1
00:00:00,000 --> 00:00:10,000
Welcome to Artificially Intelligent Marketing, a weekly podcast where we stay on top of the

2
00:00:10,000 --> 00:00:15,740
latest trends, tips and tools in the world of marketing AI, helping you get the best

3
00:00:15,740 --> 00:00:18,560
results from your marketing efforts.

4
00:00:18,560 --> 00:00:22,920
Now let's join our hosts, Paul Avery and Martin Broadhurst.

5
00:00:22,920 --> 00:00:29,000
Hello everybody, welcome to Artificially Intelligent Marketing, your favorite podcast for everything

6
00:00:29,000 --> 00:00:33,440
AI and marketing, where myself and my good friend Martin Broadhurst go through all of

7
00:00:33,440 --> 00:00:37,240
the things that happened in AI and marketing for the week, look at some tools of the week

8
00:00:37,240 --> 00:00:41,320
and just any other interesting information that we think you might find interesting to

9
00:00:41,320 --> 00:00:42,560
get on top of.

10
00:00:42,560 --> 00:00:44,880
How are you today, Martin?

11
00:00:44,880 --> 00:00:50,240
Well after the deluge of AI announcements over the past few weeks, I managed to actually

12
00:00:50,240 --> 00:00:52,440
read something that wasn't AI related.

13
00:00:52,440 --> 00:00:58,320
I managed to open a newspaper and look at the sports stories and some stuff like that.

14
00:00:58,320 --> 00:01:02,480
Apparently there's a world out there that isn't AI related.

15
00:01:02,480 --> 00:01:06,800
And you know when you read those stories, it's just Derby County bashing, isn't it?

16
00:01:06,800 --> 00:01:10,080
So is that really what you need more of that in your life?

17
00:01:10,080 --> 00:01:12,960
Yeah, for the most part it is that.

18
00:01:12,960 --> 00:01:16,520
It was yet another bleak weekend.

19
00:01:16,520 --> 00:01:18,840
We lost to Peterborough.

20
00:01:18,840 --> 00:01:21,480
That was terrible.

21
00:01:21,480 --> 00:01:24,720
Game of two halves, we were dominating in the first half but didn't score.

22
00:01:24,720 --> 00:01:26,960
So you know, that's football for you.

23
00:01:26,960 --> 00:01:34,200
That's also sort of what happened to Google in the AI race.

24
00:01:34,200 --> 00:01:36,760
Also a bleak weekend, a bleak end.

25
00:01:36,760 --> 00:01:38,240
I like the concept of that.

26
00:01:38,240 --> 00:01:40,480
That must be what it's like to be a Derby fan.

27
00:01:40,480 --> 00:01:41,480
Anyway, enough nonsense.

28
00:01:41,480 --> 00:01:42,480
Cool, cool.

29
00:01:42,480 --> 00:01:43,480
Well, let's get straight into it.

30
00:01:43,480 --> 00:01:44,480
What are we going to talk about today?

31
00:01:44,480 --> 00:01:49,680
We're going to talk about tech experts calling for a six month AI hiatus.

32
00:01:49,680 --> 00:01:56,080
We're going to talk about how Levi's has used AI generated models for some really cool stuff.

33
00:01:56,080 --> 00:01:57,080
Brian's going to take us through.

34
00:01:57,080 --> 00:02:02,080
We're going to talk about Bing has announced how it's going to try and weave advertising

35
00:02:02,080 --> 00:02:06,440
into its chat platform, which is like ChatGPT.

36
00:02:06,440 --> 00:02:11,600
We're going to look at some news that came out of the UK government about it's a white

37
00:02:11,600 --> 00:02:16,840
paper about how they're going to use AI to drive turbocharged growth.

38
00:02:16,840 --> 00:02:21,080
And we're going to look at some tweets that came out from one of the co-founders of HubSpot,

39
00:02:21,080 --> 00:02:24,760
Brian Halligan about AI and the future of content marketing.

40
00:02:24,760 --> 00:02:27,880
And then of course, we're going to have our tool of the week, which this week is going

41
00:02:27,880 --> 00:02:29,680
to be Go Charlie.

42
00:02:29,680 --> 00:02:30,960
Right.

43
00:02:30,960 --> 00:02:33,360
Let's get into story one.

44
00:02:33,360 --> 00:02:39,240
So yes, this open letter where a number of influential figures such as Elon Musk, Steve

45
00:02:39,240 --> 00:02:46,520
Wozniak and Andrew Yang have demanded a temporary hold of at least six months on AI system training

46
00:02:46,520 --> 00:02:48,200
and deployment.

47
00:02:48,200 --> 00:02:54,160
They contend that AI's increasing power is not matched by adequate safety measures, government

48
00:02:54,160 --> 00:02:59,480
regulations and strategies to address AI's far reaching societal effects.

49
00:02:59,480 --> 00:03:01,600
Here is a quote from the letter.

50
00:03:01,600 --> 00:03:06,560
Recent months have seen AI labs are locked in an out of control race to develop and deploy

51
00:03:06,560 --> 00:03:12,320
ever more powerful digital minds that no one, not even their creators, can understand, predict

52
00:03:12,320 --> 00:03:14,880
or reliably control.

53
00:03:14,880 --> 00:03:21,120
So this is a very interesting thing to see happen.

54
00:03:21,120 --> 00:03:26,080
I've got to be honest Brian, with my cynical hat on, I see that some of the most well known

55
00:03:26,080 --> 00:03:31,760
signage of this letter are competitors to open AI and that this could just be a guarded

56
00:03:31,760 --> 00:03:37,080
attempt to say, open AI, is there any chance you could just stop for six months while the

57
00:03:37,080 --> 00:03:38,680
rest of us catch up?

58
00:03:38,680 --> 00:03:40,720
What are your thoughts?

59
00:03:40,720 --> 00:03:41,720
Absolutely.

60
00:03:41,720 --> 00:03:46,240
The first thought that came to my mind, first of all, I think it's a, I get the concerns,

61
00:03:46,240 --> 00:03:47,240
right?

62
00:03:47,240 --> 00:03:50,080
There is, there are legitimate concerns and I'm sure we'll go into those, but the idea

63
00:03:50,080 --> 00:03:57,800
that we're going to go stop that for six months is, well frankly, crazy.

64
00:03:57,800 --> 00:04:06,680
Take it away from the corporate competition, but look at the kind of geopolitical landscape,

65
00:04:06,680 --> 00:04:07,680
right?

66
00:04:07,680 --> 00:04:10,760
So who is developing AI tech?

67
00:04:10,760 --> 00:04:13,800
China, Russia, EU?

68
00:04:13,800 --> 00:04:14,800
What?

69
00:04:14,800 --> 00:04:20,520
We're just supposed to imagine that American companies are going to cede their position

70
00:04:20,520 --> 00:04:24,480
to overseas organizations.

71
00:04:24,480 --> 00:04:25,480
It's a nonsense.

72
00:04:25,480 --> 00:04:30,360
I understand why a group of hyper-official figures are going to say, hey, look, this

73
00:04:30,360 --> 00:04:36,520
is something that we need to have a real discussion about because this tech is, there's potential

74
00:04:36,520 --> 00:04:38,240
for it to run away from us.

75
00:04:38,240 --> 00:04:43,080
But the idea of a, hey, let's just, let's let everyone slow down, start, take a breather,

76
00:04:43,080 --> 00:04:47,360
six months, rest, pause, and then we'll all just get back on the, on the horse in six

77
00:04:47,360 --> 00:04:48,360
months time.

78
00:04:48,360 --> 00:04:50,020
You know, six months seems arbitrary.

79
00:04:50,020 --> 00:04:54,640
It just seems like a bit of a, a way to make a bit of noise.

80
00:04:54,640 --> 00:04:58,760
And for some organizations, like you say, to try and play catch up.

81
00:04:58,760 --> 00:05:04,400
I guess there's an element of maybe knowing that this wouldn't work, but also giving yourself

82
00:05:04,400 --> 00:05:08,920
a told you so option further down the line if you need it.

83
00:05:08,920 --> 00:05:13,480
I'm like you, Martin, if I take my cynical hat back off and you read the letter, I think

84
00:05:13,480 --> 00:05:14,960
a lot of the concerns are real.

85
00:05:14,960 --> 00:05:19,400
And I think if you look at the speed of change just in March, even people who were broadly

86
00:05:19,400 --> 00:05:24,340
excited by the technology may be feeling a bit nervous about how fast things are moving.

87
00:05:24,340 --> 00:05:28,280
And I have to admit, I'm starting to think like that myself, but I don't think it'll

88
00:05:28,280 --> 00:05:32,200
have an impact on the open AI and Microsoft juggernaut at all.

89
00:05:32,200 --> 00:05:37,640
I think it's tantamount to asking every oil drilling company to stop drilling at some,

90
00:05:37,640 --> 00:05:42,720
at the same time to positively impact climate, climate change for the good of humanity.

91
00:05:42,720 --> 00:05:45,400
That's a great idea.

92
00:05:45,400 --> 00:05:51,040
But I just don't think that every single oil company or in this case, AI company is going

93
00:05:51,040 --> 00:05:55,320
to, is going to shut everything down because it only takes one to plow on regardless.

94
00:05:55,320 --> 00:05:57,760
And then they become the winner that takes all, right?

95
00:05:57,760 --> 00:05:59,160
So I just, I just don't see it.

96
00:05:59,160 --> 00:06:00,160
Yeah.

97
00:06:00,160 --> 00:06:01,160
The stakes are too high.

98
00:06:01,160 --> 00:06:02,160
Absolutely.

99
00:06:02,160 --> 00:06:04,080
There's no, and who's going to enforce this?

100
00:06:04,080 --> 00:06:08,520
Are the US government are going to pass this law change through?

101
00:06:08,520 --> 00:06:09,520
No.

102
00:06:09,520 --> 00:06:14,680
So it's impractical, but I can see why they, why they would push for it.

103
00:06:14,680 --> 00:06:15,680
Yeah, definitely.

104
00:06:15,680 --> 00:06:16,680
It's interesting.

105
00:06:16,680 --> 00:06:19,200
It's prompted some conversations, this type of thing.

106
00:06:19,200 --> 00:06:22,520
I've been having some, some people who've been listening to the podcast and sending

107
00:06:22,520 --> 00:06:25,440
me your messages and asking me my different viewpoints on different things.

108
00:06:25,440 --> 00:06:30,440
I'll be interested to get your viewpoint, Martin, but one of the questions I get is

109
00:06:30,440 --> 00:06:35,640
how scared of AIRU, Paul, because, you know, obviously you're doing a lot of reading and

110
00:06:35,640 --> 00:06:36,640
thinking about this.

111
00:06:36,640 --> 00:06:41,800
And I think ultimately, I think it's got great power and to help a lot of people be more

112
00:06:41,800 --> 00:06:42,800
efficient and effective.

113
00:06:42,800 --> 00:06:46,440
And it's going to have a massive impact for the good of humanity potentially in a lot

114
00:06:46,440 --> 00:06:50,440
of domains, especially things like healthcare and primary research.

115
00:06:50,440 --> 00:06:54,560
And on the business side, it's going to allow us to do more in less time, which I think

116
00:06:54,560 --> 00:06:57,480
is, is already starting to be shown.

117
00:06:57,480 --> 00:07:01,240
I think a key thing is going to be how we choose to use that power and how we choose

118
00:07:01,240 --> 00:07:03,720
to use it remains to be seen a bit.

119
00:07:03,720 --> 00:07:08,000
Paul Roetser at the Marketing AI Institute talks about using this to free people up to

120
00:07:08,000 --> 00:07:12,800
work fewer days and still have the same output and deliver the same outcomes.

121
00:07:12,800 --> 00:07:16,520
And I would honestly love to see that become a reality.

122
00:07:16,520 --> 00:07:19,880
You know, we'll all do three day weeks and AI can pick up the slack.

123
00:07:19,880 --> 00:07:27,200
But my fear is in this capitalist mechanism that underpins Western society, I don't think

124
00:07:27,200 --> 00:07:32,520
we can enable it without government intervention as we live in a sort of a market driven economic

125
00:07:32,520 --> 00:07:38,600
system that emphasizes the pursuit of profit where businesses compete against each other.

126
00:07:38,600 --> 00:07:43,600
So if one business decides to work three days a week in leverage AI to pick up the slack,

127
00:07:43,600 --> 00:07:47,640
it only takes one of the competitors to work five days a week with the same power and they'll

128
00:07:47,640 --> 00:07:49,880
outcompete the other companies.

129
00:07:49,880 --> 00:07:54,280
And so yeah, I'm not quite sure that we can make that happen.

130
00:07:54,280 --> 00:07:55,920
I don't know what your thoughts are on that.

131
00:07:55,920 --> 00:08:01,680
You know, I think there's a societal shift that would be required in order to kind of

132
00:08:01,680 --> 00:08:02,680
see that.

133
00:08:02,680 --> 00:08:07,560
It's in fact, have you listened to the Sam Altman interview?

134
00:08:07,560 --> 00:08:14,640
He's done a two and a half hour interview this week and he talks about, you know, there

135
00:08:14,640 --> 00:08:20,400
is going to be major economic shifts that come about through the deployment of, they're

136
00:08:20,400 --> 00:08:26,600
talking specifically about AGI in that interview and we're not quite there yet.

137
00:08:26,600 --> 00:08:33,200
But you know, things like universal basic income, if you're going to have people working

138
00:08:33,200 --> 00:08:41,800
slightly less, presumably some people at the kind of bottom might see the amount of work

139
00:08:41,800 --> 00:08:44,680
that they have generally reduced.

140
00:08:44,680 --> 00:08:52,760
So their income is lower if AGI and corporations are able to make higher profits with lower

141
00:08:52,760 --> 00:08:57,480
inputs or lower labour costs.

142
00:08:57,480 --> 00:09:01,040
Some of those profits should be more heavily taxed, in which case we should expect to see

143
00:09:01,040 --> 00:09:05,560
from the go to universal basic income scheme so that people do have the ability to earn

144
00:09:05,560 --> 00:09:08,800
a good living doing three days a week.

145
00:09:08,800 --> 00:09:15,200
And the thing Mark said, we should be able to spend our days chasing butterflies.

146
00:09:15,200 --> 00:09:16,200
Yeah, absolutely.

147
00:09:16,200 --> 00:09:21,760
I think that's a great perspective, Martin, and I love the deep questions that Paul Raitzer

148
00:09:21,760 --> 00:09:26,560
asks about this topic and to really get us all thinking because we're going to get, well,

149
00:09:26,560 --> 00:09:30,520
this open letter is because there is now a swelling belief that we're going to get to

150
00:09:30,520 --> 00:09:34,040
all of these things much faster than perhaps we all thought.

151
00:09:34,040 --> 00:09:35,720
And we have to be asking these questions.

152
00:09:35,720 --> 00:09:37,600
We have to be debating how.

153
00:09:37,600 --> 00:09:41,400
And whilst that might sound like a bit of a negative nitty-nitty, I don't mean to.

154
00:09:41,400 --> 00:09:45,400
I mean, we have to look at the system that we're operating because it's probably not

155
00:09:45,400 --> 00:09:51,240
fit for purpose for what systems we're going to have to have in place to make sure that

156
00:09:51,240 --> 00:09:56,760
AI is leveraged for the benefit of people, just not as another accelerator.

157
00:09:56,760 --> 00:10:02,160
We'll take you back to the conversation that we had last week looking at the report that

158
00:10:02,160 --> 00:10:05,280
OpenAI did on the impact of the labour market.

159
00:10:05,280 --> 00:10:11,520
80% of jobs, they estimate, will be impacted 10% by GPTs.

160
00:10:11,520 --> 00:10:16,120
19% of jobs will be impacted 50% by GPTs.

161
00:10:16,120 --> 00:10:20,480
These are substantial figures, 20% of jobs being impacted 50%.

162
00:10:20,480 --> 00:10:24,880
And if you take, you know, kind of extrapolated how, I think there are some more, rather than

163
00:10:24,880 --> 00:10:29,280
just the impact on the jobs, but the nature of the jobs that are going to be impacted.

164
00:10:29,280 --> 00:10:30,760
We've spoken about it before.

165
00:10:30,760 --> 00:10:38,240
If you're a GP, or if you're a HR consultant or a solicitor, your training has always been

166
00:10:38,240 --> 00:10:45,920
on knowledge and having knowledge on a certain corpus of text or, you know, kind of, and

167
00:10:45,920 --> 00:10:46,920
specialism.

168
00:10:46,920 --> 00:10:53,880
If you've now got GPTs that can pass the equivalent exam that you passed, or pass it at a higher

169
00:10:53,880 --> 00:10:59,920
score than the average human that passes those things, that is that, there's not as much,

170
00:10:59,920 --> 00:11:01,760
these high status jobs are going to be diminished.

171
00:11:01,760 --> 00:11:06,280
So there's a whole shift that is about to be undertaken.

172
00:11:06,280 --> 00:11:11,440
And I think there will be a resistance at first, as people say, I'm not going to trust

173
00:11:11,440 --> 00:11:13,960
my, let's say, let's take diagnosis, right?

174
00:11:13,960 --> 00:11:20,200
The thing that you go for a, to a GP for, I'm not going to trust GPT with that.

175
00:11:20,200 --> 00:11:24,080
And people will have a natural reluctance until the fact that some people, a sliver

176
00:11:24,080 --> 00:11:27,600
of people are doing that and they see good results from it and they see it working very

177
00:11:27,600 --> 00:11:31,680
well for them and then over time, more and more people, it's like self-driving cars.

178
00:11:31,680 --> 00:11:35,840
You know, we haven't gone from level zero to level five autonomous driving cars.

179
00:11:35,840 --> 00:11:41,960
We've got gradual improvements in assistance from level one, level two, level three, level

180
00:11:41,960 --> 00:11:43,200
four, level five.

181
00:11:43,200 --> 00:11:44,200
That's where we're going.

182
00:11:44,200 --> 00:11:47,800
We're just starting to see level three autonomous cars coming out into the market.

183
00:11:47,800 --> 00:11:50,540
Next it'll be level four and then it'll be level five.

184
00:11:50,540 --> 00:11:52,040
These things take time.

185
00:11:52,040 --> 00:11:58,240
I think as we start to accept GPTs assisting us, doing things that we previously said could

186
00:11:58,240 --> 00:12:03,520
only be done by a human doctor, a human lawyer, a human this that the other.

187
00:12:03,520 --> 00:12:04,520
Yeah.

188
00:12:04,520 --> 00:12:10,600
Over time that shift is going to incrementally come in and the impact is yet to be seen.

189
00:12:10,600 --> 00:12:14,160
I don't think people have, well, we know people haven't woken up to this yet.

190
00:12:14,160 --> 00:12:15,160
No.

191
00:12:15,160 --> 00:12:18,120
So let's hope, let's hope there's more going on behind the scenes than we know.

192
00:12:18,120 --> 00:12:23,520
And then we're going to talk about a UK government paper later to really start to think about

193
00:12:23,520 --> 00:12:30,200
what societal measures need to be in place in order to try and ensure that the route

194
00:12:30,200 --> 00:12:36,040
that this goes down is the route that, you know, hopefully would benefit the majority

195
00:12:36,040 --> 00:12:43,240
of, if not all of humanity rather than say a select few or actually not really unlocking

196
00:12:43,240 --> 00:12:46,560
a better life for a lot of people like it potentially could have.

197
00:12:46,560 --> 00:12:51,400
I think the other thing I often get asked about is like Skynet, Terminator, like are

198
00:12:51,400 --> 00:12:55,160
the AI's just going to like become self-aware and then kill us all?

199
00:12:55,160 --> 00:13:00,260
And I guess that's possible, but it's, I don't actually, that's not the case I tend to worry

200
00:13:00,260 --> 00:13:02,320
about the most if I do worry.

201
00:13:02,320 --> 00:13:05,320
And I try not to worry Mike because you can't worry about the things you can't control to

202
00:13:05,320 --> 00:13:11,000
some extent, but when I do worry, it's actually the sort of, not those doomsday scenarios,

203
00:13:11,000 --> 00:13:16,920
it's like accidental doomsday where we like accidentally mis-deploy AI on something or

204
00:13:16,920 --> 00:13:20,520
other and the story I've heard, must have read it on a book, I'm sure I've heard it

205
00:13:20,520 --> 00:13:23,800
in a billion podcasts as well, but the one that always sticks with me is this concept

206
00:13:23,800 --> 00:13:31,520
of creating like a paperclip master AI that's brief to develop the most effective paperclip

207
00:13:31,520 --> 00:13:36,040
that it can, manufacture it as efficiently as possible at scale.

208
00:13:36,040 --> 00:13:40,040
And then it's like told, right, go use whatever resources you need to do that.

209
00:13:40,040 --> 00:13:45,320
And of course, untouched, what it does is it uses all the world's resources to create

210
00:13:45,320 --> 00:13:50,280
the best paperclip that it then manufactures a ridiculous efficiency to turn the entire

211
00:13:50,280 --> 00:13:54,840
world into a world of paperclips, leaving no resources for humanity to actually live

212
00:13:54,840 --> 00:13:56,000
on anymore.

213
00:13:56,000 --> 00:14:00,320
And that we, and that the end of the world comes about because there's too many paperclips

214
00:14:00,320 --> 00:14:01,720
and there's not enough anything else.

215
00:14:01,720 --> 00:14:06,960
And the AI just doggedly kept doing it because that's what it's brief was.

216
00:14:06,960 --> 00:14:12,080
I hadn't really thought about that till I heard that story, but I think ironically,

217
00:14:12,080 --> 00:14:20,200
that's much more plausible doomsday scenario than Arnold Schwarzenegger style robots come

218
00:14:20,200 --> 00:14:21,200
to kill us all.

219
00:14:21,200 --> 00:14:26,560
So what we're saying is don't use an office world AI chatbot just in case.

220
00:14:26,560 --> 00:14:27,560
Yeah.

221
00:14:27,560 --> 00:14:29,400
Well, so, you know, who's going to deploy that?

222
00:14:29,400 --> 00:14:30,400
Clippy, isn't it?

223
00:14:30,400 --> 00:14:31,400
It's going to be...

224
00:14:31,400 --> 00:14:33,200
Yeah, he's going to be our overlord.

225
00:14:33,200 --> 00:14:37,200
Robin's hands are glee because this has been his plan for about 30 years and now it's all

226
00:14:37,200 --> 00:14:38,200
come like this.

227
00:14:38,200 --> 00:14:39,200
My time has come.

228
00:14:39,200 --> 00:14:42,680
I mean, ultimately what you're talking about there is an alignment issue, right?

229
00:14:42,680 --> 00:14:50,560
And alignment is one of the big topics in AI in terms of how we create models that align

230
00:14:50,560 --> 00:14:53,040
with humanity.

231
00:14:53,040 --> 00:14:56,200
And that's a big challenge because actually alignment, it means different things to different

232
00:14:56,200 --> 00:14:59,660
people and different cultures and different organizations and different interest groups.

233
00:14:59,660 --> 00:15:06,880
But I think actually those kinds of challenges of AI just getting super literal and running

234
00:15:06,880 --> 00:15:12,240
away with itself on one particular task, I'm less concerned about that.

235
00:15:12,240 --> 00:15:17,640
But I can see how it could all go horribly wrong and we'll turn into paperclips.

236
00:15:17,640 --> 00:15:22,560
Yeah, I think they're all unlikely, but certainly it's that one that...

237
00:15:22,560 --> 00:15:26,160
Not necessarily paperclips, but some unintended consequence.

238
00:15:26,160 --> 00:15:27,160
Yeah.

239
00:15:27,160 --> 00:15:32,240
Like an order consequence that we couldn't predict that you're like, oh, crumbs.

240
00:15:32,240 --> 00:15:33,240
Yeah.

241
00:15:33,240 --> 00:15:34,240
Sorry, that was me.

242
00:15:34,240 --> 00:15:35,240
Yeah.

243
00:15:35,240 --> 00:15:39,600
The entirety of...

244
00:15:39,600 --> 00:15:41,720
One city's always been turned to paperclips.

245
00:15:41,720 --> 00:15:44,760
Time to probably try and stop that AI now.

246
00:15:44,760 --> 00:15:45,760
But right.

247
00:15:45,760 --> 00:15:47,520
So I think that was an important story to cover.

248
00:15:47,520 --> 00:15:51,660
I think if you're a marketer, it's just useful to know the context of how these things are

249
00:15:51,660 --> 00:15:55,960
playing out, but not too much in the applicability there.

250
00:15:55,960 --> 00:15:58,000
Probably in your day-to-day work.

251
00:15:58,000 --> 00:16:03,760
Something that maybe is a bit more applicable is Levi's use of AI-generated models to increase

252
00:16:03,760 --> 00:16:04,760
diversity.

253
00:16:04,760 --> 00:16:08,840
Tell us a bit more about what you saw here this week, Martin.

254
00:16:08,840 --> 00:16:17,240
Levi Strauss has partnered with digital fashion studio, La La Land.ai to create AI-generated

255
00:16:17,240 --> 00:16:21,840
avatars to increase diversity amongst its models.

256
00:16:21,840 --> 00:16:26,540
Now what they're saying is La La Land uses advanced AI to create hyper-realistic models

257
00:16:26,540 --> 00:16:32,960
of various body types, ages, sizes, and skin tones for fashion brands and retailers.

258
00:16:32,960 --> 00:16:38,800
And Levi's is saying that their aim is to offer more personal, inclusive shopping experience

259
00:16:38,800 --> 00:16:43,320
for customers by enabling them to see products on models that resemble them.

260
00:16:43,320 --> 00:16:50,400
There's a lot of fanfare about this and I think it's a fascinating application.

261
00:16:50,400 --> 00:16:53,000
Nice to see some innovation in this space.

262
00:16:53,000 --> 00:16:55,640
I'm always interested to see what the big brands are doing.

263
00:16:55,640 --> 00:17:02,560
We spoke about Coca-Cola doing their partnership with OpenAI recently and this is a very practical

264
00:17:02,560 --> 00:17:03,560
example.

265
00:17:03,560 --> 00:17:06,080
I've seen things like smart mirrors.

266
00:17:06,080 --> 00:17:12,880
I know Amazon tried to do the smart retail app thing where you get the clothes that you

267
00:17:12,880 --> 00:17:17,760
like the look of and you can see them on yourself using the app.

268
00:17:17,760 --> 00:17:24,480
So this is a great example where they're using AI-generated models to increase diversity.

269
00:17:24,480 --> 00:17:31,680
They are saying that the use of AI, well, if we think about what this means for marketers,

270
00:17:31,680 --> 00:17:39,100
it just means that we can have better representation if we're working with models.

271
00:17:39,100 --> 00:17:50,600
On the flip side, if I'm a model from a minority community and the number of opportunities

272
00:17:50,600 --> 00:18:00,400
that I have to find gigs is already quite slim, I might not be looking at this entirely

273
00:18:00,400 --> 00:18:06,120
positively when I'm looking at it going, oh yes, is AI going to take jobs?

274
00:18:06,120 --> 00:18:10,600
In this sense, yes, it very much is taking my job here and now.

275
00:18:10,600 --> 00:18:16,400
So I'm kind of torn and I recognize that as a straight white man, I'm probably not the

276
00:18:16,400 --> 00:18:23,800
best voice or champion of understanding diversity and speaking about it in the right way or

277
00:18:23,800 --> 00:18:24,800
what have you.

278
00:18:24,800 --> 00:18:29,280
But I do think there's a, it didn't sit right with me when I heard this story straight up

279
00:18:29,280 --> 00:18:30,280
the back.

280
00:18:30,280 --> 00:18:36,000
I thought I appreciate and applaud the technology and the use of it.

281
00:18:36,000 --> 00:18:39,880
I don't necessarily like that the whole statement when they released it was really leading with

282
00:18:39,880 --> 00:18:42,720
that diversity message.

283
00:18:42,720 --> 00:18:49,040
To Levi's credit in the press release, they do say, look, we recognize that this isn't

284
00:18:49,040 --> 00:18:56,320
the only thing that's required in order to be a more diverse and inclusive company.

285
00:18:56,320 --> 00:18:59,620
There are other things that we're doing, but I think I thought leading with it was a bit

286
00:18:59,620 --> 00:19:03,440
like, I can see the downside as well as the positives there.

287
00:19:03,440 --> 00:19:09,360
Yeah, I think my gut, I haven't read the press release, but my gut feel is the same as yours,

288
00:19:09,360 --> 00:19:15,880
which is being able to technologically, it's really interesting.

289
00:19:15,880 --> 00:19:21,920
And I think it's only a half a step away from being easily able to dress yourself up in

290
00:19:21,920 --> 00:19:26,880
a bunch of different clothes when buying clothes online without actually having to try them

291
00:19:26,880 --> 00:19:27,880
on.

292
00:19:27,880 --> 00:19:29,400
You could see what you looked like in them potentially.

293
00:19:29,400 --> 00:19:35,000
And I know there are other technologies that are developed and used for that, but you could

294
00:19:35,000 --> 00:19:37,560
see how these would be combined.

295
00:19:37,560 --> 00:19:42,920
My Spidey sense is also tingling because I think this is a very powerful way to show

296
00:19:42,920 --> 00:19:48,680
customized ads to people based on things that you know about them in order to make the ads

297
00:19:48,680 --> 00:19:49,960
resonate more.

298
00:19:49,960 --> 00:19:58,080
We know there's lots of social science studies that show that people tend to like people

299
00:19:58,080 --> 00:20:01,320
like them more and be drawn to people who are like them.

300
00:20:01,320 --> 00:20:04,960
And I think if you see a bunch of ads that are people like you, you're going to be, you

301
00:20:04,960 --> 00:20:08,800
know, perhaps more likely to be influenced and embrace them.

302
00:20:08,800 --> 00:20:15,680
So I can see a powerful advertising component, but the bit about what the impact this has

303
00:20:15,680 --> 00:20:20,960
on those minority groups getting gigs that are probably already quite hard to get.

304
00:20:20,960 --> 00:20:22,480
I completely agree.

305
00:20:22,480 --> 00:20:24,400
That part feels a bit tone deaf to me.

306
00:20:24,400 --> 00:20:27,400
This is like, yeah, business strategy wise, I can absolutely understand why you would

307
00:20:27,400 --> 00:20:28,400
do it.

308
00:20:28,400 --> 00:20:30,880
Society wise, it feels icky as hell.

309
00:20:30,880 --> 00:20:31,880
Honestly.

310
00:20:31,880 --> 00:20:34,080
Yeah, we should probably just leave it at that.

311
00:20:34,080 --> 00:20:35,080
I think so.

312
00:20:35,080 --> 00:20:36,080
Icky as hell.

313
00:20:36,080 --> 00:20:38,520
But what are you talking about?

314
00:20:38,520 --> 00:20:40,000
Need to listen to the full podcast.

315
00:20:40,000 --> 00:20:43,680
Let's see how many times you can say icky as hell.

316
00:20:43,680 --> 00:20:44,680
Right.

317
00:20:44,680 --> 00:20:46,320
Let's get into our next story then.

318
00:20:46,320 --> 00:20:49,120
Let's talk about Bing announcing ads for chat.

319
00:20:49,120 --> 00:20:50,400
So this is kind of interesting.

320
00:20:50,400 --> 00:20:52,880
We've been talking previously like, what's going to happen?

321
00:20:52,880 --> 00:20:57,320
You know, people in chat GPT, they're in Bing, they're using these chat driven tools.

322
00:20:57,320 --> 00:21:03,240
Search engines are kind of paid for and underpinned by the fact that they show relevant ads in

323
00:21:03,240 --> 00:21:04,480
your search results.

324
00:21:04,480 --> 00:21:10,680
Very hard one would have assumed to do that in a natural chat style interface, especially

325
00:21:10,680 --> 00:21:16,000
one that ones that at the moment are starting to experiment with showing the sources that

326
00:21:16,000 --> 00:21:18,080
they're using for getting information.

327
00:21:18,080 --> 00:21:21,720
But of course, that's coming as part of the next phase of the rollout of these tools.

328
00:21:21,720 --> 00:21:26,200
So Microsoft issued a blog post that said they've begun experimenting with ads for

329
00:21:26,200 --> 00:21:31,520
Bing chat, which is sort of chat GPT through through the Bing search interface.

330
00:21:31,520 --> 00:21:36,040
If you like a couple of quotes from that blog post are that they are experimenting with

331
00:21:36,040 --> 00:21:40,640
an expanded hover experience where hovering over a link from a publisher will display

332
00:21:40,640 --> 00:21:45,120
more links from that publisher, giving the user more ways to engage and drive in more

333
00:21:45,120 --> 00:21:47,420
traffic to the publishers website.

334
00:21:47,420 --> 00:21:53,040
So I think there's an element here of trying to overcome the criticism of using other people's

335
00:21:53,040 --> 00:21:58,760
content in chat, but no real attribution and no possible driving of traffic to the original

336
00:21:58,760 --> 00:21:59,760
website.

337
00:21:59,760 --> 00:22:03,720
So we'll see how well that actually works, because I'm not sure if people are going to

338
00:22:03,720 --> 00:22:06,760
click those links, but that'd be good to experiment with.

339
00:22:06,760 --> 00:22:12,040
Another quote was, we're also exploring placing ads in the chat experience to share the ad

340
00:22:12,040 --> 00:22:16,040
revenue with partners whose content contributed to the chat response.

341
00:22:16,040 --> 00:22:22,120
So there's a couple of things here in terms of somewhat trying to deliver credit to content

342
00:22:22,120 --> 00:22:27,320
that's in the chat window and therefore some sort of revenue sharing model that I didn't

343
00:22:27,320 --> 00:22:30,520
quite understand that how that's going to work yet.

344
00:22:30,520 --> 00:22:37,400
And then the other example I saw actually showed ads like sponsored content in effect

345
00:22:37,400 --> 00:22:43,880
marked as an ad a little bit like you see on Google search now and Bing search, highlighting

346
00:22:43,880 --> 00:22:49,420
results that were brought back by the chat tool that were effectively paid for.

347
00:22:49,420 --> 00:22:54,880
So this shouldn't come as a huge surprise that Bing would look to monetize or Microsoft

348
00:22:54,880 --> 00:22:56,920
would look to monetize the tool.

349
00:22:56,920 --> 00:22:59,520
It certainly didn't take them very long, right?

350
00:22:59,520 --> 00:23:02,040
Bing chat hasn't been out very long.

351
00:23:02,040 --> 00:23:06,880
I think as a marketer, what this means is it could offer you an innovative way to reach

352
00:23:06,880 --> 00:23:11,920
your audience, especially as more and more people move towards chat based tools.

353
00:23:11,920 --> 00:23:17,200
And I think if this works for Bing and Microsoft, we should probably expect to see other search

354
00:23:17,200 --> 00:23:22,680
providers adopting a similar model in the future.

355
00:23:22,680 --> 00:23:26,680
Possibly it could come to chat GPT free quite quickly, one would assume.

356
00:23:26,680 --> 00:23:31,240
Whether or not it would make it into the paid version chat GPT plus remains to be seen.

357
00:23:31,240 --> 00:23:36,520
There's a bit of me that would hope it wouldn't, if I'm honest, but we shall see.

358
00:23:36,520 --> 00:23:42,680
I think it also opens up opportunities to start experimenting with this because we've

359
00:23:42,680 --> 00:23:47,040
talked previously, Martin, on the podcast and off the podcast.

360
00:23:47,040 --> 00:23:52,760
Where's the commercial opportunity in this, both as an advertising platform, but also

361
00:23:52,760 --> 00:23:55,240
for advertisers?

362
00:23:55,240 --> 00:23:59,680
And marketers spend so much of their time now, digital marketers, looking for the next

363
00:23:59,680 --> 00:24:02,400
digital marketing arbitrage, don't they?

364
00:24:02,400 --> 00:24:08,800
They're looking for that opportunity where marketers can identify a channel or a tactic

365
00:24:08,800 --> 00:24:13,320
that's currently low saturation that they think is going to have an oversized impact

366
00:24:13,320 --> 00:24:18,320
on their marketing results and then use it as quick as possible before it becomes oversaturated

367
00:24:18,320 --> 00:24:23,400
and audiences either tune it out or it becomes too expensive to use at scale.

368
00:24:23,400 --> 00:24:28,680
This is something that's been plaguing a lot of the real time bidding algorithms that underpin

369
00:24:28,680 --> 00:24:30,800
things like Google ads and LinkedIn ads and stuff.

370
00:24:30,800 --> 00:24:34,320
They're really expensive at this point, and especially if you work in e-commerce, it's

371
00:24:34,320 --> 00:24:37,840
hard to make margin on running those ads.

372
00:24:37,840 --> 00:24:45,600
So I guess this could be, is chat the next arbitrage opportunity for marketers to get

373
00:24:45,600 --> 00:24:51,640
in early, start testing it and see if they can get good results for a really low price?

374
00:24:51,640 --> 00:24:56,440
So I think that's how I'm starting to see this as it starts to play out.

375
00:24:56,440 --> 00:25:00,520
What are your thoughts on this bit of news, Martin?

376
00:25:00,520 --> 00:25:07,400
The first example I saw of it was from someone doing a search about the price of a car.

377
00:25:07,400 --> 00:25:12,240
How much does a certain car cost at this moment in time?

378
00:25:12,240 --> 00:25:19,140
And it had the price and it had a link to an article where this price was.

379
00:25:19,140 --> 00:25:24,400
It was such a strange ad placement and I couldn't see myself, as far as I could tell from that

380
00:25:24,400 --> 00:25:30,360
example actually, when you clicked on it, it was actually almost like a car review website.

381
00:25:30,360 --> 00:25:33,960
I don't honestly know for sure, but it didn't seem like it was taking me to a dealership

382
00:25:33,960 --> 00:25:35,120
where I was going to buy it.

383
00:25:35,120 --> 00:25:38,920
Maybe I was wrong there.

384
00:25:38,920 --> 00:25:42,920
I think you kind of hit the nail on the head when you were saying, will people actually

385
00:25:42,920 --> 00:25:45,320
click on these?

386
00:25:45,320 --> 00:25:48,400
How much are people going to engage with it?

387
00:25:48,400 --> 00:25:52,160
The other thing is how much are people going to engage with it when plugins are coming

388
00:25:52,160 --> 00:25:53,160
out?

389
00:25:53,160 --> 00:25:58,880
How much are people going to be using Bing Chat compared to ChatGPT plugins where it's

390
00:25:58,880 --> 00:26:03,800
connected to Kayak and I can just do my search and it's going to tell me the itinerary that

391
00:26:03,800 --> 00:26:07,440
I can choose and give me the hotel that I've asked it to find and I'll just click on it

392
00:26:07,440 --> 00:26:08,440
that way.

393
00:26:08,440 --> 00:26:15,760
So yeah, I'm really interested to see how users actually interact with it.

394
00:26:15,760 --> 00:26:22,120
I suspect this is a first roll of the dice by Bing who has just gone, right, let's get

395
00:26:22,120 --> 00:26:24,440
something out and see what happens.

396
00:26:24,440 --> 00:26:26,760
Will this be the long-term solution?

397
00:26:26,760 --> 00:26:31,000
Will they look to actually use, I mean, it's almost certain that we're going to have some

398
00:26:31,000 --> 00:26:37,480
sort of remarketing based elements to this where chat topics will be used in the same

399
00:26:37,480 --> 00:26:43,120
way that search topics will be used and if you say, what's the best itinerary for a holiday

400
00:26:43,120 --> 00:26:47,800
in Crete, next thing you'll see is ads chasing you around the web for holidays in Crete.

401
00:26:47,800 --> 00:26:50,680
That kind of makes sense.

402
00:26:50,680 --> 00:26:52,480
I can see that being implemented.

403
00:26:52,480 --> 00:26:58,840
But yeah, these particular in-context ad units, the other thing is does that, as an end user,

404
00:26:58,840 --> 00:27:08,840
do I then trust the chat less and anything that harms the credibility or if I start to

405
00:27:08,840 --> 00:27:12,840
see that actually this response just feels entirely sponsored and it's not giving me

406
00:27:12,840 --> 00:27:19,440
the content that I'm after, are people going to shift to different LLM?

407
00:27:19,440 --> 00:27:24,000
So when there's open source models that are being run, they're just as big and powerful

408
00:27:24,000 --> 00:27:28,800
as GPT-4 or what have you, people just look elsewhere.

409
00:27:28,800 --> 00:27:31,520
So yeah, I mean, ultimately we're at day one, are we?

410
00:27:31,520 --> 00:27:39,280
And I think the market is, yes, if now there is an opportunity to capitalise on this as

411
00:27:39,280 --> 00:27:42,200
a new channel, then go for it.

412
00:27:42,200 --> 00:27:46,800
But don't be surprised if in six months time the whole thing has completely shifted and

413
00:27:46,800 --> 00:27:48,840
these ad units no longer exist.

414
00:27:48,840 --> 00:27:51,000
Yeah, absolutely.

415
00:27:51,000 --> 00:27:54,860
It's interesting when you say that and you give the kayak example and of course when

416
00:27:54,860 --> 00:27:59,200
ChatGPT launched its plugins as we covered in last week's episode, if you haven't listened

417
00:27:59,200 --> 00:28:07,480
to it yet, available by all good podcast subscription services, the first 20 that were selected,

418
00:28:07,480 --> 00:28:11,680
it was quite an interesting mix, but they were like Expedia and Kayak both in there,

419
00:28:11,680 --> 00:28:12,680
right?

420
00:28:12,680 --> 00:28:20,120
Both massively battling for where their almost entire model now is based on people who Google

421
00:28:20,120 --> 00:28:26,000
for holidays and things and then providing price comparisons and that type of stuff.

422
00:28:26,000 --> 00:28:30,520
Did Kayak and Expedia and those other platforms that were in the first 20 plugins to date

423
00:28:30,520 --> 00:28:33,560
pay for that opportunity, right?

424
00:28:33,560 --> 00:28:38,400
Because you could imagine that it probably would be worth paying for in order to be one

425
00:28:38,400 --> 00:28:40,200
of those first plugins.

426
00:28:40,200 --> 00:28:44,320
How does it work when all of Kayak's competitors get a plugin?

427
00:28:44,320 --> 00:28:49,320
Do you sort of drift towards your favourite plugin and you have to activate the plugin

428
00:28:49,320 --> 00:28:50,320
to make that search?

429
00:28:50,320 --> 00:28:52,600
That's how the plugins work at the moment.

430
00:28:52,600 --> 00:28:54,280
That's how it works right now.

431
00:28:54,280 --> 00:28:56,840
The plugins, so it's going to be a brand thing, isn't it?

432
00:28:56,840 --> 00:28:57,840
Absolutely.

433
00:28:57,840 --> 00:28:58,840
So then what do you end up with?

434
00:28:58,840 --> 00:29:04,440
Some sort of app marketplace where a little bit like you have on sort of Play Store and

435
00:29:04,440 --> 00:29:09,960
the Apple Store now, where you can sponsor specific search terms in the app store to

436
00:29:09,960 --> 00:29:12,080
make sure your app is up at the top.

437
00:29:12,080 --> 00:29:17,720
And is that how this all then starts to the monetisation of the aspect of this, which

438
00:29:17,720 --> 00:29:20,100
of course for most marketers is not going to be super relevant.

439
00:29:20,100 --> 00:29:26,320
But in terms of how the platforms evolve and how they make money and how businesses like

440
00:29:26,320 --> 00:29:31,280
Kayak and other businesses that want you searching their inventories, their product lists or

441
00:29:31,280 --> 00:29:34,560
whatever, you leverage the tools to drive commercial value.

442
00:29:34,560 --> 00:29:38,000
It's going to be fascinating because it is day one, like you said, Martin, and I think

443
00:29:38,000 --> 00:29:43,760
there's going to be a lot of twists and turns in this story as we see and as the businesses

444
00:29:43,760 --> 00:29:48,840
involved, Microsoft, Google and others have to snake and pivot around and test things

445
00:29:48,840 --> 00:29:53,240
to figure out what's going to work from a commercial standpoint because quite hard to

446
00:29:53,240 --> 00:29:57,200
imagine and bet for sure on what will, I would say.

447
00:29:57,200 --> 00:29:58,200
Yeah.

448
00:29:58,200 --> 00:30:02,160
And these systems are going to be expensive to run, expensive to develop.

449
00:30:02,160 --> 00:30:08,520
At some point they've got to start generating revenue and users are very quickly going to

450
00:30:08,520 --> 00:30:10,280
have diversity of options.

451
00:30:10,280 --> 00:30:16,880
The other day I was sat there with Bard open, with Bing chat open and with chat GPT open

452
00:30:16,880 --> 00:30:18,840
all on the same screen.

453
00:30:18,840 --> 00:30:21,880
Already we have three chat interfaces that start talking to it.

454
00:30:21,880 --> 00:30:25,520
I can already make these decisions.

455
00:30:25,520 --> 00:30:26,520
Competition is already rife.

456
00:30:26,520 --> 00:30:31,520
So yeah, you've got to make sure that it doesn't detract from the user experience because as

457
00:30:31,520 --> 00:30:39,880
we know UX is such a differentiator, hence chat GPT becoming the fastest product knowledge

458
00:30:39,880 --> 00:30:47,000
ever versus GPT 3 having been around for two years, barely having made a dent.

459
00:30:47,000 --> 00:30:48,000
Yeah.

460
00:30:48,000 --> 00:30:49,000
So funny, isn't it?

461
00:30:49,000 --> 00:30:50,000
Right.

462
00:30:50,000 --> 00:30:52,480
I think probably time to move on to the next story.

463
00:30:52,480 --> 00:30:54,000
This is with you, Martin.

464
00:30:54,000 --> 00:31:01,040
Tell us about this UK government news around this artificial intelligence white paper.

465
00:31:01,040 --> 00:31:07,600
So regulation in AI is a hot topic at the moment and there's a lot of it in the ether,

466
00:31:07,600 --> 00:31:15,400
in the US, China and EU, all of which looking at this area and the UK now that we live in

467
00:31:15,400 --> 00:31:21,280
post-Brexit Britain, obviously isn't going to be part of the EU regulatory framework.

468
00:31:21,280 --> 00:31:28,040
So the UK government has unveiled what they're calling a world leading approach to innovation

469
00:31:28,040 --> 00:31:33,000
in artificial intelligence to turbocharge growth.

470
00:31:33,000 --> 00:31:37,180
This white paper talks about a few interesting things.

471
00:31:37,180 --> 00:31:42,240
One that the main thing to come out of it is actually the UK's approach is going to

472
00:31:42,240 --> 00:31:44,040
be quite light touch.

473
00:31:44,040 --> 00:31:51,440
So what they're saying is that AI contributed already 3.7 billion to the UK economy and

474
00:31:51,440 --> 00:31:58,800
going forward UK government's approach to regulating AI will be based on five guiding

475
00:31:58,800 --> 00:32:04,240
principles of safety, transparency, fairness, accountability and contestability and there's

476
00:32:04,240 --> 00:32:07,360
definitions of each of those.

477
00:32:07,360 --> 00:32:13,440
What they're saying is that they're not going to put in a dedicated regulator overseeing

478
00:32:13,440 --> 00:32:14,440
AI.

479
00:32:14,440 --> 00:32:27,840
It's going to be existing regulators in place to have AI as part of their regulatory remit.

480
00:32:27,840 --> 00:32:37,120
So existing regulators are going to have to adapt and they are saying that they are empowering

481
00:32:37,120 --> 00:32:41,960
existing regulators to take their own tailored approaches based on the different sectors.

482
00:32:41,960 --> 00:32:48,640
So banking can take their own approach, help and what have you can all take their own approach.

483
00:32:48,640 --> 00:32:57,240
In order to enable businesses to kind of figure out what will or won't work in any regulatory

484
00:32:57,240 --> 00:33:05,800
environment they are launching a £2 million fund to basically enable sandbox trials.

485
00:33:05,800 --> 00:33:10,000
So what they'll say is businesses can basically experiment and work alongside the regulators

486
00:33:10,000 --> 00:33:15,960
and put ideas together and they can figure out whether their uses of AI fit within whatever

487
00:33:15,960 --> 00:33:23,080
the regulators in that industry seem to be appropriate.

488
00:33:23,080 --> 00:33:25,360
So what of it?

489
00:33:25,360 --> 00:33:32,480
Well it's been criticised in some corners as being too light touch and placing it at

490
00:33:32,480 --> 00:33:37,440
odds with global trends in AI regulation.

491
00:33:37,440 --> 00:33:41,360
Some people saying that consumer groups and privacy activists saying that it's not going

492
00:33:41,360 --> 00:33:44,360
to go far enough.

493
00:33:44,360 --> 00:33:55,240
There is no statutory body or statutory mechanism to enforce anything.

494
00:33:55,240 --> 00:34:00,400
It's all very, it's so light touch it's kind of a real hands off approach from a legislation

495
00:34:00,400 --> 00:34:01,400
perspective.

496
00:34:01,400 --> 00:34:06,600
It's all devolved, all of the power is going to be devolved to the regulators.

497
00:34:06,600 --> 00:34:11,720
And when you think about the regulators as is, how well funded are they?

498
00:34:11,720 --> 00:34:18,160
How much extra resource are they going to have to monitor or manage and regulate all

499
00:34:18,160 --> 00:34:22,800
of these new technologies which are as we've just discussed moving at kind of break net

500
00:34:22,800 --> 00:34:23,800
speed.

501
00:34:23,800 --> 00:34:27,240
They're going to be able to keep up with these changes.

502
00:34:27,240 --> 00:34:33,480
So I said that this is at odds with the rest of the world.

503
00:34:33,480 --> 00:34:42,320
In the USA they've got the Algorithmic Accountability Act 2022 which is a voluntary AI framework

504
00:34:42,320 --> 00:34:48,320
but companies are required to assess the impact of AI in however they're using it.

505
00:34:48,320 --> 00:34:53,920
China is introducing specific laws addressing AI risks with mandatory user notifications

506
00:34:53,920 --> 00:34:58,120
when AI algorithms are involved when they deal with a business.

507
00:34:58,120 --> 00:35:03,840
So if a business uses AI as part of a kind of profiling, maybe it's kind of risk management

508
00:35:03,840 --> 00:35:12,020
or insurance thing, the consumer has to be told that there is an AI mechanism to that.

509
00:35:12,020 --> 00:35:14,580
So they've actually enforced that.

510
00:35:14,580 --> 00:35:17,840
And that's moved past the kind of discussion stage or the proposal stage.

511
00:35:17,840 --> 00:35:20,520
This is actually being rolled out as we speak.

512
00:35:20,520 --> 00:35:27,120
The EU has the Artificial Intelligence Act which is kind of still being worked on.

513
00:35:27,120 --> 00:35:31,000
No firm paper yet.

514
00:35:31,000 --> 00:35:36,840
It's not been fully introduced but that's going to include some interesting mechanisms

515
00:35:36,840 --> 00:35:43,640
such as grading AI products by potential harm.

516
00:35:43,640 --> 00:35:51,800
So for instance, an AI that analyzes emails and assesses whether they are spam or not

517
00:35:51,800 --> 00:35:59,040
is going to be lower harm than something that uses biometric profiling for anything.

518
00:35:59,040 --> 00:36:00,480
Or that makes paper clips.

519
00:36:00,480 --> 00:36:07,080
Or that turns all of the world's resources into paper clips.

520
00:36:07,080 --> 00:36:12,760
So the EU one is the one with the broadest scope and it actually goes so far as to say

521
00:36:12,760 --> 00:36:15,600
that there are some AI uses that are prohibited.

522
00:36:15,600 --> 00:36:23,440
So for instance, social grading by governments is prohibited under the proposals from the

523
00:36:23,440 --> 00:36:29,480
EU which I guess is kind of probably a response to people's concerns that they've heard about

524
00:36:29,480 --> 00:36:34,400
the way that China does the social credit score and different mechanisms.

525
00:36:34,400 --> 00:36:35,400
I don't know, maybe not.

526
00:36:35,400 --> 00:36:36,400
I've not read the pay hook.

527
00:36:36,400 --> 00:36:42,560
It's kind of in line with the EU's understandable stances in a number of areas.

528
00:36:42,560 --> 00:36:50,720
In GDPR in terms of protecting people's privacy and ensuring that they're not categorized

529
00:36:50,720 --> 00:36:54,720
and treated differently based on what factors.

530
00:36:54,720 --> 00:36:56,920
AI is going to start grading people.

531
00:36:56,920 --> 00:37:02,760
I think A, a big sense and B, I think it's much more in line with things the EU has done

532
00:37:02,760 --> 00:37:06,680
in the past.

533
00:37:06,680 --> 00:37:14,160
With regard to the UK law, I think there's a wider commercial reality for any companies

534
00:37:14,160 --> 00:37:18,080
that are looking to develop their own products or services using AI.

535
00:37:18,080 --> 00:37:20,880
And that is that if they're developing these products and services, they're probably going

536
00:37:20,880 --> 00:37:27,440
to want to develop them to sell them in markets such as the US, the EU or China.

537
00:37:27,440 --> 00:37:34,040
So much like when GDPR was introduced, this resulted in many American organizations basically

538
00:37:34,040 --> 00:37:40,320
aligning with GDPR standards, right, with EU standards, because they recognized that

539
00:37:40,320 --> 00:37:43,600
it was useful and beneficial to them.

540
00:37:43,600 --> 00:37:48,520
And actually GDPR ended up becoming the gold standard for data protection regulation around

541
00:37:48,520 --> 00:37:49,520
the world.

542
00:37:49,520 --> 00:37:54,000
It's kind of the blueprint that people take and adapt for their own markets.

543
00:37:54,000 --> 00:37:59,800
So I think we'll see the same ultimately down the line with AI regulation.

544
00:37:59,800 --> 00:38:04,640
So countries around the world, depending on their sphere of influence on a geopolitical

545
00:38:04,640 --> 00:38:11,760
scale will either look to use the AI algorithmic accountability act in America.

546
00:38:11,760 --> 00:38:17,920
If they're maybe in North America, maybe Canada, Mexico, they'll kind of lean on the American

547
00:38:17,920 --> 00:38:22,560
framework in the EU kind of sphere of influence.

548
00:38:22,560 --> 00:38:26,100
Countries will look at the EU framework and what have you.

549
00:38:26,100 --> 00:38:31,060
So on the face of it, the UK government's approach of being an outlier in this domain

550
00:38:31,060 --> 00:38:35,560
does have some upsides.

551
00:38:35,560 --> 00:38:41,780
They can say, you know, we're supporting innovation, we're not going to be heavy handed on regulation.

552
00:38:41,780 --> 00:38:42,780
There are potential benefits.

553
00:38:42,780 --> 00:38:48,520
I can see that some businesses might want to use the UK as a bit of a testbed.

554
00:38:48,520 --> 00:38:55,960
And then maybe if a product or some sort of innovation is quite successful, maybe they

555
00:38:55,960 --> 00:39:03,000
can then lean on EU regulators, US regulators to say, you know, this isn't harmful.

556
00:39:03,000 --> 00:39:10,120
Can you change the rules to enable us to introduce the similar mechanism in your market?

557
00:39:10,120 --> 00:39:14,040
We can become a bit of a testing ground.

558
00:39:14,040 --> 00:39:19,600
That said, there's also a bit of me which is slightly cynical where I go, you know,

559
00:39:19,600 --> 00:39:24,120
is this just an easy way for the government to say, we just don't have the resource to

560
00:39:24,120 --> 00:39:25,120
enforce this?

561
00:39:25,120 --> 00:39:33,800
We're basically outsourcing the regulation because we're not the big players in the world.

562
00:39:33,800 --> 00:39:38,520
We're not setting regulatory standards in the way that we have done in the past.

563
00:39:38,520 --> 00:39:43,720
That's down to the EU, the US, and we'll just kind of leave it to the market to figure out

564
00:39:43,720 --> 00:39:46,160
where to go.

565
00:39:46,160 --> 00:39:51,560
From a kind of commercial investment decision, are businesses that are working heavily in

566
00:39:51,560 --> 00:39:58,480
AI going to be looking at London as a place to invest for AI development?

567
00:39:58,480 --> 00:40:02,920
Are we going to be looking at the UK and go, that's the place to develop because they've

568
00:40:02,920 --> 00:40:06,400
got less regulation in this area?

569
00:40:06,400 --> 00:40:09,520
I don't think so because of the reasons I've outlined previously.

570
00:40:09,520 --> 00:40:11,240
I think they're going to be looking.

571
00:40:11,240 --> 00:40:15,480
If the UK is the place to develop, I think it's going to be based on different factors

572
00:40:15,480 --> 00:40:21,720
like our skills, it could be levels of investment, what have you.

573
00:40:21,720 --> 00:40:26,320
I don't think this light of regulation is going to be it because as I mentioned, if

574
00:40:26,320 --> 00:40:32,280
you're developing an AI product or service, you're going to be selling into the EU and

575
00:40:32,280 --> 00:40:35,880
the US, so you're probably aligning with their frameworks as is.

576
00:40:35,880 --> 00:40:36,880
Yeah.

577
00:40:36,880 --> 00:40:40,320
Anyway, regardless of where you actually, if you developed it in the UK, where it was

578
00:40:40,320 --> 00:40:43,240
a bit looser, that makes perfect sense to me, Martin.

579
00:40:43,240 --> 00:40:49,040
That sounds like it's the type of nuanced interplay that we're seeing happening at the

580
00:40:49,040 --> 00:40:54,560
regulation level, which as marketers, we want to know how strict the rules are going to

581
00:40:54,560 --> 00:40:58,880
be on the people who are making the tools that we want to use because of course, the

582
00:40:58,880 --> 00:41:02,440
more regulation there is, the more tightly controlled the feature sets will be and the

583
00:41:02,440 --> 00:41:06,160
harder it will probably be to bring products to market, one assumes.

584
00:41:06,160 --> 00:41:12,720
That will then have a downstream impact on what we do and what tools we can use.

585
00:41:12,720 --> 00:41:18,720
I guess also, potentially even what we can do as marketers in terms of how far does this

586
00:41:18,720 --> 00:41:24,760
regulation stretch into content that you produce using AI and tactics that you use.

587
00:41:24,760 --> 00:41:27,240
You mentioned retargeting earlier, right?

588
00:41:27,240 --> 00:41:30,600
There's a version of the world where the AI is going to be much better at retargeting

589
00:41:30,600 --> 00:41:37,160
than even the mechanisms that we have now and profiling people effectively and how acceptable

590
00:41:37,160 --> 00:41:38,160
is that going to be?

591
00:41:38,160 --> 00:41:39,160
Fascinating one.

592
00:41:39,160 --> 00:41:46,080
Yeah, I think auditability as well, being able to dig into the models.

593
00:41:46,080 --> 00:41:52,440
This is something that the EU regulation is going to be big on, which is being able to

594
00:41:52,440 --> 00:41:57,040
interrogate the weights and biases basically within the models.

595
00:41:57,040 --> 00:41:59,920
That's something that I think marketers are going to have to be aware of going forward.

596
00:41:59,920 --> 00:42:06,640
If you're using a product, maybe you're buying some AI services, you've got to be prepared

597
00:42:06,640 --> 00:42:09,480
to maybe audit those underlying models as well.

598
00:42:09,480 --> 00:42:16,440
That's going to be hard when OpenAI, who was supposed to be in the title, is now not sharing

599
00:42:16,440 --> 00:42:21,760
those details and so that you don't know how the model works.

600
00:42:21,760 --> 00:42:27,320
Maybe OpenAI will have to be more open because regulation will require all large language

601
00:42:27,320 --> 00:42:33,280
models to show us how the weights and biases and probability matrices and all those other

602
00:42:33,280 --> 00:42:34,280
things work.

603
00:42:34,280 --> 00:42:35,280
Fascinating.

604
00:42:35,280 --> 00:42:37,840
Well, Martin, that was a good one.

605
00:42:37,840 --> 00:42:38,840
Thanks for sharing that with us.

606
00:42:38,840 --> 00:42:39,840
You've got another one for us as well.

607
00:42:39,840 --> 00:42:45,000
You've got some tweets that you saw Brian Halligan, co-founder of HubSpot, share about

608
00:42:45,000 --> 00:42:49,320
AI and the future content marketing that I think you've nicely summarized for us so that

609
00:42:49,320 --> 00:42:51,560
we don't have to go through the whole thing ourselves.

610
00:42:51,560 --> 00:42:52,560
Yeah.

611
00:42:52,560 --> 00:42:55,360
What does AI mean for the future of content marketing?

612
00:42:55,360 --> 00:42:57,240
Got some really good replies.

613
00:42:57,240 --> 00:42:59,880
It was a fascinating thread to read through.

614
00:42:59,880 --> 00:43:05,240
I was trying to find some broad categories because there was quite a lot in there.

615
00:43:05,240 --> 00:43:13,600
And this came off the back of GPT-4 and just this kind of step-changing quality of output.

616
00:43:13,600 --> 00:43:22,080
And you could just see that people's concern over low quality content where SEO had once

617
00:43:22,080 --> 00:43:26,340
ruled what's its future going to be.

618
00:43:26,340 --> 00:43:34,080
So I would say that it kind of got it down to five key themes really.

619
00:43:34,080 --> 00:43:41,760
So one of the things that came up a couple of times was people talking about AIO, artificial

620
00:43:41,760 --> 00:43:42,760
intelligence.

621
00:43:42,760 --> 00:43:47,680
Not AIAIO, because that would be AIAIO.

622
00:43:47,680 --> 00:43:52,200
The artificial intelligence optimization becoming the new SEOs.

623
00:43:52,200 --> 00:43:55,360
And Nelson Joyce replied saying AIO is the new SEO.

624
00:43:55,360 --> 00:44:00,320
And this kind of goes back to what we were saying earlier about how do you get your content

625
00:44:00,320 --> 00:44:01,600
into the Bing chat?

626
00:44:01,600 --> 00:44:06,200
For the past couple of years, people have been talking about how do you write a blog

627
00:44:06,200 --> 00:44:09,640
so that you end up getting in the featured snippet on Google.

628
00:44:09,640 --> 00:44:12,560
That's been a big thing for content marketers.

629
00:44:12,560 --> 00:44:19,880
Well how do you make sure that your content is now referenced in Bing chat or that your

630
00:44:19,880 --> 00:44:22,720
kind of model or framework, whatever it is that you want to talk about, is the thing

631
00:44:22,720 --> 00:44:25,280
that the AI puts forward to the readers.

632
00:44:25,280 --> 00:44:27,520
So that's going to be an interesting one, once the AIO plays out.

633
00:44:27,520 --> 00:44:34,200
Because if you take it on the paradigm that we've operated in with large language models

634
00:44:34,200 --> 00:44:39,860
up until about two weeks ago, there was a hard cut off with the data.

635
00:44:39,860 --> 00:44:45,080
If you were waiting to optimize your data on the old model, it's been trained up to

636
00:44:45,080 --> 00:44:49,880
September 21, when's the next model going to be updated and released?

637
00:44:49,880 --> 00:44:52,400
And then imagine if you didn't get your results straight away.

638
00:44:52,400 --> 00:44:54,080
It's like the core updates in Google.

639
00:44:54,080 --> 00:44:57,760
At least they come every six months or something.

640
00:44:57,760 --> 00:45:05,560
Having to wait three years for a new LLM update would be a bit painful.

641
00:45:05,560 --> 00:45:11,040
So now that they're actually connected to the web and we've got plugins and we've got

642
00:45:11,040 --> 00:45:15,960
crawlers and all sorts of integrations, then yeah, I think AIO is going to be something

643
00:45:15,960 --> 00:45:19,660
that people need to pay attention to.

644
00:45:19,660 --> 00:45:21,920
Another theme, human centric content.

645
00:45:21,920 --> 00:45:24,160
We've spoken about this before.

646
00:45:24,160 --> 00:45:29,920
People are going to have to have more personality, real humans, faces talking like this, you

647
00:45:29,920 --> 00:45:36,040
know, conversations with words and emotions and laughter and mistakes, all of that kind

648
00:45:36,040 --> 00:45:39,800
of stuff that humans do that AIs don't do so well.

649
00:45:39,800 --> 00:45:41,720
That's going to become increasingly important.

650
00:45:41,720 --> 00:45:42,720
Yeah.

651
00:45:42,720 --> 00:45:47,800
Humanizing it and really, yeah, allowing us to empathize and connect with it in a different

652
00:45:47,800 --> 00:45:50,760
way to robotically generated text.

653
00:45:50,760 --> 00:45:56,360
Yeah, I think so that's going to be a trust content will be king was something that one

654
00:45:56,360 --> 00:45:59,440
of the tweets said.

655
00:45:59,440 --> 00:46:02,720
So they're just going to be, so, you know, what does it mean for the future of content

656
00:46:02,720 --> 00:46:03,720
marketing?

657
00:46:03,720 --> 00:46:04,960
Evolving tools and roles.

658
00:46:04,960 --> 00:46:10,160
So new tools and in-house jobs or agencies for AI driven content.

659
00:46:10,160 --> 00:46:19,320
Brett, someone called Brett said, QA as a service for AI driven content development.

660
00:46:19,320 --> 00:46:23,520
Which I thought was an interesting, interesting one.

661
00:46:23,520 --> 00:46:27,880
I'm not entirely sure I understood it fully, but you know, I totally get that.

662
00:46:27,880 --> 00:46:37,120
So one of the things I've read is AI as much as anything, chat GPT is a powerful tool for

663
00:46:37,120 --> 00:46:41,800
experts because you can ask it to produce a bit of content, but you also have the necessary

664
00:46:41,800 --> 00:46:48,400
expertise to identify when it's hallucinating and talking absolute nonsense.

665
00:46:48,400 --> 00:46:55,320
And so QA as a service would, to me is getting an expert in that domain to review the output

666
00:46:55,320 --> 00:47:01,840
of chat GPT and making sure it hasn't spoken a load of crap, basically, which I can, I

667
00:47:01,840 --> 00:47:07,400
can actually imagine, especially because that's part of the value that Bystrata brings, right?

668
00:47:07,400 --> 00:47:12,640
Like we, when we produce content for clients as a life science specialist with writers

669
00:47:12,640 --> 00:47:17,000
with PhDs and stuff like that, we can understand the content that we're creating, you know,

670
00:47:17,000 --> 00:47:21,440
have peer to peer conversations with client subject matter experts and produce really

671
00:47:21,440 --> 00:47:26,120
good content first time because we understand the topic.

672
00:47:26,120 --> 00:47:32,760
I can quite imagine that if I hired a new marketing grad tomorrow who didn't have a

673
00:47:32,760 --> 00:47:36,960
science background and then asked chat GPT to write me a blog about confocal microscopy,

674
00:47:36,960 --> 00:47:42,600
that person that quite understandably wouldn't be able to tell fact from fiction in chat

675
00:47:42,600 --> 00:47:43,600
GPT's output.

676
00:47:43,600 --> 00:47:47,320
But one of my writers would because they've been working in this field a number of years

677
00:47:47,320 --> 00:47:51,040
and they've got the scientific expertise to, to QA the content.

678
00:47:51,040 --> 00:47:53,520
So yeah, I'm not surprised to see that.

679
00:47:53,520 --> 00:47:57,960
I hadn't quite thought of it as a service, to be honest.

680
00:47:57,960 --> 00:48:02,800
Because I think it, I think great content is not just, is it technically accurate?

681
00:48:02,800 --> 00:48:07,960
I think there's a lot of other things that make content great in terms of narrative structure,

682
00:48:07,960 --> 00:48:15,200
covering the key points, piffiness, readability in marketing, being honestly persuasive and

683
00:48:15,200 --> 00:48:20,660
compelling and encouraging a person to take another action, all of which AI is getting

684
00:48:20,660 --> 00:48:22,960
better at, but I don't think is there yet.

685
00:48:22,960 --> 00:48:27,560
So that I think is where I guess the question is what do you consider QA?

686
00:48:27,560 --> 00:48:29,120
But yeah, that's an interesting one.

687
00:48:29,120 --> 00:48:34,080
And yeah, something I think I will be keeping an eye on as an idea.

688
00:48:34,080 --> 00:48:41,880
Yeah, and the last one I want to mention in terms of the themes was increased alignment

689
00:48:41,880 --> 00:48:43,600
and faster response to market events.

690
00:48:43,600 --> 00:48:50,840
So it's just going to enable people to respond quicker to changing market conditions or environments.

691
00:48:50,840 --> 00:48:59,080
Wait, so Hello Margo said way faster response to new problems or events.

692
00:48:59,080 --> 00:49:07,440
And I think again, it's just that we've spoken about this before AI content efficiency, you

693
00:49:07,440 --> 00:49:09,800
can get things done quicker.

694
00:49:09,800 --> 00:49:14,560
So yeah, I think it's just going to enable people to produce more quickly.

695
00:49:14,560 --> 00:49:15,880
It was an interesting discussion.

696
00:49:15,880 --> 00:49:22,320
I think the main thing for me though was in fact, oh, there, here we go.

697
00:49:22,320 --> 00:49:23,320
Kit Bodner.

698
00:49:23,320 --> 00:49:26,120
So Kit Bodner used to be the CMO at HubSpot.

699
00:49:26,120 --> 00:49:31,680
I believe he replied saying content marketing will be driven by the things AI isn't as good

700
00:49:31,680 --> 00:49:32,680
at.

701
00:49:32,680 --> 00:49:35,880
Point of view, personality, people connection.

702
00:49:35,880 --> 00:49:38,480
Education content will be fully commoditized.

703
00:49:38,480 --> 00:49:40,240
Creators will win.

704
00:49:40,240 --> 00:49:45,880
Brand will have to partner and hire creators to have differentiation in the market.

705
00:49:45,880 --> 00:49:48,240
That is a hell of a quote.

706
00:49:48,240 --> 00:49:51,400
I think I agree with every single part of that.

707
00:49:51,400 --> 00:49:53,520
I actually thought that was really interesting.

708
00:49:53,520 --> 00:49:54,520
Right.

709
00:49:54,520 --> 00:49:57,040
So basically all the stories for this week.

710
00:49:57,040 --> 00:50:00,960
And then we were going to look at a tool of the week this week, which is GoCharlie.

711
00:50:00,960 --> 00:50:03,640
You've been playing with GoCharlie, Martin, tell us about it.

712
00:50:03,640 --> 00:50:13,240
GoCharlie is an AI writing assistant, much in the same vein as tools like Jasper and

713
00:50:13,240 --> 00:50:17,640
Writer and well, take your pick.

714
00:50:17,640 --> 00:50:19,560
There's been loads of them.

715
00:50:19,560 --> 00:50:21,120
What makes GoCharlie different?

716
00:50:21,120 --> 00:50:28,200
No, because let's be honest, now that we have such easy interfaces to interact with the

717
00:50:28,200 --> 00:50:34,000
likes of ChatGPT and say, write me a blog introduction or write me some email subject

718
00:50:34,000 --> 00:50:36,320
lines.

719
00:50:36,320 --> 00:50:42,240
The value that we used to get from these writing assistants has largely been diminished because

720
00:50:42,240 --> 00:50:46,600
the instruct power is so capable in ChatGPT.

721
00:50:46,600 --> 00:50:50,260
However, GoCharlie is a bit different.

722
00:50:50,260 --> 00:50:56,840
They've been, the team at GoCharlie has done a great job at really extending the power

723
00:50:56,840 --> 00:51:00,640
of LLMs and integrating kind of multimodal capabilities.

724
00:51:00,640 --> 00:51:03,560
So I think the team there have done a brilliant job.

725
00:51:03,560 --> 00:51:05,680
Shout out to Brennan and the team.

726
00:51:05,680 --> 00:51:08,520
I think they've done, like I say, done some good work.

727
00:51:08,520 --> 00:51:11,600
For example, Content Repurposer.

728
00:51:11,600 --> 00:51:14,240
This is a fantastic little tool.

729
00:51:14,240 --> 00:51:22,400
So Content Repurposer allows you to add some content, closing the name, right?

730
00:51:22,400 --> 00:51:31,320
So you can add a link to a YouTube video or you can upload an MP4 or an MP3.

731
00:51:31,320 --> 00:51:33,360
You can upload audio as well.

732
00:51:33,360 --> 00:51:38,600
And then you can then choose the output that you want and you can get it to write a blog

733
00:51:38,600 --> 00:51:39,600
post.

734
00:51:39,600 --> 00:51:42,240
You can get it to summarize the content.

735
00:51:42,240 --> 00:51:46,880
You can get it to do some LinkedIn content or Twitter content.

736
00:51:46,880 --> 00:51:48,480
And it's really, really good.

737
00:51:48,480 --> 00:51:53,560
If you take, for instance, the audio from a podcast, I can't imagine where this example

738
00:51:53,560 --> 00:51:57,960
might have come from, but if you take the audio recording of a podcast and stick it

739
00:51:57,960 --> 00:52:02,600
into the Content Repurposer, it will write you a blog post summarizing that content.

740
00:52:02,600 --> 00:52:05,520
So they've done a really good job with that.

741
00:52:05,520 --> 00:52:14,960
You've got, if you create a blog post using one of these tools, it introduces suggested

742
00:52:14,960 --> 00:52:15,960
images as well.

743
00:52:15,960 --> 00:52:21,880
And it actually puts in prompts that you can put into an image generator like Dali or Stable

744
00:52:21,880 --> 00:52:24,160
Diffusion as well.

745
00:52:24,160 --> 00:52:28,600
It does have an image generation tool in it, so it doesn't bake it in.

746
00:52:28,600 --> 00:52:31,020
I think that will come down the line.

747
00:52:31,020 --> 00:52:35,640
But I really like the product, the interface, it's quite cheeky.

748
00:52:35,640 --> 00:52:41,680
They've got Charlie is this little illustrated cartoon dog.

749
00:52:41,680 --> 00:52:45,080
There's a little Easter egg mode where you can turn it on and it turns all the icons

750
00:52:45,080 --> 00:52:48,200
to rashes of bacon.

751
00:52:48,200 --> 00:52:49,200
Who doesn't love bacon?

752
00:52:49,200 --> 00:52:54,720
It's quite playful, but actually, who doesn't like bacon?

753
00:52:54,720 --> 00:52:55,720
Don't at me, vegetarians.

754
00:52:55,720 --> 00:53:01,360
I was just thinking a lot of people don't like bacon, but you like bacon and that's

755
00:53:01,360 --> 00:53:02,880
absolutely your choice.

756
00:53:02,880 --> 00:53:03,880
I know.

757
00:53:03,880 --> 00:53:04,880
Bacon, yeah.

758
00:53:04,880 --> 00:53:11,000
So the interface is easy to use.

759
00:53:11,000 --> 00:53:16,360
They've got loads of templates for things like product description, subject lines, blog

760
00:53:16,360 --> 00:53:17,920
introductions, that kind of stuff.

761
00:53:17,920 --> 00:53:20,440
They're all very good and very capable.

762
00:53:20,440 --> 00:53:26,680
What I think they're really pushing the envelope on is bringing in these other technologies,

763
00:53:26,680 --> 00:53:30,400
for instance, like the transcription and the content repurposing.

764
00:53:30,400 --> 00:53:33,680
Being able to take a YouTube video, particularly if it's something that's quite short, if you

765
00:53:33,680 --> 00:53:38,240
get something like 15 minutes long, stick it in and it turns it into a blog in one click

766
00:53:38,240 --> 00:53:40,960
and you go, okay, that's cool.

767
00:53:40,960 --> 00:53:42,720
So are there any limits with it?

768
00:53:42,720 --> 00:53:46,720
I know a lot of the writing tools, you have to pay more and more and more to get more

769
00:53:46,720 --> 00:53:51,120
words and more characters, is that the similar thing with GoCharlie?

770
00:53:51,120 --> 00:53:52,120
Yeah.

771
00:53:52,120 --> 00:54:00,640
So there are, here's on their plans, to be honest, I'm not sure of the, they have a teams

772
00:54:00,640 --> 00:54:07,160
plan, an unlimited plan and a kind of entry plan, I believe.

773
00:54:07,160 --> 00:54:09,920
They've just launched the teams plan.

774
00:54:09,920 --> 00:54:11,360
Oh, I'm just looking.

775
00:54:11,360 --> 00:54:14,000
One monthly charge, unlimited content for growth.

776
00:54:14,000 --> 00:54:16,560
So the charge is a little higher than some of the other tools.

777
00:54:16,560 --> 00:54:22,000
So if you, the freelance entry one's $39 a month, but there's no caps on it, which for

778
00:54:22,000 --> 00:54:26,520
a lot of the tools that are cheaper than that, there are character or word limit caps on

779
00:54:26,520 --> 00:54:27,520
that.

780
00:54:27,520 --> 00:54:28,720
So that, that could make it interesting.

781
00:54:28,720 --> 00:54:34,960
I love that some of the things you described there in terms of being able to feed it audio

782
00:54:34,960 --> 00:54:37,320
and have it create the blog post.

783
00:54:37,320 --> 00:54:42,040
I did that all this week in the most convoluted way possible probably by the sounds of things,

784
00:54:42,040 --> 00:54:45,800
because I use whisper API to transcribe the audio of our last podcast.

785
00:54:45,800 --> 00:54:53,080
And then I fed the transcription in sections into chat GPT four and had it create blog

786
00:54:53,080 --> 00:54:54,400
sections off the back of it.

787
00:54:54,400 --> 00:54:58,400
And I couldn't help but think while I was doing it, there must be a better tool and

788
00:54:58,400 --> 00:55:00,920
a better way out there to do that.

789
00:55:00,920 --> 00:55:02,440
How good are the blog summaries?

790
00:55:02,440 --> 00:55:04,720
Are they, are they decent and decent length?

791
00:55:04,720 --> 00:55:08,680
Because when I've used other tools like this in the past, they've turned an hour's transcription

792
00:55:08,680 --> 00:55:13,760
into about 300 words and completely missed a lot of the value and the nuance of the actual

793
00:55:13,760 --> 00:55:15,200
thing I was trying to feed it.

794
00:55:15,200 --> 00:55:16,200
Yeah.

795
00:55:16,200 --> 00:55:17,200
So there is, there is a limit.

796
00:55:17,200 --> 00:55:21,520
Um, and I think that comes down to, I don't know what tools they're using in terms of

797
00:55:21,520 --> 00:55:25,240
whisper API, um, or what have you.

798
00:55:25,240 --> 00:55:30,560
I put in a transcript from one of our, sorry, I put in the audio recording from our blog.

799
00:55:30,560 --> 00:55:31,560
Uh, sorry.

800
00:55:31,560 --> 00:55:34,880
I'll get me words out in a second.

801
00:55:34,880 --> 00:55:37,080
This isn't real human content with real mistakes.

802
00:55:37,080 --> 00:55:40,880
Real, real genuine human mistakes.

803
00:55:40,880 --> 00:55:47,240
One of our recent episodes, I uploaded the content and it basically missed the, a good

804
00:55:47,240 --> 00:55:52,920
chunk of the podcast, probably like 25% of the content was missing from the summary.

805
00:55:52,920 --> 00:55:57,680
Um, so yeah, it's what output length itself though, it was a decent blog length.

806
00:55:57,680 --> 00:56:01,320
It was, you know, if memory serves me right, that one was around a thousand words.

807
00:56:01,320 --> 00:56:05,000
So it wasn't short, but it didn't capture everything.

808
00:56:05,000 --> 00:56:11,800
I think there are some, some limitations there, but the tool that cracks that, the tool that

809
00:56:11,800 --> 00:56:18,400
cracks getting a full length transcript, breaking it down into the appropriate number of tokens,

810
00:56:18,400 --> 00:56:21,840
feeding that through and then piecing it all back together.

811
00:56:21,840 --> 00:56:26,840
That is going to be, well, as someone who has to edit, summarize and write show notes

812
00:56:26,840 --> 00:56:31,120
for a podcast, I think that's going to be fantastic.

813
00:56:31,120 --> 00:56:32,120
Well, me too.

814
00:56:32,120 --> 00:56:33,640
Uh, someone who does similar.

815
00:56:33,640 --> 00:56:37,880
Um, I mean, the blog post that I produced in the end was drifting towards 3000 words.

816
00:56:37,880 --> 00:56:41,780
So it was extensive because I really wanted to do justice to the stories that we'd covered

817
00:56:41,780 --> 00:56:46,520
because not everybody has the time or inclination to listen to a podcast.

818
00:56:46,520 --> 00:56:51,840
Um, but yes, I do think, I do think that that tool would be valuable.

819
00:56:51,840 --> 00:56:56,880
And if, as we suspect more brands lean into human content, which probably ends up being

820
00:56:56,880 --> 00:57:03,120
audio and video, because that's how you know, it's a human, at least at the moment, um,

821
00:57:03,120 --> 00:57:06,560
more and more, we're probably going to want to have access to these tools to do those

822
00:57:06,560 --> 00:57:12,240
summaries and those blog posts and those social posts that do the original recordings justice.

823
00:57:12,240 --> 00:57:16,120
Because at the moment, the only way I could find was that convoluted way I described.

824
00:57:16,120 --> 00:57:18,480
And it sounds like GoCharlie's getting there.

825
00:57:18,480 --> 00:57:23,680
It sounds honestly better than a number of the other tools I've tried, but still not

826
00:57:23,680 --> 00:57:27,720
quite maybe what we ideally really want.

827
00:57:27,720 --> 00:57:30,160
Yeah, it's not far off.

828
00:57:30,160 --> 00:57:31,160
It's certainly not far off.

829
00:57:31,160 --> 00:57:34,000
I think the way that the team are developing at the moment, you know, I see they've got

830
00:57:34,000 --> 00:57:40,400
a very active community that, um, that they're really taking on board, uh, user feedback.

831
00:57:40,400 --> 00:57:43,840
Another thing I didn't mention actually, they have image generation, so you can create,

832
00:57:43,840 --> 00:57:45,200
um, 4k images.

833
00:57:45,200 --> 00:57:50,840
So they're nice upscaled images, but they've also got in-painting, which I think is a nice

834
00:57:50,840 --> 00:57:56,800
feature because, um, you don't see that on, on so many of the image generation tools.

835
00:57:56,800 --> 00:57:57,800
Yeah.

836
00:57:57,800 --> 00:58:01,400
You can highlight those bits that you, you don't have to go create a new image.

837
00:58:01,400 --> 00:58:04,720
You can highlight a little bit of the image like, Oh, can you change this from orange

838
00:58:04,720 --> 00:58:05,720
to red?

839
00:58:05,720 --> 00:58:08,160
Or can you remove that football from the background or whatever?

840
00:58:08,160 --> 00:58:09,160
Interesting.

841
00:58:09,160 --> 00:58:11,560
I think I'm going to have to go and have a play with that one.

842
00:58:11,560 --> 00:58:12,560
It sounds intriguing, right?

843
00:58:12,560 --> 00:58:15,400
So thank you very much for taking us through that and telling us a bit more about your

844
00:58:15,400 --> 00:58:17,400
experiences with GoCharlie.

845
00:58:17,400 --> 00:58:18,400
Cool.

846
00:58:18,400 --> 00:58:22,720
I think that's us done for another episode, isn't it, Martin?

847
00:58:22,720 --> 00:58:24,200
I think that is.

848
00:58:24,200 --> 00:58:26,760
Well, as always, thank you for your time.

849
00:58:26,760 --> 00:58:28,680
Always lovely to hang out with you too, the listeners.

850
00:58:28,680 --> 00:58:31,000
I hope you've enjoyed that and that it's been beneficial.

851
00:58:31,000 --> 00:58:34,680
If it's been useful for you and you've not subscribed yet, please do subscribe.

852
00:58:34,680 --> 00:58:38,840
And if you know other marketers who might benefit from listening to our ramblings, um,

853
00:58:38,840 --> 00:58:43,080
and that they won't be too mad at you for you sending this, this podcast, then please

854
00:58:43,080 --> 00:58:46,800
share it with your friends and your colleagues and, uh, and see if they like it as well.

855
00:58:46,800 --> 00:58:49,240
So more thing before we go on.

856
00:58:49,240 --> 00:58:51,320
We have the Twitter account.

857
00:58:51,320 --> 00:58:53,960
Oh, no, I Twitter account.

858
00:58:53,960 --> 00:58:57,800
So please do follow us and send us your feedback.

859
00:58:57,800 --> 00:59:00,720
It's AI Marketing Pod.

860
00:59:00,720 --> 00:59:03,120
So follow us on Twitter, AI Marketing Pod.

861
00:59:03,120 --> 00:59:04,560
Say hello, tell us your thoughts.

862
00:59:04,560 --> 00:59:09,200
I'm so glad you said that, Martin, because not only would I love like Martin would love

863
00:59:09,200 --> 00:59:14,800
for you to send us your thoughts, what we're really looking for as well is Ninja Marketing

864
00:59:14,800 --> 00:59:15,800
Applications.

865
00:59:15,800 --> 00:59:21,520
If you've done something cool with AI that's sort of beyond the bog standard, I asked Chat

866
00:59:21,520 --> 00:59:24,520
GPT to give me a blog post and then it did.

867
00:59:24,520 --> 00:59:29,320
We'd love to hear from, from you because some of the, and maybe it's one that you did yourself

868
00:59:29,320 --> 00:59:30,960
and we'd love to like talk about what you've done.

869
00:59:30,960 --> 00:59:34,960
You could even come on the show and we'll happily interview you about what you've done.

870
00:59:34,960 --> 00:59:38,400
Or maybe it's just something cool that you saw that you think we should see so that we

871
00:59:38,400 --> 00:59:41,660
can bring it up here and share it with the, with the rest of the community.

872
00:59:41,660 --> 00:59:42,660
Please do.

873
00:59:42,660 --> 00:59:46,520
Because I think there's some really exciting stuff happening in the margins where people

874
00:59:46,520 --> 00:59:50,580
are just doing something creative with these tools, thinking outside the box and getting

875
00:59:50,580 --> 00:59:51,920
some really amazing results.

876
00:59:51,920 --> 00:59:59,160
We talked about the success of the Shopify, uh, cold, uh, cold outreach email program

877
00:59:59,160 --> 01:00:00,160
last week.

878
01:00:00,160 --> 01:00:03,680
And if you've got them, we'd love to hear you hit us up on the Twitter's and share them

879
01:00:03,680 --> 01:00:09,160
with us and ask you questions and share your best applications and all that good stuff.

880
01:00:09,160 --> 01:00:10,160
Nice one.

881
01:00:10,160 --> 01:00:11,160
Thank you, Martin.

882
01:00:11,160 --> 01:00:13,200
I will look forward to you speaking to you next time.

883
01:00:13,200 --> 01:00:14,200
See you next week.

884
01:00:14,200 --> 01:00:15,200
Bye.

885
01:00:15,200 --> 01:00:16,200
Cheers.

886
01:00:16,200 --> 01:00:36,120
Thank you for listening to artificially intelligent marketing to stay on top of the latest trends,

887
01:00:36,120 --> 01:00:40,000
tips and tools in the world of marketing AI.

888
01:00:40,000 --> 01:00:41,760
Be sure to subscribe.

889
01:00:41,760 --> 01:00:46,760
We look forward to seeing you again next week.