49.200 --> 00:08:52.640
"Okay, okay, I should have checked more&nbsp;
than the top 50 before recording this,

182
00:08:52.640 --> 00:08:56.480
"'cos it turns out that the&nbsp;
highest-rated sex thing is 'orgasm',

183
00:08:56.480 --> 00:09:00.000
"and it got to number 69, and I&nbsp;
swear I'm not making that up."

184
00:09:00.000 --> 00:09:04.960
The best creatures are bees, then&nbsp;
emperor penguins, then hedgehogs.

185
00:09:04.960 --> 00:09:08.720
The best colours are black, then&nbsp;
blue, and the worst is brown.

186
00:09:08.720 --> 00:09:12.000
Love doesn't even make it into&nbsp;
the top 100, it's down at 137,

187
00:09:12.000 --> 00:09:14.320
next to Vitamin C and cryptography,

188
00:09:14.320 --> 00:09:17.680
and if that doesn't prove my audience&nbsp;
isn't representative of the wider world,

189
00:09:17.680 --> 00:09:19.040
I don't know what does.

190
00:09:19.040 --> 00:09:20.800
Actually, I do know what does,&nbsp;&nbsp;

191
00:09:20.800 --> 00:09:24.080
heterosexuality lost more&nbsp;
than 50% of its match-ups,

192
00:09:24.080 --> 00:09:27.920
while bisexuality was ranked&nbsp;
only one item below doctors.

193
00:09:27.920 --> 00:09:31.600
Yes, it is ridiculous to try&nbsp;
and rank everything like this.

194
00:09:31.600 --> 00:09:34.960
But the results do reveal things&nbsp;
about this group of people,

195
00:09:34.960 --> 00:09:37.280
about the folks who tend&nbsp;
to watch videos like this.

196
00:09:37.280 --> 00:09:40.880
And perhaps the most revealing&nbsp;
thing is what placed first.

197
00:09:40.880 --> 00:09:43.920
It doesn't just tell you about the&nbsp;
needs and desires of this audience,

198
00:09:43.920 --> 00:09:46.560
it's also something about&nbsp;
the times we're living in.

199
00:09:46.560 --> 00:09:49.280
If we weren't in what seems&nbsp;
to be such a rough year,

200
00:09:49.280 --> 00:09:52.960
if I were giving this talk to a live&nbsp;
audience, like I originally planned to,

201
00:09:52.960 --> 00:09:55.200
instead of a standing in&nbsp;
front of a green screen and

202
00:09:55.200 --> 00:09:57.520
talking to a camera in a tiny apartment,

203
00:09:57.520 --> 00:10:00.400
well, then, in that case maybe the&nbsp;
results would have been different.

204
00:10:00.960 --> 00:10:05.520
But the best thing, according&nbsp;
to this audience in mid-2020:

205
00:10:05.520 --> 00:10:06.240
the best thing is...

206
00:10:07.600 --> 00:10:08.100
sleep.

207
00:10:08.640 --> 00:10:10.080
Have a good night, folks.

208
00:10:10.080 --> 00:10:12.480
I ran the polling site for&nbsp;
this video on Fasthosts,

209
00:10:12.480 --> 00:10:15.520
a web hosting company with more&nbsp;
than twenty years' experience.

210
00:10:15.520 --> 00:10:19.520
Their dedicated servers can have up&nbsp;to
10Gbps connectivity&nbsp;and unlimited bandwidth,

211
00:10:19.520 --> 00:10:20.960
and their CloudNX platform&nbsp;&nbsp;

212
00:10:20.960 --> 00:10:23.280
lets you configure and&nbsp;
scale your hosting hardware

213
00:10:23.280 --> 00:10:25.760
in real time with no upfront costs.

214
00:10:25.760 --> 00:10:28.000
All of their servers and&nbsp;
engineers are based in the UK.

215
00:10:28.000 --> 00:10:32.160
And if you are too, then you can go the&nbsp;
link on screen or in the description

216
00:10:32.160 --> 00:10:35.760
to enter their competition to win&nbsp;
a tech bundle and dream PC setup&nbsp;&nbsp;

217
00:10:35.760 --> 00:10:37.360
worth up to £5,000.

218
00:10:38.800 --> 00:10:41.200
If you can answer the techie test&nbsp;
question they asked me to write.

219
00:10:41.200 --> 00:10:44.720
Which is: what's the HTTP&nbsp;
response code for 'OK'?

220
00:10:45.280 --> 00:10:46.880
Terms and conditions&nbsp;
are over on their site,

221
00:10:46.880 --> 00:10:59.840
the closing date is 31st&nbsp;
October 2020: good luck!

