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 --> 00:00:15,760 latest trends, tips, and tools in the world of marketing AI, helping you get the best 3 00:00:15,760 --> 00:00:18,560 results from your marketing efforts. 4 00:00:18,560 --> 00:00:22,160 Now let's join our hosts, Paul Avery and Martin Broadhurst. 5 00:00:22,160 --> 00:00:28,120 Hello everyone, welcome to episode 16 of Artificially Intelligent Marketing. 6 00:00:28,120 --> 00:00:32,680 I am joined as always by my very good friend, Martin Broadhurst. 7 00:00:32,680 --> 00:00:34,080 How are you Martin, and what have you been up to? 8 00:00:34,080 --> 00:00:35,080 I'm very good. 9 00:00:35,080 --> 00:00:38,600 It's been a busy week in the world of AI for me. 10 00:00:38,600 --> 00:00:43,360 Been delivering some workshops, two workshops this week, one on getting started with Chat 11 00:00:43,360 --> 00:00:48,360 GPT, and the other one was very much an email marketing focused one, but we did a bit of 12 00:00:48,360 --> 00:00:52,840 a deep dive into how AI can help with email marketing in that. 13 00:00:52,840 --> 00:00:58,580 In one of the sessions though, we got completely derailed as people started having a conversation 14 00:00:58,580 --> 00:01:03,200 amongst themselves about universal basic income and how they might need to find new ways to 15 00:01:03,200 --> 00:01:07,920 make a living in the future as AI takes their jobs. 16 00:01:07,920 --> 00:01:09,600 Hopefully they weren't too worried about that. 17 00:01:09,600 --> 00:01:14,680 I'm sure you were showing them how as the human in the loop, they're very much driving 18 00:01:14,680 --> 00:01:15,680 the system. 19 00:01:15,680 --> 00:01:17,960 It's not driving them, at least at this point in time. 20 00:01:17,960 --> 00:01:23,800 Yeah, very much so, telling them how it can augment their role as opposed to completely 21 00:01:23,800 --> 00:01:24,800 replace. 22 00:01:24,800 --> 00:01:25,800 Yeah, it was an interesting discussion. 23 00:01:25,800 --> 00:01:26,800 How about yourself? 24 00:01:26,800 --> 00:01:27,800 What have you been up to this week? 25 00:01:27,800 --> 00:01:28,800 Oh, what have I been up to? 26 00:01:28,800 --> 00:01:32,320 It's been lots of, I've been around about so much this week. 27 00:01:32,320 --> 00:01:35,920 It's been lots of new client engagement conversations, which has been fantastic. 28 00:01:35,920 --> 00:01:39,880 I've met some really interesting companies this week. 29 00:01:39,880 --> 00:01:49,120 I met one that's doing some really great stuff with using AI as part of analyzing 3D protein 30 00:01:49,120 --> 00:01:52,720 structures, which was really interesting. 31 00:01:52,720 --> 00:01:53,720 That was brilliant. 32 00:01:53,720 --> 00:01:56,920 Yeah, just getting involved in a few client projects where I could help. 33 00:01:56,920 --> 00:01:58,800 So yeah, good week for me. 34 00:01:58,800 --> 00:02:01,640 Basically in the office, less out and about, which was nice. 35 00:02:01,640 --> 00:02:07,440 Yeah, bit of a breather after you were gallivanting up north of the border the other week. 36 00:02:07,440 --> 00:02:08,440 Indeed. 37 00:02:08,440 --> 00:02:11,960 So we're going to be out and about next week and the week after delivering some strategy 38 00:02:11,960 --> 00:02:13,920 sessions and meeting with some different clients. 39 00:02:13,920 --> 00:02:17,800 So yeah, I'm going to be back on the road. 40 00:02:17,800 --> 00:02:19,360 Maybe we'll have to do our recording on the road. 41 00:02:19,360 --> 00:02:21,720 That would be exciting. 42 00:02:21,720 --> 00:02:24,080 What have we got for our listeners this week? 43 00:02:24,080 --> 00:02:28,280 Well, as usual, we've got our main stories, our short snippets and our tool of the week. 44 00:02:28,280 --> 00:02:35,200 So the main stories we're going to cover are Meta talking AI for marketers at the Cannes 45 00:02:35,200 --> 00:02:41,920 Lions advertising event this week. 46 00:02:41,920 --> 00:02:46,600 We've got loads of generative image news this week, so we've summarized all of that into 47 00:02:46,600 --> 00:02:48,240 one big story. 48 00:02:48,240 --> 00:02:53,840 We'll talk about Meta's new all-in-one voice synthesis tool, the demo videos of which sound 49 00:02:53,840 --> 00:02:56,240 really cool, really cool, very interesting. 50 00:02:56,240 --> 00:02:59,880 And we're going to talk about Pickup, which is a very popular project management tool, 51 00:02:59,880 --> 00:03:01,960 especially in agency land. 52 00:03:01,960 --> 00:03:07,480 And it's made its AI powered assistant available more widely and actually brings a lot of tools. 53 00:03:07,480 --> 00:03:13,280 And I think maybe is a little indicator as to where we might see a number of other tools 54 00:03:13,280 --> 00:03:14,280 go in the future. 55 00:03:14,280 --> 00:03:15,580 So we're going to dig into that. 56 00:03:15,580 --> 00:03:19,160 The tool of the week this week is going to be Perplexity AI. 57 00:03:19,160 --> 00:03:21,880 Right, let's look at our short snippets. 58 00:03:21,880 --> 00:03:24,880 We'll rattle through these at Moirapido. 59 00:03:24,880 --> 00:03:28,680 We're 11 Labs, a voice technology research company that we've mentioned before on the 60 00:03:28,680 --> 00:03:36,680 podcast and a world leader in AI audio synthesis, etc. has raised another chunk of cash, $90 61 00:03:36,680 --> 00:03:41,680 million as part of a Series A to continue the research and development and improvement 62 00:03:41,680 --> 00:03:46,480 of its product and taking that product more widely to market. 63 00:03:46,480 --> 00:03:49,200 Vimeo has launched some new AI tools. 64 00:03:49,200 --> 00:03:54,000 So if you're a Vimeo user to host your videos, you'll see that there are three new features, 65 00:03:54,000 --> 00:03:59,800 including a script generator, a teleprompter and a text based video editor that identifies 66 00:03:59,800 --> 00:04:04,440 and allows you to remove filler words, pauses and awkward moments, etc. 67 00:04:04,440 --> 00:04:06,680 Perhaps a little descript like. 68 00:04:06,680 --> 00:04:11,120 So I think if you're on the standard plan or above, you get access to these by default. 69 00:04:11,120 --> 00:04:14,640 So if you're a Vimeo user, go and have a look and have a little play. 70 00:04:14,640 --> 00:04:20,800 On the video front, Loom, which is a screen recording tool and video hosting tool that 71 00:04:20,800 --> 00:04:28,040 a lot of people use, launched an AI generative tool to auto title videos and also include 72 00:04:28,040 --> 00:04:30,520 a short summary underneath the video. 73 00:04:30,520 --> 00:04:32,600 So it's done auto transcripts for a while. 74 00:04:32,600 --> 00:04:36,600 I'm a Loom user, which is why I was interested in this. 75 00:04:36,600 --> 00:04:43,760 I do a lot of how to videos for various use cases, and I have to admit the auto title 76 00:04:43,760 --> 00:04:48,120 and auto summarizer have saved me a lot of time, but it's also really handy if you've 77 00:04:48,120 --> 00:04:53,560 got lots of videos like I have to just quickly read the AI generated summary and then it 78 00:04:53,560 --> 00:04:56,360 reminds you, oh yeah, that's video, that video, that's the video. 79 00:04:56,360 --> 00:05:00,400 And I think in many cases may even save people having to watch the video. 80 00:05:00,400 --> 00:05:03,480 So if you're a Loom user, check that out. 81 00:05:03,480 --> 00:05:07,720 Have you tested that to see if it does that to your old videos then, the ones that were 82 00:05:07,720 --> 00:05:08,720 pre this? 83 00:05:08,720 --> 00:05:11,080 It won't do it to my old videos. 84 00:05:11,080 --> 00:05:15,360 It's one of the first things that's gone up on the forums is can we have this to apply 85 00:05:15,360 --> 00:05:18,120 to our old videos? 86 00:05:18,120 --> 00:05:21,240 I don't know how they would manage for that or whether they'd be worried about an absolute 87 00:05:21,240 --> 00:05:22,840 deluge of users, right? 88 00:05:22,840 --> 00:05:27,480 If we've got 500 videos and we'll just push them all in in one go, that could be a bit 89 00:05:27,480 --> 00:05:28,480 of a problem. 90 00:05:28,480 --> 00:05:32,760 The other thing is I fed it a massive video and it went, sorry, computer says no on that 91 00:05:32,760 --> 00:05:37,160 one and it was like over an hour long and it just said, no, I can't do it. 92 00:05:37,160 --> 00:05:43,360 So it's been perfect for my five and 10 minute screen share how tos, longer form videos, 93 00:05:43,360 --> 00:05:45,800 not so much. 94 00:05:45,800 --> 00:05:51,160 In other news this week, we had inflection deputing its own foundational AI model to 95 00:05:51,160 --> 00:05:54,720 rival Google and open AI's LLMs. 96 00:05:54,720 --> 00:06:01,600 It's an AI startup that has inflection one and it also has a conversational agent called 97 00:06:01,600 --> 00:06:02,600 PI. 98 00:06:02,600 --> 00:06:10,480 The model is roughly GPT 3.5 size and it has capabilities that are on par or superior to 99 00:06:10,480 --> 00:06:15,480 other models around GPT 3.5 according to the company. 100 00:06:15,480 --> 00:06:25,440 I've actually used this because you can access it through WhatsApp, like a little chat friend 101 00:06:25,440 --> 00:06:31,120 and you can just use it as a brainstorm partner for a wide variety of topics, whether that's 102 00:06:31,120 --> 00:06:38,120 like business questions, marketing questions, mindfulness questions, anything you like. 103 00:06:38,120 --> 00:06:40,680 It's kind of like your personal assistant in your pocket. 104 00:06:40,680 --> 00:06:45,720 I found it to be really good when I was doing some stuff where I wanted to like be consistent 105 00:06:45,720 --> 00:06:50,600 in my mindfulness mind and I asked it to remind me at seven o'clock every day, it's time to 106 00:06:50,600 --> 00:06:54,200 do 10 minutes of mindfulness and it just never reminds me. 107 00:06:54,200 --> 00:06:57,240 So I messaged it back and I go, I don't know if you know, but you forgot to remind me again. 108 00:06:57,240 --> 00:07:01,280 It's like, oh, I'm so sorry for all that I forgot to remind you, but I'm glad you pointed 109 00:07:01,280 --> 00:07:05,040 this out because I'll definitely remind you tomorrow and then it doesn't again. 110 00:07:05,040 --> 00:07:09,720 So I've kind of, there's some good stuff of it, but some areas for improvement. 111 00:07:09,720 --> 00:07:11,920 And useful feedback for them, I'm sure. 112 00:07:11,920 --> 00:07:14,720 Yeah, we'll get them on the podcast. 113 00:07:14,720 --> 00:07:15,720 You're all welcome on the podcast. 114 00:07:15,720 --> 00:07:16,720 Obviously we've got Zach. 115 00:07:16,720 --> 00:07:19,720 Zacherberg is on next week as we're going to discuss. 116 00:07:19,720 --> 00:07:20,720 My phone. 117 00:07:20,720 --> 00:07:25,320 Is this before or after his fight with Elon Musk? 118 00:07:25,320 --> 00:07:29,800 I'm going to go and do a live stream of said fight and then I'll interview him straight 119 00:07:29,800 --> 00:07:30,800 after. 120 00:07:30,800 --> 00:07:32,520 In the cage, straight away. 121 00:07:32,520 --> 00:07:36,960 Regardless of what state he's in, for those of you who don't know what we're talking about, 122 00:07:36,960 --> 00:07:41,920 there's been a bit of banter on the old Twittersphere between Elon Musk and Mark Zuckerberg about 123 00:07:41,920 --> 00:07:49,160 having a cage fight because Zuckerberg is doing quite a lot of martial arts training 124 00:07:49,160 --> 00:07:51,000 at the moment. 125 00:07:51,000 --> 00:07:55,480 And supposedly Dana White has been in the middle of them. 126 00:07:55,480 --> 00:08:00,440 UFC president Dana White has been in the middle of them and spoke to both of them and he claims 127 00:08:00,440 --> 00:08:02,440 that they're deadly serious about doing this. 128 00:08:02,440 --> 00:08:03,440 It's ridiculous. 129 00:08:03,440 --> 00:08:08,800 Fire of the century pay-per-view, $20 giggy cash in. 130 00:08:08,800 --> 00:08:14,480 I've genuinely seen stories saying $100 pay-per-view and all the money goes to charity. 131 00:08:14,480 --> 00:08:15,800 I think I would probably pay it. 132 00:08:15,800 --> 00:08:19,480 It would just be fascinating to see how that would play out. 133 00:08:19,480 --> 00:08:20,680 Last couple of short snippets then. 134 00:08:20,680 --> 00:08:27,960 We've got, according to Group IB, over 100,000 chat GPT accounts were breached between June 135 00:08:27,960 --> 00:08:31,440 22 and May 23 and shared on the dark web. 136 00:08:31,440 --> 00:08:39,240 We talked previously about companies banning employees from using chat GPT because of potential 137 00:08:39,240 --> 00:08:42,280 sharing of proprietary information. 138 00:08:42,280 --> 00:08:44,000 This might make a few people nervous. 139 00:08:44,000 --> 00:08:49,200 You might have shared some proprietary information and then have those accounts, their account 140 00:08:49,200 --> 00:08:51,200 was one of those that was hacked. 141 00:08:51,200 --> 00:09:00,040 So I think this is a perfect example of how OpenAI probably need to figure out how to 142 00:09:00,040 --> 00:09:05,880 make it a really robust software product, not something they sort of went to market 143 00:09:05,880 --> 00:09:07,440 very quickly with. 144 00:09:07,440 --> 00:09:12,320 They now need to provide sort of enterprise level security, but hey, that's probably a 145 00:09:12,320 --> 00:09:14,280 discussion for another day. 146 00:09:14,280 --> 00:09:20,800 And then the last story this week was according to a recent UK wide survey by email specialist 147 00:09:20,800 --> 00:09:28,680 Instiller, one in two people can't spot an AI generated email and less than one in three, 148 00:09:28,680 --> 00:09:33,640 and you basically don't care about AI and their brand communications. 149 00:09:33,640 --> 00:09:36,680 So what does that tell us as marketers? 150 00:09:36,680 --> 00:09:40,840 Perhaps we can experiment with a bit of AI to auto personalize our emails because a 151 00:09:40,840 --> 00:09:42,440 bunch of people are not going to spot it. 152 00:09:42,440 --> 00:09:45,840 And even those that do may not care that much as long as they're getting valuable information, 153 00:09:45,840 --> 00:09:47,600 which I think was interesting. 154 00:09:47,600 --> 00:09:53,400 We've talked before on the podcast Martin about lies, damn lies and statistics and the 155 00:09:53,400 --> 00:09:57,120 questions you ask and the groups that you ask and how you can basically get whatever 156 00:09:57,120 --> 00:09:59,600 answer you want, depending on how you frame all those things. 157 00:09:59,600 --> 00:10:03,640 But I thought it was an interesting enough one to chuck into our short snippets. 158 00:10:03,640 --> 00:10:04,640 Yeah. 159 00:10:04,640 --> 00:10:10,280 On that very point, I saw an interesting Twitter thread where someone had tweeted an infographic 160 00:10:10,280 --> 00:10:15,880 and it said the percentage of studios, and I think it was game development and movie 161 00:10:15,880 --> 00:10:20,520 studios, any kind of creative studios, percentage of studios using different AI tools. 162 00:10:20,520 --> 00:10:26,320 And they had a big list of them in the graph and it said 22% were using auto GPT. 163 00:10:26,320 --> 00:10:34,680 And someone else had quote tweeted this particular graph and said, I'm absolutely convinced there 164 00:10:34,680 --> 00:10:39,400 are zero users of auto GPT, zero, none. 165 00:10:39,400 --> 00:10:43,320 And then a great conversation thread below where it was just loads of people saying, 166 00:10:43,320 --> 00:10:46,320 yeah, I used auto GPT for an afternoon. 167 00:10:46,320 --> 00:10:49,480 It was novel, but didn't really do anything. 168 00:10:49,480 --> 00:10:51,080 It's not fit for purpose. 169 00:10:51,080 --> 00:10:54,480 Nobody is using this in a commercial serious setting. 170 00:10:54,480 --> 00:10:57,000 That is interesting. 171 00:10:57,000 --> 00:11:01,480 One of my friends who used it basically could have been the one who tweeted about the afternoon, 172 00:11:01,480 --> 00:11:02,480 had a real runner. 173 00:11:02,480 --> 00:11:04,040 It is very, very capable. 174 00:11:04,040 --> 00:11:05,360 It's a mill that we had on the podcast. 175 00:11:05,360 --> 00:11:10,400 I see for listeners, knows how to get the best out of these things and found it very, 176 00:11:10,400 --> 00:11:15,000 very difficult to get it to do anything meaningful in a way that you could rely on the output. 177 00:11:15,000 --> 00:11:22,360 So yeah, it's, we're in that part of the, of the development of these tools where they 178 00:11:22,360 --> 00:11:27,600 can almost do a lot of great stuff, but they're just too unreliable in a number of areas to 179 00:11:27,600 --> 00:11:29,880 make them your go-to tool. 180 00:11:29,880 --> 00:11:30,880 Yeah. 181 00:11:30,880 --> 00:11:36,400 And there are some real subtleties, even like a really simple example with prompting. 182 00:11:36,400 --> 00:11:41,200 So I was playing around this week with a chat GPT. 183 00:11:41,200 --> 00:11:47,240 It was actually on the workshop and I was showing people how using chat GPT and specifically 184 00:11:47,240 --> 00:11:55,000 chat GPT4, you could ask it to create documents from scratch using Google Apps Script to create 185 00:11:55,000 --> 00:12:00,000 you, for example, the particular example I was using was a customer feedback form. 186 00:12:00,000 --> 00:12:02,760 So if you're designing a customer survey, you've got to come up with the questions, 187 00:12:02,760 --> 00:12:06,480 you've got to come up with all of the, the rating systems and what have you. 188 00:12:06,480 --> 00:12:10,500 So you just ask it to write you a survey using best practice. 189 00:12:10,500 --> 00:12:12,680 It writes that if you're happy with that, great. 190 00:12:12,680 --> 00:12:16,480 And then say, create this using Google Apps Script. 191 00:12:16,480 --> 00:12:20,160 And then it will give you the script and then you drop that in Google Apps Script and it 192 00:12:20,160 --> 00:12:23,240 will create you the form in seconds. 193 00:12:23,240 --> 00:12:28,160 So like you have your customer feedback form done in no time at all. 194 00:12:28,160 --> 00:12:33,080 And I was showing this to the, to the room full of people and the first response that 195 00:12:33,080 --> 00:12:39,440 it gave me, I noticed a very subtle element of the output, which basically meant that 196 00:12:39,440 --> 00:12:40,960 it wasn't going to create a new form. 197 00:12:40,960 --> 00:12:46,440 It was going to try to edit an existing form and I had to give it my existing form ID. 198 00:12:46,440 --> 00:12:52,280 And I spotted that and corrected it and just changed my prompt to say, create a new form. 199 00:12:52,280 --> 00:12:54,800 So just a very slight subtle difference. 200 00:12:54,800 --> 00:12:58,360 But in that session, everyone was like, how did you spot that? 201 00:12:58,360 --> 00:12:59,360 How did you know that? 202 00:12:59,360 --> 00:13:03,080 And I thought that's interesting that it's only when you're using these tools day in 203 00:13:03,080 --> 00:13:06,600 day out that you can, you can see them and adjust them. 204 00:13:06,600 --> 00:13:12,600 And you know, going back to what we said previously about UI and UX being the kind of key differentiator, 205 00:13:12,600 --> 00:13:16,140 when you've got to have that tinkerer, you've got to have that kind of tinkerer mindset 206 00:13:16,140 --> 00:13:17,720 to make something work. 207 00:13:17,720 --> 00:13:19,840 Most people don't have that tinkerer mindset. 208 00:13:19,840 --> 00:13:24,440 So if it doesn't just work for them straight away, they're just going to go, oh, it was 209 00:13:24,440 --> 00:13:27,760 novel it was interesting, but no, not for me. 210 00:13:27,760 --> 00:13:31,000 And that's kind of the response I got from people in that room when I showed them that 211 00:13:31,000 --> 00:13:35,440 they were like, oh, that's, it's cool that it can do that, but don't think harm. 212 00:13:35,440 --> 00:13:38,000 I think I'll just spend the hour making the form. 213 00:13:38,000 --> 00:13:39,120 Well and this is it right? 214 00:13:39,120 --> 00:13:44,840 Like that will slow adoption and maybe we'll all learn over time just precisely enough 215 00:13:44,840 --> 00:13:47,840 how to prompt these things to get what we like out of it. 216 00:13:47,840 --> 00:13:49,640 Like the actual intended output. 217 00:13:49,640 --> 00:13:56,080 I think also your story there, you spotted that you needed to give it the ID, right? 218 00:13:56,080 --> 00:14:03,280 So again, it's an informed person running the process who can check the output and go, 219 00:14:03,280 --> 00:14:04,280 that's not right. 220 00:14:04,280 --> 00:14:08,400 I think somebody who wouldn't be able to see the technical aspects there or understand 221 00:14:08,400 --> 00:14:12,040 them or know they're important might be like, oh, this is broken. 222 00:14:12,040 --> 00:14:13,040 Yeah. 223 00:14:13,040 --> 00:14:15,960 Right and then you've got a problem, solve it. 224 00:14:15,960 --> 00:14:21,000 And I think even I was reading another one of the 10,000 bajillion AI newsletters that 225 00:14:21,000 --> 00:14:28,320 I'm signed up to about just how important ongoing prompting is like expect to go in 226 00:14:28,320 --> 00:14:33,520 and having to have five to 10 prompt conversation to get what you want. 227 00:14:33,520 --> 00:14:37,600 And if you go in with that mindset, then you're probably going to get some really good outputs 228 00:14:37,600 --> 00:14:39,680 out of say chat GPT. 229 00:14:39,680 --> 00:14:44,880 And I've seen people produce some really great starter prompts that just basically frame 230 00:14:44,880 --> 00:14:45,920 the conversation. 231 00:14:45,920 --> 00:14:48,360 Like this is who you are, chat GPT. 232 00:14:48,360 --> 00:14:50,400 This is the type of thing we're going to try and achieve. 233 00:14:50,400 --> 00:14:53,560 I want you to ask me really great questions to help improve the output. 234 00:14:53,560 --> 00:14:56,240 But then eventually I'm going to ask you to produce the output. 235 00:14:56,240 --> 00:15:00,300 And then there's a bit of back and forth and it coaches you through extracting the information 236 00:15:00,300 --> 00:15:01,300 it needs from you. 237 00:15:01,300 --> 00:15:05,280 And I felt it was really powerful, but that's not to your point, what people are expecting 238 00:15:05,280 --> 00:15:06,280 when they log in. 239 00:15:06,280 --> 00:15:07,280 No. 240 00:15:07,280 --> 00:15:14,640 It's going to be fascinating to see if, because there was also back along Sam Altman and the 241 00:15:14,640 --> 00:15:20,400 team at OpenAI were on a variety of podcasts, he said something similar, which is when we 242 00:15:20,400 --> 00:15:24,600 get these tools where they need to be, you won't need to be a prompt engineer because 243 00:15:24,600 --> 00:15:29,760 we want the tools to be able to figure out what you want quickly and easily without having 244 00:15:29,760 --> 00:15:33,320 to learn how to get what you want. 245 00:15:33,320 --> 00:15:38,360 Obviously that would remove this, but I don't think we're going to get to that because half 246 00:15:38,360 --> 00:15:41,280 the time it's because the human doesn't really know what they want. 247 00:15:41,280 --> 00:15:44,280 Like it's in their brain, but they're not articulated in any clear way. 248 00:15:44,280 --> 00:15:47,200 And if they said it to another human, the other human won't have a clue either. 249 00:15:47,200 --> 00:15:48,840 So of course the machine doesn't know. 250 00:15:48,840 --> 00:15:49,840 Yeah. 251 00:15:49,840 --> 00:15:50,840 Yeah. 252 00:15:50,840 --> 00:15:55,000 If you were telling a member of staff or a colleague, you'd be going kind of back and 253 00:15:55,000 --> 00:15:57,120 forth and figuring it out on the fly with them. 254 00:15:57,120 --> 00:15:58,120 So yeah, absolutely. 255 00:15:58,120 --> 00:16:00,520 It's always going to be iterative, isn't it? 256 00:16:00,520 --> 00:16:02,520 Again, it's like it's an assistant. 257 00:16:02,520 --> 00:16:03,520 Yeah. 258 00:16:03,520 --> 00:16:08,240 We called it the alien intern when I was at the SAMPS event the other week. 259 00:16:08,240 --> 00:16:11,080 And I do think that is the best way to look at it, right? 260 00:16:11,080 --> 00:16:15,680 Give it a precise brief because it doesn't have the background of knowledge and expertise 261 00:16:15,680 --> 00:16:18,040 that you do because you really have to spell it out. 262 00:16:18,040 --> 00:16:22,880 And every now and again, it will come up with something completely like it doesn't understand 263 00:16:22,880 --> 00:16:27,120 how people work or how the world works because it's not a human. 264 00:16:27,120 --> 00:16:28,120 Right. 265 00:16:28,120 --> 00:16:30,240 Let's crack into our main stories. 266 00:16:30,240 --> 00:16:35,800 The first main story this week is about this event, Martin. 267 00:16:35,800 --> 00:16:38,600 You're going to take us through what you saw here. 268 00:16:38,600 --> 00:16:39,600 Yeah. 269 00:16:39,600 --> 00:16:45,280 So Meta have been one of the headline sponsors of the big annual advertising event, the Cannes 270 00:16:45,280 --> 00:16:47,920 Lions. 271 00:16:47,920 --> 00:16:53,960 And this is where all of the major advertisers, the big players all get together. 272 00:16:53,960 --> 00:16:57,800 Everyone that's making waves in the advertising industry and there's a big award as well for 273 00:16:57,800 --> 00:17:00,000 the best campaigns and what have you. 274 00:17:00,000 --> 00:17:04,640 But Meta have been talking about their investments in AI and what that means for businesses and 275 00:17:04,640 --> 00:17:06,400 for marketers in particular. 276 00:17:06,400 --> 00:17:08,720 So they said they've been investing in AI powered technologies. 277 00:17:08,720 --> 00:17:09,720 We know this. 278 00:17:09,720 --> 00:17:17,120 We've been covering their research papers and new product launches every single week. 279 00:17:17,120 --> 00:17:21,580 And obviously a big area of focus for them is generative AI. 280 00:17:21,580 --> 00:17:26,840 So they said there's three areas of focus for them, which includes AI assistance, AI 281 00:17:26,840 --> 00:17:30,640 stickers and AI media editing. 282 00:17:30,640 --> 00:17:33,400 Now the assistance is exactly as you'd expect it to be. 283 00:17:33,400 --> 00:17:39,600 It's going to be assistance for businesses running through the likes of Messenger, WhatsApp 284 00:17:39,600 --> 00:17:43,800 and DMs and Instagram and things like that. 285 00:17:43,800 --> 00:17:50,080 And these are going to be used for things like customer experience, sorry, not customer 286 00:17:50,080 --> 00:17:56,240 experience, customer support, but also providing interesting customer experiences. 287 00:17:56,240 --> 00:18:02,760 So you can totally imagine interacting with certain brand voices or influencers that have 288 00:18:02,760 --> 00:18:06,680 aligned with a brand that you'll be able to kind of play games with them or just kind 289 00:18:06,680 --> 00:18:09,320 of Q&A them. 290 00:18:09,320 --> 00:18:10,320 That kind of thing. 291 00:18:10,320 --> 00:18:14,320 They're talking about bringing brand experiences to these assistants. 292 00:18:14,320 --> 00:18:17,140 There seems to be some suggestion as well that they're really looking at turning these 293 00:18:17,140 --> 00:18:19,000 into AI agents as well. 294 00:18:19,000 --> 00:18:26,000 So how we saw Google talking about Bard being able to book tickets for you and like you 295 00:18:26,000 --> 00:18:28,520 can actually make purchases and transact through there. 296 00:18:28,520 --> 00:18:36,560 I think that's where Meta have also got an eye on these developing in the future as well. 297 00:18:36,560 --> 00:18:38,120 There was also these AI stickers. 298 00:18:38,120 --> 00:18:42,060 Now this one was an interesting one to me because I hadn't really paid any attention 299 00:18:42,060 --> 00:18:43,720 to stickers at all. 300 00:18:43,720 --> 00:18:47,880 Have you used stickers in any meaning for sense? 301 00:18:47,880 --> 00:18:48,880 Not at all. 302 00:18:48,880 --> 00:18:52,560 In fact, when you said it, I was like, I don't even know what that is. 303 00:18:52,560 --> 00:18:57,760 Yeah, so I've been kind of vaguely aware if you're in WhatsApp, you can post a sticker 304 00:18:57,760 --> 00:19:02,400 and I've seen, you know, is it like business fish was one that was going around a while 305 00:19:02,400 --> 00:19:08,880 ago of it's a little sticker of a fish in a business soup and it's in like different 306 00:19:08,880 --> 00:19:09,880 scenarios. 307 00:19:09,880 --> 00:19:11,640 Anyway, it's ridiculous. 308 00:19:11,640 --> 00:19:14,640 It sounds it. 309 00:19:14,640 --> 00:19:17,360 I was in a group chat the other week. 310 00:19:17,360 --> 00:19:22,960 I was on a stag do and somebody turned one of the photos from the stag weekend into a 311 00:19:22,960 --> 00:19:26,200 sticker and posted it into the group. 312 00:19:26,200 --> 00:19:27,600 Now I know what you mean. 313 00:19:27,600 --> 00:19:28,600 Yes. 314 00:19:28,600 --> 00:19:30,000 Just like those little images. 315 00:19:30,000 --> 00:19:32,400 You just had to say stag do and I got it. 316 00:19:32,400 --> 00:19:33,400 No. 317 00:19:33,400 --> 00:19:34,400 Yeah, right. 318 00:19:34,400 --> 00:19:35,400 I'm there. 319 00:19:35,400 --> 00:19:36,400 I'm on board. 320 00:19:36,400 --> 00:19:40,800 And I loved the way that, so the one that I saw is someone had, you know, cropped, basically 321 00:19:40,800 --> 00:19:48,080 cropped the the silhouette of my mate and did a cut out of them and turned into a sticker. 322 00:19:48,080 --> 00:19:53,400 Businesses are going to be able to use these stickers and AI generate these stickers and 323 00:19:53,400 --> 00:19:58,720 basically have people use them, posting them in conversations. 324 00:19:58,720 --> 00:20:01,960 There's a particular phrase that they talked about. 325 00:20:01,960 --> 00:20:07,920 They said, we can imagine a world where businesses are able to create customer stickers tied 326 00:20:07,920 --> 00:20:13,640 to moments in time or a marketing campaign that would allow for a new creative medium 327 00:20:13,640 --> 00:20:15,680 to engage and delight their customers. 328 00:20:15,680 --> 00:20:16,920 Do you know what? 329 00:20:16,920 --> 00:20:21,880 I hear that part and I look at this and I'm like, yeah, AI driven customer support and 330 00:20:21,880 --> 00:20:27,480 customer experience through generative AI that then frees up humans to do other things 331 00:20:27,480 --> 00:20:29,520 and could be done personalized at scale. 332 00:20:29,520 --> 00:20:31,400 Huge commercial value. 333 00:20:31,400 --> 00:20:33,440 AI generated stickers. 334 00:20:33,440 --> 00:20:36,740 I'm like, I don't even understand what the point of that is. 335 00:20:36,740 --> 00:20:38,440 So you know what's going to happen? 336 00:20:38,440 --> 00:20:39,440 That'll blow up. 337 00:20:39,440 --> 00:20:42,880 How ridiculous, because it could be a commercial value. 338 00:20:42,880 --> 00:20:47,240 And I can't quite see it because I'm not an expert in that area and I'm probably going 339 00:20:47,240 --> 00:20:48,360 to end up being way off. 340 00:20:48,360 --> 00:20:52,240 But the juxtaposition of those seems quite strange to me. 341 00:20:52,240 --> 00:20:54,560 Yeah, it is, isn't it? 342 00:20:54,560 --> 00:20:56,360 But again, I'm like you, I don't really see it. 343 00:20:56,360 --> 00:21:02,320 But before you know it, half a trillion in market cap will be added to Meta based purely 344 00:21:02,320 --> 00:21:03,680 on their sticker strategy. 345 00:21:03,680 --> 00:21:06,920 I mean, they must know so much more than us. 346 00:21:06,920 --> 00:21:11,480 So one assumes there's big commercial opportunity in offering AI generated stickers, but I'd 347 00:21:11,480 --> 00:21:12,520 love to see it. 348 00:21:12,520 --> 00:21:17,040 They say billions of stickers are sent on Meta platforms every month, billions every 349 00:21:17,040 --> 00:21:18,040 month. 350 00:21:18,040 --> 00:21:21,480 I mean, I've seen one in like the past six months. 351 00:21:21,480 --> 00:21:23,080 I think it must be young people, Mark. 352 00:21:23,080 --> 00:21:24,880 I think it must be young people. 353 00:21:24,880 --> 00:21:25,880 Yeah. 354 00:21:25,880 --> 00:21:26,880 We're not young anymore. 355 00:21:26,880 --> 00:21:27,880 We've got to own that. 356 00:21:27,880 --> 00:21:28,880 Right? 357 00:21:28,880 --> 00:21:31,760 That ship has well and truly sailed. 358 00:21:31,760 --> 00:21:36,880 I did that speaking of being young, I went to a like a day rave on this stag day the 359 00:21:36,880 --> 00:21:39,800 other week and my word did I feel old. 360 00:21:39,800 --> 00:21:43,520 There was a 25 year old and I thought you look old. 361 00:21:43,520 --> 00:21:50,240 So if you look old and I'm considerably older than you, crikey. 362 00:21:50,240 --> 00:21:52,080 You had me at day rave. 363 00:21:52,080 --> 00:21:54,480 That sounds like what an old person would go to. 364 00:21:54,480 --> 00:21:56,200 Anyway, we digress. 365 00:21:56,200 --> 00:21:57,760 Tell us more about this, friend. 366 00:21:57,760 --> 00:21:59,720 It ended at 10pm. 367 00:21:59,720 --> 00:22:00,720 It's fantastic. 368 00:22:00,720 --> 00:22:03,160 It's ideal you're in bed by then. 369 00:22:03,160 --> 00:22:04,160 Perfect. 370 00:22:04,160 --> 00:22:05,160 Anyway, yeah. 371 00:22:05,160 --> 00:22:07,360 So what else did they introduce? 372 00:22:07,360 --> 00:22:09,920 They announced their AI media editing. 373 00:22:09,920 --> 00:22:20,920 So image generation models and basically being able to edit videos, so AI generated videos. 374 00:22:20,920 --> 00:22:22,640 Things are going to help creative campaigns. 375 00:22:22,640 --> 00:22:28,720 So you can take some of your existing imagery and maybe put a seasonal twist on it without 376 00:22:28,720 --> 00:22:30,440 having to do a whole new reshoot. 377 00:22:30,440 --> 00:22:35,920 That's quite an interesting concept if you've done some like product videography and then 378 00:22:35,920 --> 00:22:44,240 you want to put it in a Halloween setting or Christmas setting, you can do that using 379 00:22:44,240 --> 00:22:45,240 AI. 380 00:22:45,240 --> 00:22:50,920 So again, speaking about where the commercial applications for these kinds of things come 381 00:22:50,920 --> 00:22:51,920 in, totally get it. 382 00:22:51,920 --> 00:23:00,920 I can see how that is going to dramatically help creative teams, small business owners, 383 00:23:00,920 --> 00:23:01,920 large marketing teams. 384 00:23:01,920 --> 00:23:08,560 Doesn't matter who, you're going to be able to use that to get better creative done quickly 385 00:23:08,560 --> 00:23:13,880 or maybe not even better creative, just more variations of existing creative for more suitable 386 00:23:13,880 --> 00:23:18,000 applications like I say seasonality and what have you. 387 00:23:18,000 --> 00:23:23,880 Who knows, you can maybe even turn those interesting creatives into stickers and that becomes your 388 00:23:23,880 --> 00:23:25,920 new lead campaign. 389 00:23:25,920 --> 00:23:32,960 Yeah, I think I'm going to have to see that to believe that, but okay. 390 00:23:32,960 --> 00:23:36,480 And like I said, definitely turn out to be wrong because there's so much smarter people 391 00:23:36,480 --> 00:23:41,200 who know more about this than me who clearly think this is going to be a big thing. 392 00:23:41,200 --> 00:23:49,000 But yeah, all in a whole load of tools there for marketers. 393 00:23:49,000 --> 00:23:50,960 Interesting stuff. 394 00:23:50,960 --> 00:23:51,960 Interesting stuff. 395 00:23:51,960 --> 00:23:52,960 Okay. 396 00:23:52,960 --> 00:23:59,600 Well, on the generative side, there was an absolute ton of generative image news this 397 00:23:59,600 --> 00:24:00,680 week. 398 00:24:00,680 --> 00:24:10,920 So there was Mid Journey 5.2, so improvements to Mid Journey that includes a zoom out feature. 399 00:24:10,920 --> 00:24:17,600 So, we've talked a bit about Clip Drops, Uncrop and we've talked a bit about Photoshop, Adobe 400 00:24:17,600 --> 00:24:26,440 Photoshop's ability to produce really quite accurate expansions of images, filling in, 401 00:24:26,440 --> 00:24:30,120 you know, allowing you to basically see the area around an image that wasn't real, but 402 00:24:30,120 --> 00:24:32,080 it actually looks quite real. 403 00:24:32,080 --> 00:24:36,160 So it's not a surprise maybe that Mid Journey is looking to get in on the act there with 404 00:24:36,160 --> 00:24:38,920 their zoom out feature. 405 00:24:38,920 --> 00:24:43,720 I've been having a little play with this and it actually works reasonably well, which I 406 00:24:43,720 --> 00:24:47,080 guess it would given the fact that the other tools have managed to do this pretty good 407 00:24:47,080 --> 00:24:48,680 as well. 408 00:24:48,680 --> 00:24:54,860 There is a small increase in image quality there as well, I think. 409 00:24:54,860 --> 00:25:02,200 So then on the back of that, we also have Stability AI has launched a new AI model called 410 00:25:02,200 --> 00:25:09,200 SDXL 0.9, which you can access, I think, on Dream Studio and probably on Clip Drop as 411 00:25:09,200 --> 00:25:13,440 well, which produces more photorealistic images. 412 00:25:13,440 --> 00:25:18,360 And I haven't had a play with that one, but the examples I've seen certainly look quite 413 00:25:18,360 --> 00:25:19,360 a significant improvement. 414 00:25:19,360 --> 00:25:21,920 I don't know if you've managed to have a go with this yet. 415 00:25:21,920 --> 00:25:25,360 Yeah, I had a play with it this afternoon. 416 00:25:25,360 --> 00:25:29,760 I mean, it was released about 12 hours ago, so not had a great deal of time to jump into 417 00:25:29,760 --> 00:25:31,480 it, but they look really smart. 418 00:25:31,480 --> 00:25:32,880 I think they're doing a good job with that. 419 00:25:32,880 --> 00:25:35,640 I tested it with a bunch of different situations. 420 00:25:35,640 --> 00:25:44,320 I tried photorealistic images of snowboarding on the Alps and that kind of thing and they 421 00:25:44,320 --> 00:25:45,320 looked fantastic. 422 00:25:45,320 --> 00:25:50,000 Certainly, the reason I picked that one as well is I had a comparison from the old model 423 00:25:50,000 --> 00:25:54,080 that I could do light for light comparison and it was definitely a step change. 424 00:25:54,080 --> 00:25:55,080 Right. 425 00:25:55,080 --> 00:25:57,640 So I think this is good news for those people. 426 00:25:57,640 --> 00:26:05,520 I've had feedback and I found it myself that accessing mid-journey through Discord and 427 00:26:05,520 --> 00:26:10,840 then trying to subscribe and it's quite hard. 428 00:26:10,840 --> 00:26:12,840 It feels a bit technical. 429 00:26:12,840 --> 00:26:14,600 Discord hasn't got the best user interface in the world. 430 00:26:14,600 --> 00:26:15,600 It's a bit confusing. 431 00:26:15,600 --> 00:26:19,680 And I think a lot of people have either given up or just decided not to use mid-journey 432 00:26:19,680 --> 00:26:23,760 because Discord and the way that you have to access it is so hard. 433 00:26:23,760 --> 00:26:29,240 And I think seeing Stability AI's models improve where they can get closer to the outputs of 434 00:26:29,240 --> 00:26:31,640 mid-journey is going to make a lot of people smile. 435 00:26:31,640 --> 00:26:35,000 It won't be long before mid-journey have got their own web apps though. 436 00:26:35,000 --> 00:26:40,040 They are asking for beta testers on that at the moment. 437 00:26:40,040 --> 00:26:46,120 So mid-journey is going to be bringing that to the table before long. 438 00:26:46,120 --> 00:26:47,120 Good stuff. 439 00:26:47,120 --> 00:26:51,200 I think that is definitely going to make access easier for people and probably help stimulate 440 00:26:51,200 --> 00:26:53,880 their commercial development as well. 441 00:26:53,880 --> 00:26:57,480 There's more generative image news. 442 00:26:57,480 --> 00:27:04,560 We've got Meta launching their AI image creation model iJEPA, which can complete unfinished 443 00:27:04,560 --> 00:27:08,080 images more accurately than existing models. 444 00:27:08,080 --> 00:27:12,000 And they will be open sourcing the training code and model checkpoints. 445 00:27:12,000 --> 00:27:18,400 So if this ends up being commercialized, it's just another sort of tool in the toolbox for 446 00:27:18,400 --> 00:27:25,400 creatives to amend their images and add things that weren't there before or remove aspects 447 00:27:25,400 --> 00:27:27,680 of images that aren't appropriate. 448 00:27:27,680 --> 00:27:33,200 So for example, some of the examples that are in the blog post from Meta announcing 449 00:27:33,200 --> 00:27:37,160 this show sections of images completely cut out. 450 00:27:37,160 --> 00:27:42,240 So you could imagine, I don't know, an animal obscured by a bus. 451 00:27:42,240 --> 00:27:45,360 You could only see the back end of the animal, but you want to see the front end. 452 00:27:45,360 --> 00:27:49,360 Its ability to paint that back in looks quite interesting. 453 00:27:49,360 --> 00:27:51,560 So I think that's going to be quite cool. 454 00:27:51,560 --> 00:27:55,360 So lots of improvements in image generation. 455 00:27:55,360 --> 00:27:58,400 We've talked a lot about image generation on the show. 456 00:27:58,400 --> 00:28:03,400 And I think these are getting more and more impressive all the time. 457 00:28:03,400 --> 00:28:10,800 And this has all come to a head this week when Marvel Studios used AI, we believe Mid-Journey, 458 00:28:10,800 --> 00:28:16,400 unsubstantiated at this point, to create the opening of its new series on Disney Plus Secret 459 00:28:16,400 --> 00:28:17,640 Invasion. 460 00:28:17,640 --> 00:28:25,060 So this is not some designers tinkering with some creative ideas as inspiration for a small 461 00:28:25,060 --> 00:28:26,600 campaign for a small brand. 462 00:28:26,600 --> 00:28:36,520 This is Marvel and Disney, arguably some of the biggest sort of brands on the planet, 463 00:28:36,520 --> 00:28:40,240 using these tools to generate the opening credits. 464 00:28:40,240 --> 00:28:45,440 And there's been a bit of a hullabalaa, hasn't there, about this this week, Martin? 465 00:28:45,440 --> 00:28:46,440 There has. 466 00:28:46,440 --> 00:28:48,840 It has ruffled some feathers. 467 00:28:48,840 --> 00:28:56,580 And you can understand why when the Hollywood screenwriters strike is currently on. 468 00:28:56,580 --> 00:29:00,760 And one of the things that they've mentioned in that is that they want to stop studios 469 00:29:00,760 --> 00:29:05,720 from using AI for writing purposes. 470 00:29:05,720 --> 00:29:14,480 So to see this actually being used on a flagship series from Marvel, Disney is quite big. 471 00:29:14,480 --> 00:29:20,720 I've got a few tweets here from people commenting on it. 472 00:29:20,720 --> 00:29:22,320 Secret Invasion intro is AI generated. 473 00:29:22,320 --> 00:29:23,320 I'm devastated. 474 00:29:23,320 --> 00:29:29,440 I believe AI to be unethical, dangerous and designed solely to eliminate artists careers. 475 00:29:29,440 --> 00:29:33,720 Spent almost half a year working on this show and had a fantastic experience working with 476 00:29:33,720 --> 00:29:36,280 the most amazing people I've ever met. 477 00:29:36,280 --> 00:29:37,700 That's from a Jeff Simpson. 478 00:29:37,700 --> 00:29:42,520 So people working closely on the show were voicing their concerns. 479 00:29:42,520 --> 00:29:45,360 Yeah, and you can understand it really. 480 00:29:45,360 --> 00:29:49,440 I mean, it's the type of thing that would have usually been produced by a company and 481 00:29:49,440 --> 00:29:50,540 by a team. 482 00:29:50,540 --> 00:29:57,680 You would imagine a fair bit of money gets spent on these things and keeps a reasonable 483 00:29:57,680 --> 00:30:00,720 number of people busy producing those things. 484 00:30:00,720 --> 00:30:01,800 So it's really interesting. 485 00:30:01,800 --> 00:30:08,200 The argument that I'll paraphrase probably badly made by Marvel around this was they 486 00:30:08,200 --> 00:30:10,880 felt it was in keeping with the show. 487 00:30:10,880 --> 00:30:16,320 Because obviously with Mid-Journey you get not quite photorealistic images. 488 00:30:16,320 --> 00:30:22,440 They look a bit strange and the whole premise of this series is nothing is as it seems. 489 00:30:22,440 --> 00:30:28,100 And it's basically about some shape-shifting creatures who can look like people. 490 00:30:28,100 --> 00:30:35,160 So they kind of wanted it to have that unrealistic quality, this morphing quality that I think 491 00:30:35,160 --> 00:30:40,000 you can get some quite cool stuff out of Mid-Journey when you're trying to do that. 492 00:30:40,000 --> 00:30:47,840 But if you ask me, my bull poo meter was going ding, ding, ding, ding, because that sounded 493 00:30:47,840 --> 00:30:52,200 just like a good excuse as to why they did it. 494 00:30:52,200 --> 00:30:53,640 Maybe they did it for money reasons. 495 00:30:53,640 --> 00:30:55,680 Maybe somebody went, oh, wouldn't it be cool? 496 00:30:55,680 --> 00:30:59,840 Let's use AI for this and then just didn't realize the backlash it was going to cause. 497 00:30:59,840 --> 00:31:00,840 I don't know. 498 00:31:00,840 --> 00:31:01,840 But interesting. 499 00:31:01,840 --> 00:31:11,760 I mean, personally, it's difficult to criticize it when you're talking about these technologies 500 00:31:11,760 --> 00:31:13,080 every week as we are doing. 501 00:31:13,080 --> 00:31:17,200 I can understand why people are doing, but ultimately there was a team of people sat 502 00:31:17,200 --> 00:31:19,960 behind this doing the prompting, doing the editing. 503 00:31:19,960 --> 00:31:26,640 I'm sure they didn't just type in some words into Mid-Journey and get this out in an instant. 504 00:31:26,640 --> 00:31:29,160 It wasn't like a one-shot thing. 505 00:31:29,160 --> 00:31:33,840 Someone's had to prompt it in all the different scenes, stitch it together, edit it, do a 506 00:31:33,840 --> 00:31:35,800 lot of work over the top of it. 507 00:31:35,800 --> 00:31:38,600 So yeah, I think maybe the... 508 00:31:38,600 --> 00:31:44,040 I can understand why people feel upset, but there were people being paid to do that. 509 00:31:44,040 --> 00:31:50,720 I'm sure a company or someone took a fee and that hasn't done people out of work too much. 510 00:31:50,720 --> 00:31:51,720 Yeah. 511 00:31:51,720 --> 00:31:55,720 I'd be interested to see what the workflow looked like and how much time it took versus 512 00:31:55,720 --> 00:32:02,280 going down a more traditional route because iteration being a bit of a theme for the whole 513 00:32:02,280 --> 00:32:05,880 podcast as we started, but certainly today, you're right. 514 00:32:05,880 --> 00:32:09,160 I don't think they would have got the outputs they wanted without some iteration. 515 00:32:09,160 --> 00:32:11,600 And of course, it's an intro to a TV show. 516 00:32:11,600 --> 00:32:13,000 It lasts about a minute. 517 00:32:13,000 --> 00:32:15,200 There's quite a few images in there. 518 00:32:15,200 --> 00:32:19,720 Yes, I suspect it would have taken longer to create a bunch of those images from scratch, 519 00:32:19,720 --> 00:32:25,880 but it's not like you said, a morning's worth of work and they already add production ready 520 00:32:25,880 --> 00:32:28,680 images and they literally just needed to stitch them together. 521 00:32:28,680 --> 00:32:30,000 I doubt that was the case. 522 00:32:30,000 --> 00:32:31,160 So be interested to see. 523 00:32:31,160 --> 00:32:32,520 I'd love to know more about that. 524 00:32:32,520 --> 00:32:35,920 If you were involved in the process of putting this together and you want to come on the 525 00:32:35,920 --> 00:32:39,720 podcast, tell us what workflow you went through so that actually we can get a bit of feel 526 00:32:39,720 --> 00:32:45,000 for what the creative process looks like when you're creating TV titles using AI. 527 00:32:45,000 --> 00:32:48,600 We'd love to hear about that. 528 00:32:48,600 --> 00:32:53,760 Let's move on from generative image news to another meta story. 529 00:32:53,760 --> 00:32:55,600 This is the meta episode. 530 00:32:55,600 --> 00:32:58,280 They didn't sponsor it, Martin, so I am really unhappy. 531 00:32:58,280 --> 00:33:00,680 I'm going to be straight on the phone to my pal. 532 00:33:00,680 --> 00:33:01,680 Down to the cash. 533 00:33:01,680 --> 00:33:02,680 Yeah, yeah. 534 00:33:02,680 --> 00:33:06,240 If we're going to feature this much, man, we, you know, there's got to be, it's got 535 00:33:06,240 --> 00:33:10,000 to be a little something in it. 536 00:33:10,000 --> 00:33:12,260 I'm teasing, obviously. 537 00:33:12,260 --> 00:33:16,640 This one's about Meta's new text speech tool called Voicebox. 538 00:33:16,640 --> 00:33:18,920 Tell us about this, Martin, because the video is pretty cool, right? 539 00:33:18,920 --> 00:33:21,120 Yeah, I would recommend people go check out the blog. 540 00:33:21,120 --> 00:33:24,000 We'll put it in the show notes as usual. 541 00:33:24,000 --> 00:33:27,880 So this is an all-in-one generative speech tool and we've seen generative speech tools. 542 00:33:27,880 --> 00:33:31,000 We spoke about 11 labs at the start of the session. 543 00:33:31,000 --> 00:33:35,360 They are doing a synthetic voice and their product's really cool. 544 00:33:35,360 --> 00:33:42,440 But what Meta have done with this particular model is really over and above what we've 545 00:33:42,440 --> 00:33:44,040 seen before by some way. 546 00:33:44,040 --> 00:33:46,080 So this is a text-to-speech synthesis. 547 00:33:46,080 --> 00:33:50,560 You type in some words and a voice will say those words. 548 00:33:50,560 --> 00:33:57,240 It synthesizes speech across six languages, including taking your voice in one language 549 00:33:57,240 --> 00:34:02,680 so you can kind of train it and then it will make your voice speak in another language. 550 00:34:02,680 --> 00:34:03,680 So that's pretty cool. 551 00:34:03,680 --> 00:34:09,720 Because, you know, I don't speak Portuguese, but we could make me speak Portuguese. 552 00:34:09,720 --> 00:34:15,440 It can also perform tasks that it wasn't trained on, which is this emergent capability that 553 00:34:15,440 --> 00:34:18,080 we see in LLMs. 554 00:34:18,080 --> 00:34:22,440 It has noise removal, which in the demo is really neat. 555 00:34:22,440 --> 00:34:29,000 There's a clip of Mark Zuckerberg speaking with a dog barking in the background and then 556 00:34:29,000 --> 00:34:30,880 it cuts that out completely. 557 00:34:30,880 --> 00:34:37,920 So where we've seen similar things from other providers, where voices can be optimized to 558 00:34:37,920 --> 00:34:40,800 podcasts like with Adobe Studio. 559 00:34:40,800 --> 00:34:41,800 Is it Adobe Studio? 560 00:34:41,800 --> 00:34:42,800 Is that what it's called? 561 00:34:42,800 --> 00:34:43,800 Something like that. 562 00:34:43,800 --> 00:34:46,720 Adobe Enhance, I think the one we use on the podcast this year. 563 00:34:46,720 --> 00:34:47,720 Yeah. 564 00:34:47,720 --> 00:34:50,280 So it's like that, but more so. 565 00:34:50,280 --> 00:34:57,920 So it can give it that crisp audio feel, but also remove background noise quite strongly 566 00:34:57,920 --> 00:34:58,920 as well. 567 00:34:58,920 --> 00:35:08,320 It has language conversion, so it will take your voice and change the language, even when 568 00:35:08,320 --> 00:35:10,440 the two things are written in different parts. 569 00:35:10,440 --> 00:35:14,880 I wasn't fully sure I understood that, but it all sounds very cool and it's 20 times 570 00:35:14,880 --> 00:35:19,840 faster than current models and it outperforms single purpose models. 571 00:35:19,840 --> 00:35:23,080 So models that have been trained specifically for that task. 572 00:35:23,080 --> 00:35:24,920 And it does that through in-contact learning. 573 00:35:24,920 --> 00:35:27,080 So incredibly capable. 574 00:35:27,080 --> 00:35:28,440 Would love to get my hands on it. 575 00:35:28,440 --> 00:35:31,120 I do like playing with these kinds of tools. 576 00:35:31,120 --> 00:35:36,680 Lovo.ai and Jenny we featured on the show previously. 577 00:35:36,680 --> 00:35:40,760 And that's pretty cool, but just falls down in some areas. 578 00:35:40,760 --> 00:35:46,280 Unfortunately, as much as I would like to get my hands on it, Meta are not opening this 579 00:35:46,280 --> 00:35:47,440 up to public release. 580 00:35:47,440 --> 00:35:49,180 This is not going to be available at the moment. 581 00:35:49,180 --> 00:35:53,640 This is very much still in their research paper and hey, look, isn't this a cool thing 582 00:35:53,640 --> 00:35:55,920 that we've done, FaZe? 583 00:35:55,920 --> 00:36:00,880 So I'm sure they're going to commercialize it through the AI assistants and AI agents 584 00:36:00,880 --> 00:36:02,780 that we spoke about earlier in the episode. 585 00:36:02,780 --> 00:36:08,280 You can start to see how all of these things are pieced together. 586 00:36:08,280 --> 00:36:12,840 Get a celebrity, train their voice, stick that into the AI agent. 587 00:36:12,840 --> 00:36:18,280 Then all of a sudden you've got a nice interactive avatar where you're speaking to the latest 588 00:36:18,280 --> 00:36:24,680 celebrity on Love Island or Sportstar or whatever it may be. 589 00:36:24,680 --> 00:36:26,600 So until then, they're not doing it. 590 00:36:26,600 --> 00:36:30,440 And the reason they say they're not doing it is because of the potential for misuse, 591 00:36:30,440 --> 00:36:31,680 which I do buy. 592 00:36:31,680 --> 00:36:36,520 I do think that this does have a great potential for misuse based on the quality of the outputs 593 00:36:36,520 --> 00:36:38,560 we see on the demo. 594 00:36:38,560 --> 00:36:44,780 But also I can see that from a commercial perspective, it would be quite beneficial 595 00:36:44,780 --> 00:36:49,680 to just keep that on under wraps for a bit and say, no, you're not having it. 596 00:36:49,680 --> 00:36:55,680 You're going to integrate this into our tools and you lot can figure it out yourself if 597 00:36:55,680 --> 00:36:57,920 you want to use something similar. 598 00:36:57,920 --> 00:37:01,160 Yeah, it's really cool. 599 00:37:01,160 --> 00:37:03,720 And it just makes you wonder how far it can go. 600 00:37:03,720 --> 00:37:07,840 So the audiobook industry surely is about to fundamentally change. 601 00:37:07,840 --> 00:37:12,520 I mean, I could probably have Stephen Fry read every book I've got on my Kindle at some 602 00:37:12,520 --> 00:37:17,760 point because he's wonderful, by the way, when he does narrate the stories and I'm sure 603 00:37:17,760 --> 00:37:22,120 they could never capture his performance, if I'm really honest. 604 00:37:22,120 --> 00:37:27,800 But are we getting to the point where you buy a Kindle book and you can have it, the 605 00:37:27,800 --> 00:37:30,880 audiobook version in effect for the same price? 606 00:37:30,880 --> 00:37:31,880 I don't know, right? 607 00:37:31,880 --> 00:37:37,160 I currently buy quite a lot of audiobooks and Kindle books and sometimes I've got copies, 608 00:37:37,160 --> 00:37:40,720 the audio and the Kindle version. 609 00:37:40,720 --> 00:37:43,360 But will anyone need to go in a booth again? 610 00:37:43,360 --> 00:37:48,360 Or will we be able to A, use any famous person who commercialises their voice for this year 611 00:37:48,360 --> 00:37:49,360 use? 612 00:37:49,360 --> 00:37:54,880 Or, as I've seen with a number of other models, train these systems on about one minute of 613 00:37:54,880 --> 00:38:00,640 speech from us or the author and have them narrate their own audiobook without having 614 00:38:00,640 --> 00:38:01,640 to do the work. 615 00:38:01,640 --> 00:38:07,800 Yeah, I mean, when you start thinking about the actual productisation of this, that is 616 00:38:07,800 --> 00:38:09,880 absolutely where to go. 617 00:38:09,880 --> 00:38:17,280 Books are booming as an industry and the idea that, you know, self-published authors, a 618 00:38:17,280 --> 00:38:22,120 friend of mine at the moment is currently about to self-publish a sports performance 619 00:38:22,120 --> 00:38:23,120 book. 620 00:38:23,120 --> 00:38:28,360 And I said I would be thought about doing it as an audiobook and hadn't, well, then 621 00:38:28,360 --> 00:38:32,120 we discussed it and spoke about, you know, you can get a studio, you can go in, record 622 00:38:32,120 --> 00:38:34,720 it over eight hours or something. 623 00:38:34,720 --> 00:38:37,480 And the look on her face was just like, oh no, that sounds terrible. 624 00:38:37,480 --> 00:38:43,440 But if you could turn it into an audiobook and the press of a button, absolutely should 625 00:38:43,440 --> 00:38:44,440 do it. 626 00:38:44,440 --> 00:38:45,440 Absolutely. 627 00:38:45,440 --> 00:38:51,560 And I'm guessing you could have it speak for you in a very confident, eloquent manner, 628 00:38:51,560 --> 00:38:54,520 just using the timbre of your voice. 629 00:38:54,520 --> 00:38:59,240 Whereas you know, if a person might sit in a booth and feel a bit nervous and stumble 630 00:38:59,240 --> 00:39:03,520 over their words and not feel very confident about it, whereas the AI is going to have 631 00:39:03,520 --> 00:39:04,520 none of that, right? 632 00:39:04,520 --> 00:39:08,080 It's going to mimic your voice and then deliver this wonderful, clear version. 633 00:39:08,080 --> 00:39:11,280 So yeah, this is really interesting. 634 00:39:11,280 --> 00:39:16,480 And as marketers, obviously it's going to give us the ability to perhaps produce things 635 00:39:16,480 --> 00:39:20,280 in the voice of our SMEs, but without having to bother our SMEs, right? 636 00:39:20,280 --> 00:39:25,840 Imagine running a webinar where you can see the slides, but you can't see the person. 637 00:39:25,840 --> 00:39:30,480 And actually the webinar was scripted and you had one of your SMEs or your CEO as the 638 00:39:30,480 --> 00:39:34,240 voice, but you never really had to bother them and use their time. 639 00:39:34,240 --> 00:39:35,240 It could be pretty powerful. 640 00:39:35,240 --> 00:39:36,240 Right. 641 00:39:36,240 --> 00:39:37,920 Let's do our last story. 642 00:39:37,920 --> 00:39:39,320 We're cracking on through. 643 00:39:39,320 --> 00:39:44,000 This is about the project management tool ClickUp, which many people probably on the 644 00:39:44,000 --> 00:39:45,000 podcast will use. 645 00:39:45,000 --> 00:39:49,480 And if not, it's worth having a look because it's pretty expansive in terms of its capabilities. 646 00:39:49,480 --> 00:39:53,720 It is a project management tool, but it's got a Google Docs built in it. 647 00:39:53,720 --> 00:39:55,560 It's got automations built in it. 648 00:39:55,560 --> 00:40:00,080 And now it's got an AI powered assistant built into it. 649 00:40:00,080 --> 00:40:06,080 So if you look at the press release on the ClickUp website on their blog, it claims that 650 00:40:06,080 --> 00:40:13,000 it has a hundred plus AI tools optimized for every role and use case, which is pretty interesting. 651 00:40:13,000 --> 00:40:17,640 And that there's a, they've teamed up with dozens of experts to create fully templatized 652 00:40:17,640 --> 00:40:21,840 prompts to make it even easier for people to use. 653 00:40:21,840 --> 00:40:27,460 The I think one of the interesting things here is the text generation capabilities. 654 00:40:27,460 --> 00:40:32,460 So you can highlight any text within ClickUp, like a document, a comment, task description, 655 00:40:32,460 --> 00:40:37,840 and then you can use the AI toolbar to improve the content, or you can make it longer or 656 00:40:37,840 --> 00:40:40,540 shorter or simplify the writing style. 657 00:40:40,540 --> 00:40:44,120 So again, a lot of the things we've come to see in a lot of the writing tools that are 658 00:40:44,120 --> 00:40:48,840 out there, but now this is being baked into the project management tool that you use. 659 00:40:48,840 --> 00:40:50,920 Do you need your writing tool anymore? 660 00:40:50,920 --> 00:40:51,920 That's the question, right? 661 00:40:51,920 --> 00:40:54,320 If you're a ClickUp user, do you need Jasper? 662 00:40:54,320 --> 00:40:55,320 Right. 663 00:40:55,320 --> 00:40:58,960 So it's a tool that you can use to create a text, but it's also a tool that you can 664 00:40:58,960 --> 00:41:00,920 use to generate text from scratch. 665 00:41:00,920 --> 00:41:05,680 So you can use the text to generate text from scratch, which, you know, will help you with 666 00:41:05,680 --> 00:41:11,320 creating emails and blog posts, and it will summarize transcripts and other things that 667 00:41:11,320 --> 00:41:12,320 you, that you give it. 668 00:41:12,320 --> 00:41:17,320 So it's almost like ClickUp is becoming really more than a project management tool. 669 00:41:17,320 --> 00:41:20,320 And if ClickUp is doing this, we should expect other tools to do this. 670 00:41:20,320 --> 00:41:24,760 And this really fits in with something we've talked about a lot, Martin, in terms of which 671 00:41:24,760 --> 00:41:28,400 way, unless they can get a really good deal, like on a lifetime deal, they should not be 672 00:41:28,400 --> 00:41:34,960 subscribing for a year because if they're using six, seven, eight different tools, they're 673 00:41:34,960 --> 00:41:39,920 going to be consolidated into Copilot or Bard or ClickUp. 674 00:41:39,920 --> 00:41:41,880 And I think we're going to see a lot more of this. 675 00:41:41,880 --> 00:41:46,320 The use cases that ClickUp talk about offering are for marketing, you could easily write 676 00:41:46,320 --> 00:41:49,760 a case study or generate a creative brief. 677 00:41:49,760 --> 00:41:53,680 For sales, they suggest you could help use AI to help you create a territory plan, write 678 00:41:53,680 --> 00:41:56,720 prospecting emails for product development. 679 00:41:56,720 --> 00:42:03,200 You could write product requirement documentation, design a user testing study for engineering. 680 00:42:03,200 --> 00:42:07,120 You could create a sprint retrospective report, generate a model schema. 681 00:42:07,120 --> 00:42:09,680 I don't even know what that stuff is. 682 00:42:09,680 --> 00:42:11,760 Model schema part, I mean, project management. 683 00:42:11,760 --> 00:42:15,080 You can create a statement of work, spin up a project timeline. 684 00:42:15,080 --> 00:42:20,000 And all of this power is coming into ClickUp AI for an extra $5 per user per month. 685 00:42:20,000 --> 00:42:25,360 So it's already quite low cost, something's like the business level is like $12 a user. 686 00:42:25,360 --> 00:42:30,040 Imagine all the time and as you just saved and all the other tools you don't need, the 687 00:42:30,040 --> 00:42:31,640 competition is heating up. 688 00:42:31,640 --> 00:42:33,480 Now, ClickUp already has a lot of users. 689 00:42:33,480 --> 00:42:37,760 So when we've talked about motes on the show before, ClickUp has a moat because it's got 690 00:42:37,760 --> 00:42:40,000 really a lot of users already. 691 00:42:40,000 --> 00:42:41,600 But that's pretty cool. 692 00:42:41,600 --> 00:42:45,680 So if you've never had a play with ClickUp and you have a project management tool that 693 00:42:45,680 --> 00:42:48,540 you don't love or you're thinking, do you know what, we could probably be better organized 694 00:42:48,540 --> 00:42:52,240 if we actually just used a tool to enable it. 695 00:42:52,240 --> 00:42:57,920 Generative AI stuff that's been added to ClickUp might be worth going and having a look at. 696 00:42:57,920 --> 00:42:59,720 Right. 697 00:42:59,720 --> 00:43:04,400 Last bit today then is tool of the week, which is Perplexity AI. 698 00:43:04,400 --> 00:43:07,440 Martin, tell us about Perplexity. 699 00:43:07,440 --> 00:43:11,680 Perplexity is a really cool search engine. 700 00:43:11,680 --> 00:43:18,400 And it was when GPD launched and people were saying it didn't have real time data and all 701 00:43:18,400 --> 00:43:20,240 of this kind of thing. 702 00:43:20,240 --> 00:43:28,440 And before we got plugins and before we got browse with ChatGPT, Perplexity popped up and 703 00:43:28,440 --> 00:43:31,440 it's available on perplexity.ai. 704 00:43:31,440 --> 00:43:38,240 And it started out as just a very simple search box, which would you put in your search query 705 00:43:38,240 --> 00:43:43,400 and it is powered by ChatGPT. 706 00:43:43,400 --> 00:43:51,160 Or should I say GPT 3.5 and actually it does have some GPT-4 capabilities as well now. 707 00:43:51,160 --> 00:43:52,520 But basically it's a search engine. 708 00:43:52,520 --> 00:43:58,520 If you go on there now, you can sign up and you can put a profile in which will help tailor 709 00:43:58,520 --> 00:44:01,680 the results to you, but you can ask it anything. 710 00:44:01,680 --> 00:44:06,360 In fact, on the home screen, as you go on there, there are some examples already. 711 00:44:06,360 --> 00:44:08,320 So I'm just going to click on the one that I see now. 712 00:44:08,320 --> 00:44:12,160 It says, world's most livable cities 2023. 713 00:44:12,160 --> 00:44:18,540 And it's very similar to the kind of Google's generative search experience. 714 00:44:18,540 --> 00:44:24,720 So if you click on that result, it doesn't bring up a bunch of websites you can visit. 715 00:44:24,720 --> 00:44:28,880 Instead it answers the question and it says the Economist Intelligence Units Livability 716 00:44:28,880 --> 00:44:32,680 Index 2023 is ranked the world's most livable cities, et cetera. 717 00:44:32,680 --> 00:44:36,800 And it goes on and lists all of the cities, but it provides references as well. 718 00:44:36,800 --> 00:44:43,200 So against every fact that it provides, it gives you a reference to a source that you 719 00:44:43,200 --> 00:44:48,360 can then go and dig into that further. 720 00:44:48,360 --> 00:44:51,600 It's available as a Chrome extension. 721 00:44:51,600 --> 00:44:55,000 Now the Chrome extension is the tool that I use every day. 722 00:44:55,000 --> 00:44:57,140 It has a summarized feature. 723 00:44:57,140 --> 00:45:02,020 So when you're on any web page, you can just click on the perplexity icon and it brings 724 00:45:02,020 --> 00:45:06,680 up saying, you can ask it anything, but there's a big summarize button. 725 00:45:06,680 --> 00:45:11,800 You can just say this page and it will then summarize whatever the article is that you're 726 00:45:11,800 --> 00:45:15,200 reading and it will give you a kind of paragraph about that. 727 00:45:15,200 --> 00:45:19,240 But the neat thing about it is that it has, rather than it just being a summarization 728 00:45:19,240 --> 00:45:21,920 tool, it actually has a prompt window as well. 729 00:45:21,920 --> 00:45:26,640 So you can say, like you would with ChatGBT or something like that, you can say summarize 730 00:45:26,640 --> 00:45:33,040 it in the style of a pirate or summarize using only bullet points or something like that. 731 00:45:33,040 --> 00:45:34,940 So it's really neat. 732 00:45:34,940 --> 00:45:36,280 It's available on Android. 733 00:45:36,280 --> 00:45:40,020 It's available on iOS. 734 00:45:40,020 --> 00:45:42,760 It saves your threads as well. 735 00:45:42,760 --> 00:45:47,280 So if you go in and start going back and forth because it allows you to ask followups and 736 00:45:47,280 --> 00:45:48,280 all of that. 737 00:45:48,280 --> 00:45:50,400 It's basically a full search engine. 738 00:45:50,400 --> 00:45:55,160 Not only that, when you go on the homepage, I forgot this, this is a really neat feature. 739 00:45:55,160 --> 00:46:00,320 In the search box itself, in the bottom left hand corner, it says there's a little icon 740 00:46:00,320 --> 00:46:02,720 which will by default be all. 741 00:46:02,720 --> 00:46:05,760 And that basically searches across the whole of the internet. 742 00:46:05,760 --> 00:46:11,920 But if you want to get more specific and you want to focus specifically on maybe academic, 743 00:46:11,920 --> 00:46:19,000 so looking at published academic papers, or if you want to dig into YouTube or Reddit 744 00:46:19,000 --> 00:46:27,080 or Wikipedia, you could even plug into Wolfram Alpha if you need that kind of the maths based 745 00:46:27,080 --> 00:46:29,880 computational knowledge engine element. 746 00:46:29,880 --> 00:46:32,180 They partnered with Wolfram as well. 747 00:46:32,180 --> 00:46:38,600 So you can actually change the way that it searches and control it with quite a bit of 748 00:46:38,600 --> 00:46:39,600 detail. 749 00:46:39,600 --> 00:46:41,040 It's available for free. 750 00:46:41,040 --> 00:46:45,280 If you upgrade, you can actually pay $20 a month. 751 00:46:45,280 --> 00:46:54,460 If you choose to upgrade, you get access to, I think you get GPT-4 based responses. 752 00:46:54,460 --> 00:47:00,280 So there's a few add-ons available there that they try to... 753 00:47:00,280 --> 00:47:02,560 Squeeze in trying to... 754 00:47:02,560 --> 00:47:07,000 Yeah, trying to incentivise you to start paying as well. 755 00:47:07,000 --> 00:47:14,400 But they were featured on the Microsoft Azure blog recently about a company that's built 756 00:47:14,400 --> 00:47:18,720 on top of the Azure and OpenAI stack. 757 00:47:18,720 --> 00:47:19,720 It's a great tool. 758 00:47:19,720 --> 00:47:27,240 I've been following the founder on LinkedIn and Twitter since I first came across it. 759 00:47:27,240 --> 00:47:30,240 And yeah, can't speak highly enough about it really. 760 00:47:30,240 --> 00:47:34,680 It's one of my daily driver AI tools. 761 00:47:34,680 --> 00:47:35,680 Interesting. 762 00:47:35,680 --> 00:47:40,600 Yeah, I think I'll be having a play with that, especially the ability to search academic 763 00:47:40,600 --> 00:47:41,600 papers. 764 00:47:41,600 --> 00:47:44,960 I think for those of us working in the life sciences, that's worth playing with. 765 00:47:44,960 --> 00:47:50,560 I think we spoke about this last week, but I recently ran a test as part of the SAMS 766 00:47:50,560 --> 00:47:55,920 presentation that I was doing that we talked about last week, whereby I used a bunch of 767 00:47:55,920 --> 00:47:59,280 different tools to ask what the top five papers on stem cells were. 768 00:47:59,280 --> 00:48:00,960 And they all said different things. 769 00:48:00,960 --> 00:48:05,520 And some of the tools like Bard made up citations that weren't real. 770 00:48:05,520 --> 00:48:10,360 In fact, they pretty much all made up citations that weren't real in one way or another. 771 00:48:10,360 --> 00:48:14,200 So it'd be really interesting to see if this is better, because at this point, you would 772 00:48:14,200 --> 00:48:22,140 argue can you really trust Chachipiti or Bard or Claude if you are doing research or writing 773 00:48:22,140 --> 00:48:27,560 content about the life sciences where you need to be able to bring up and cite accurately 774 00:48:27,560 --> 00:48:30,840 real papers, which you can't trust. 775 00:48:30,840 --> 00:48:31,840 It makes papers up. 776 00:48:31,840 --> 00:48:36,160 So I'm going to have a play with this and see if it is any better. 777 00:48:36,160 --> 00:48:40,640 Thank you for sharing that with us and the listeners, Martin. 778 00:48:40,640 --> 00:48:41,640 Much appreciated. 779 00:48:41,640 --> 00:48:44,800 So I think that brings us to the end for today. 780 00:48:44,800 --> 00:48:48,240 Thank you very much for your time as always, Martin. 781 00:48:48,240 --> 00:48:54,960 It's been a pleasure just to want to update everyone on some other big news. 782 00:48:54,960 --> 00:48:58,000 The transfer window has opened last week. 783 00:48:58,000 --> 00:49:02,280 So Derby County have been making their first signings of the summer. 784 00:49:02,280 --> 00:49:04,400 So we've made four signings. 785 00:49:04,400 --> 00:49:05,400 Very exciting. 786 00:49:05,400 --> 00:49:09,520 We did resign Scott Lopes, the keeper, though he's a third choice, I would say. 787 00:49:09,520 --> 00:49:14,360 I think he's probably going to be more of a coaching role backup third place on the 788 00:49:14,360 --> 00:49:15,360 bench. 789 00:49:15,360 --> 00:49:17,960 But they're good to see some faces coming back into the squad. 790 00:49:17,960 --> 00:49:18,960 I think you'll agree. 791 00:49:18,960 --> 00:49:19,960 Yeah. 792 00:49:19,960 --> 00:49:24,800 I think the only thing about this entire episode, I think the thing that will make people angry 793 00:49:24,800 --> 00:49:26,400 is why was that not at the beginning? 794 00:49:26,400 --> 00:49:28,960 Because that's clearly the most important news of the week. 795 00:49:28,960 --> 00:49:29,960 Yeah. 796 00:49:29,960 --> 00:49:30,960 Yeah. 797 00:49:30,960 --> 00:49:32,960 No, I honestly, it nearly slipped my mind outrageously. 798 00:49:32,960 --> 00:49:35,840 But got it in there in the end, better late than never. 799 00:49:35,840 --> 00:49:36,840 You did well. 800 00:49:36,840 --> 00:49:40,080 I think people will breathe a sigh of relief that they also got a little bit of Derby County 801 00:49:40,080 --> 00:49:41,640 news and they can feel better about that. 802 00:49:41,640 --> 00:49:45,560 We're going to make Derby County the new Wrexham. 803 00:49:45,560 --> 00:49:50,400 And if we get them on Disney+, the intro credits will obviously be created with Mid Journey 804 00:49:50,400 --> 00:49:54,600 and other voice synthesis tools that we might use if Matt let us play with their model. 805 00:49:54,600 --> 00:49:58,240 But as always, thank you to you dear listener. 806 00:49:58,240 --> 00:49:59,240 We hope you've enjoyed. 807 00:49:59,240 --> 00:50:04,080 Please do share this with your marketing pals if you think they find it interesting. 808 00:50:04,080 --> 00:50:05,500 Subscribe if you haven't already. 809 00:50:05,500 --> 00:50:11,640 You can follow us on the Twitter's if you go to AI Marketing Pod. 810 00:50:11,640 --> 00:50:12,640 That's right. 811 00:50:12,640 --> 00:50:13,720 That's two weeks in a row. 812 00:50:13,720 --> 00:50:14,720 I agree. 813 00:50:14,720 --> 00:50:21,240 Or go to artificiallyintelligentmarketing.com where you can subscribe to the episodes and 814 00:50:21,240 --> 00:50:24,400 you'll get an email in your inbox whenever a new one is out, which of course makes sure 815 00:50:24,400 --> 00:50:25,400 life much easier. 816 00:50:25,400 --> 00:50:27,040 So please do do that as well. 817 00:50:27,040 --> 00:50:29,200 Other than that, we'll see you next week. 818 00:50:29,200 --> 00:50:30,200 Cheers, Martin. 819 00:50:30,200 --> 00:50:31,200 Bye. 820 00:50:31,200 --> 00:50:32,200 Bye. 821 00:50:32,200 --> 00:50:37,360 Thank you for listening to Artificially Intelligent Marketing. 822 00:50:37,360 --> 00:50:43,420 To stay on top of the latest trends, tips and tools in the world of marketing AI, be 823 00:50:43,420 --> 00:50:45,160 sure to subscribe. 824 00:50:45,160 --> 00:50:59,160 We look forward to seeing you again next week.