24 April 2024
Vincent Valentine and Matt Rouse Talk about Art.ai and Your Future AI Girlfriend (or Boyfriend) - E255
.jpg)
In this episode, we delve into a fascinating conversation between Matt Raus and Vincent Valentine, exploring the intersection of AI, technology, and the future of creative industries. Vincent sheds light on innovative projects such as Cognitive AI and Art AI, offering valuable insights into the transformative power of AI in shaping the realms of music, movies, and language translation. Our discussion navigates through the latest advancements in AI technology, touching on topics like voice generation, character mapping, and the creation of multilingual content.
Join us as we uncover the potential impact of AI on various creative domains and envision a future where artificial intelligence plays a pivotal role in shaping artistic expressions and communication.
https://art.ai
https://girlfriend.ai/
https://matthewrouse.com/
"Will AI Take My Job?" The book. Available on Amazon.
[00:00:01]
AI Narrator:
Welcome to the Digital Marketing Masters podcast with your host, Matt Raus.
[00:00:07] Matt Rouse:
Hey. Welcome everybody back to Digital Marketing Masters. I'm your host, Matt Rouse. My guest today is Vincent Valentine. How are you doing, Vincent?
[00:00:15] Vincent Valentine:
I'm doing great. How's your day been? It has been kinda crazy. You got a, like, a flash snowstorm on the second day of spring here. But you're in Barcelona. Right? So no snow here. Yeah. Barcelona, Spain. Nice and sunny at the moment. Heading into spring. So, yes, beautiful Barcelona. Have you ever been to Barcelona?
[00:00:34] Matt Rouse:
I have not. I would like to. Well, if you ever come over Hit some of the cafes and stuff there. I hear they have good cafes, some good food, obviously. Nova Scotia, you know, all we got here is is lobsters the size of small dogs and, you know, scallops the size of baseballs. So that part is good also. But, hey, let's talk for a minute about Cognitive AI. Do you want to give us, like, just maybe a quick once over on what you do, what Cognitive AI is?
[00:01:03] Vincent Valentine:
Yeah. I'm Vincent Ballantyne. I'm the CEO of Cognitive AI. We're a holding company for develops Vance AI solutions, basically.
[00:01:13] Matt Rouse:
Nice. I think a good example for somebody who's who's, you know, like, what does that actually mean? A good example of that would be the art dotai. Do you wanna talk about that for a second? Yeah. Sure. We're we've we're developing
[00:01:25] Vincent Valentine:
multiple IPs at the moment, and the first one we're actually building is called arts.ai. Nice short domain. Mhmm. And, basically, it's a social media platform for social media for AI artists. So think of it a bit like Instagram. Right. But it's just focused on AI creation. So the art isn't just for art, it's art artificial, basically. So images, videos, music, anything 3 d, you know, as as AI evolves, now there's gonna be a lot more to to offer on there. Like, we've generated music, sound effects, voices. It's gonna expand as AI grows as well. So the idea is just basically you can create a profile, showcase your work, the people like what they say, they will follow you just like Instagram, and you'll be able to download free of charge as well the images to use on your website, you know, use on print materials. So it's a bit like a shutter stock as well, but only for AI generated content.
And with time, as we grow, we will plan to monetize it so to allow the AI creators to actually earn a living out of this as well. So they can start charging for prompts like micro transactions or even selling the art themselves.
[00:02:41] Matt Rouse:
Right. Is that more like a kind of a creator economy based thing like you have on YouTube where you get a Yeah. It's easy to use kind of thing.
[00:02:50] Vincent Valentine:
Kind of. Yeah. We're we're still working on on what we're gonna do with it, but we kinda see it as like a hybrid. So it's a bit like Instagram mixed with Patreon mixed with Shutterstock. So it's it's kind of like a mix of all 3 of them. We're still very early in the build right now, so we're still deciding which direction we're gonna go, but Like, you're gonna have to see what you're gonna do. Yeah. I've tried that. It
[00:03:16] Matt Rouse:
takes a lot of prompting. It'll it'll cheerfully tell you all the different things you should do, that each one is the best.
[00:03:22] Vincent Valentine:
Exactly. It's, I use AI a lot, and building the business plan was so much easier with GPT.
[00:03:29] Matt Rouse:
Yeah. Okay.
[00:03:30] Vincent Valentine:
It's insane what you can learn from this.
[00:03:33] Matt Rouse:
You know, one of my favorite things to to do with chat JPT is to somebody sends me a document. Right? They're like, oh, here's the agreement between our companies kind of thing, and I just pump it in there and say, tell me 5 things about this I need to know. Get the information I need and get out without having to skim a 30 page insurance document.
[00:03:52] Vincent Valentine:
Exactly. Yeah. Even with contracts, it's great. You can put I've I've done, used contracts in the past where you just throw it in there and say, what is the negative points to this document? And it will just list them for you instead of going through like 40 pages of a document, which you don't have time for. It's just brilliant.
[00:04:08] Matt Rouse:
I don't I've stopped using GPT altogether now. I'm actually using perplexity. I don't know if you've used that before. I have, actually. I kinda find that I like I like perplexity a lot. I built, like, an entire kind of custom instruction set for GPT 4 that seems to make it better at following instructions, and it seems to work well for the uses that I have. If I need large context, you know, I use Claude and, yeah, for other stuff I use perplexity. My daughter really likes pie, but, unfortunately, I don't know how long pie is gonna be around for. They just had a bit of a shuffle. Yeah.
[00:04:41] Vincent Valentine:
That is kind of what inspired us to build Girlfriend was Py, because it's, because it was, is it DeepMind in is involved with Py or one of the developers?
[00:04:53] Matt Rouse:
The lead developer came from DeepMind, but he just went to open the new division from Microsoft. Right. Okay. Bunch of their staff jumped ship and went with them. So Mhmm. That's something. Because,
[00:05:05] Vincent Valentine:
because PIE seems great, but I just don't see It's really good. Did it?
[00:05:10] Matt Rouse:
They just don't have any way to monetize it, I think, is the problem. Mhmm. Like Exactly.
[00:05:15] Vincent Valentine:
It's, it's the voice, when you speak to it as well, there's always a little bit of a delay, I've noticed.
[00:05:21] Matt Rouse:
Yeah. And I think that's pretty much solved. Like, if you look at something like groq, like, groq, the one that's software, not the Elon Musk's b s The LPUs,
[00:05:34] Vincent Valentine:
basically, where it's super quick. Yeah. I've been researching that. Yeah. So fast.
[00:05:39] Matt Rouse:
If you guys have not seen that Grok, look for Grok. It's a you can see a demo of it online. You can go on to their site, and you can use it. It is instantaneous. Like, you ask it a question, and the answer's already there. Like Mhmm. It seems like, you know, like, it's it's not quite to the point where it knows what you're gonna ask before you ask it, but that's probably the only way it would be faster.
[00:06:02] Vincent Valentine:
Yeah. I did a test with it with because you could use Mistral. You can use, Llama 2 on there. And with the demo, you just have to ask it one single question, and it's like paragraph after paragraph in 1 in less than one second. It's, it's pretty remarkable.
[00:06:18] Matt Rouse:
Yeah. I don't know. I mean, I see the the, you know, the idea of how they're doing it. But the other thing that's really interesting is that it's using different hardware, right, than
[00:06:30] Vincent Valentine:
It looks like the the very when I've done my research into them, they're very cryptic of how they're actually doing it. So is it an algorithm they're using just with hardware acceleration? I have no idea, but you're obviously keeping it under wraps.
[00:06:46] Matt Rouse:
Yeah. I think somebody I heard talking about it was saying that there's something about the way that it pipelines the information, and so they were using chips that are specifically for that task, but they didn't say what those chips were. Yeah. But, I mean, they're probably just Intel server, you know, whatever the newest version of the Xeon is or something. Mhmm. I can't remember now. I used to work at Intel. I can't remember. I should probably know that. But, anyway, so you you mentioned girlfriend dotai. Do you wanna explain what that is?
[00:07:18] Vincent Valentine:
Girlfriend dotai is, yeah, very black mirror right now. It's, we're we've bought the domain Girlfriend dotai. Mhmm. And we're building it's safer work for 1. It's no adult content like most, but but if you do research for AI girlfriends, they're very seedy, very adult like. Right. We're totally opposite from this. We're we're we're looking at ethics. We want AI to help the individual. So we're literally creating, like, the, the the best AI girlfriend slash companion for any male that is single trying to bet themselves. So basically we're, we're using, using it to be a companion, but at the same time to push you to improve yourself. Now if you improve your confidence and speaking to women to going out and actually doing things basically.
[00:08:07] Matt Rouse:
And when when you say it makes people better, I mean, is it like I I know that you're under talking about, like, mental state kind of thing, but is that I mean, I guess the question is how are you making them better? Or is it like just encouragement? Is it helping make good decisions? Or
[00:08:27] Vincent Valentine:
we have to create a first thing, we need to get to know the individual because everyone's different. Mhmm. So the onboarding procedure will be very personal and it's only data that we store is very, very private. It's never leaked to anybody. So the first phase, what was what we're currently working on right now is basically getting to know the individual. No asking about them, about literally everything about the person. So it understands them how to speak to you and in terms of like, emotions as well. Like we've found AI tech that we can use where we can detect in your voice of your current state as well. So if you're stressed, we will we should be able to detect that you're stressed, and then try and de stress you with, you know, based based on what we're training the the model with right now. Same way if you're happy to keep you elevated, if you're sad, no, it's, we're using a lot of like therapy methods to to actually help the individual better become a better person, basically.
[00:09:27] Matt Rouse:
Right. How do you keep something like that from I don't know. I guess, like, from going off the rails a bit, you know, because it's
[00:09:36] Vincent Valentine:
LMLs are kinda known for doing it. Yes. We're we're we're we're protecting the LLM with we're we're basically wrapping each request. So we can't. And there's one thing what we can't do is instruct somebody to do something like we can't say, you don't know if the L and M is gonna say, jump off a building and someone dies and then we get sued for it. So we have to be very non direct. We have to basically act like a therapist would do. Right. Where we'll suggest, like, no, 80 percent of men would find a partner if they go to a date night or something like that. We've given suggestions, but it's up to you as the individual to follow that or or ignore it, basically.
[00:10:15] Matt Rouse:
Right. Is it some sort of, like,
[00:10:17] Vincent Valentine:
mixture of experts model or something where you're using an LLM to check the LLM kind of thing? Or Well, we're gonna be using, like, reinforcement training over with the first phase. So, we've got lists of questions that we ask the individual. And we're just basically, we're gonna just see because the first time we we we're gonna launch this, it's not gonna be good. It's we need to train the model. Now every AI developer that I spoke to said, look, don't expect it to be the best straight away. So we're just experimenting with ideas how to, like, refine the model to become better. And then once we get it to a certain level, that's when we will start now releasing it to the public. We just don't know how long it's gonna take. And is this something where you guys are training your own
[00:10:58] Matt Rouse:
specific model? Or is it more like you're using another model and using, like, I don't know, RAG or something to help back up the data? We're gonna be using a lot of experts in this. Like, like, relationship
[00:11:10] Vincent Valentine:
experts know know to basically help us, assist us based on what would you do if you were to, like, relationship therapist. Right. How would you go from a to b? And this is what we're learning a lot with the professionals at the moment. So it's just finding out what what no. How people are and how they react. And there's there's certain ways that you can actually map a person's personality, which I didn't even understand until, like, 2 weeks ago. And there's I think it's called can't remember the exact name of it, but every human has like 16 different types of personality traits. And based on a set of questions, they know exactly what your personality is. And then obviously with that personality, then you can develop on top of it. So if you're an introvert or extrovert or no, there's just so much psychology into this and it's, for me, it's I'm learning something new every single day regarding it.
[00:12:03] Matt Rouse:
Did you ever see that movie Her?
[00:12:05] Vincent Valentine:
I was about to come to I was gonna say that I've got a system gal. I've got an interesting story with us. My business partly, partner, Andy Booth. We started in January, and the first project was Arts. And then I was telling them about her. I've only watched halfway through it, but I got excited from it. So I was telling him, no. Watch it. Watch it. Watch it. So he did eventually watch it, and he just was like, I want to buy the domain right now. So he went away, bought girlfriend. Io and just went, right, Vincent, here's the name.
Let's develop something. And I was like, okay. So we are taking a deep dive into it. Obviously, with the movie, it's a bit negative because
[00:12:43] Matt Rouse:
he falls in love with the the AI operating system. This is not like designed to be a girlfriend. It's just, an AI.
[00:12:51] Vincent Valentine:
Right? Exactly. But technically an AGI. Right. But exactly. Yeah. So, with the movie, she kinda gets to know him. And then I've noticed with certain parts of the movie when he's sad, because he's going through a breakup, she's comfort in him, like sending him messages, sending him pictures, writing songs for him. And this was like 11 years ago, this movie was made. And we currently have this technology today to do exactly the same thing if we wanted to. Obviously, not at that level, but we're getting very, very close to it.
[00:13:19] Matt Rouse:
Yeah. I thought that was really if you guys haven't seen that movie, definitely go check it out. It's called Her. I've seen some things. Like, you see a lot of people using chat gbt kind of out of the box as their, like, therapist, which I would not recommend. I mean, it's yeah. It's quote somebody to talk to, but it's not really a somebody. Right? So then you've got, like, character AI. Mostly people making AIs that seem slightly like celebrities or Yeah. Elon Musk and yeah. I've seen it. Yeah. Yeah. Yeah. Glad we don't need any more of those. I think one's enough for the planet. No. Exactly. Yeah. But The biggest part of this, though, is making the LLM
[00:14:01] Vincent Valentine:
sound realistic. Yeah. So because when you're speaking to an LLM, you'd know you're speaking to a bot, basically. And this is what we're working on is the the human element to it, the responsiveness from the user. And what we're doing with Girlfriend is we're trying to make it feel as real as possible. We want the user to be immersed into the tech, and this is where we're going to find the secret sauce in this. And once we nail this, this is when it's gonna feel like a real person. And I think once you get once you blur the lines between that, this is when it goes very sci fi, basically. Right. Definitely some interesting stuff going on in that space when you look at, like, customer service bots and things like that Mhmm. Where you can interrupt the bot, and it'll, you know, wait and then, like, respond differently
[00:14:44] Matt Rouse:
using systems to kind of really, really cut down on the response time. If you look at something like Sym Theory where you could have, you know, use their bot to call I forgot what the name of the program they're using in it is, but you could have this bot that will go out and call, you know, make you a dinner appointment or reservation or something like that. Mhmm. But it has a big delay, especially on introductions and stuff. Like the person answers the phone, they're like, hello.
[00:15:10] Vincent Valentine:
Hi. I'm Rachel. You know, like it's just too long. The latency is the key with this. And this is one of the biggest thing I spoke to the AI developers about. I said the first thing is latency. We can't be delayed. Now you, if you're speaking to somebody, you need them to actually be processing as you're speaking. And it's,
[00:15:29] Matt Rouse:
we've solved this type of thing, but it does require a lot of compute. That's that's your new trouble. Yeah. Well, you know what? The other thing with that I mean, you guys obviously will know this already, but the idea is, like, if you're building an application right now and there is a lot of latency in it, there's not gonna be a lot of latency 12 months from now. Right? Because the the software will be faster,
[00:15:50] Vincent Valentine:
the chips will be faster, like, making sure that your connection to the cloud and all those kind of things are optimized so that Mhmm. You can cut down on that side of the latency latency because the processors will be able to do it. Yeah. Yeah. The the biggest problem is you're joining loads of dots together. So you got, like, text voice and then processing that and then and then it's sending it to the LLM, grabbing the response back, but then grabbing the response and turning it into a voice. And then also with this automatic voice gen, detection, there's a lot of roadblocks that you have to try and, avoid with this. But look at NVIDIA this week with the new hardware that they've released. Yeah. They say 30 30 times faster.
There's some is it something like 4 trillion parameters it can support? It's over it's it's crazy.
[00:16:39] Matt Rouse:
Yeah. I mean, the models that are gonna come out like your GPT 6 level model is gonna be just off the charts. Right? Mhmm. The other thing that's really interesting is people use a lot of things. Like, you can hear what I'm thinking about what I'm gonna ask you. I'm slowing down my speech to give myself more time to think about it before I ask you the question. Or people will use ums and ahs and and filler words and things like that. Mhmm. You know, those can be worked in to help, you know, with lag time and stuff like that as well. But
[00:17:08] Vincent Valentine:
Exactly. This is the thing. It's like, when you're talking to somebody, you hear mhmm. And then this this could be a simple MP 3 file. Right. So as you're speaking, but the thing is, what you could do to reduce latency even more is that we know the questions we're gonna ask you. So if it's a yes or no, no Boolean request, like, is it yes or no? We can pre render that if you didn't understand. So we know if you say yes, we can instantly give you the response. And this is where we're we're if we ask the right questions in the right, way, we know there's only 2 responses. We can Right. Literally cache the response with this.
[00:17:45] Matt Rouse:
Yeah. And there's also that you know, the idea from video games and stuff where it's it's only caching the, assets it knows it's gonna use in advance. So Yeah. If it knows it's going to ask 1 of 3 questions coming up, you can process all those in advance and then have them kind of waiting in the wings ready to go.
[00:18:05] Vincent Valentine:
Exactly. But they have to be personalized for that person as well because that response will be, no. You'd be they'll be speaking back, no. Hey, Derek or hey, Vincent.
[00:18:14] Matt Rouse:
So it's it still requires a lot of compute to actually do this, but, I'm sure we'll figure it out. Yeah. So do you guys have anything else kinda waiting in the wings there? Are those your kinda two main projects right now?
[00:18:27] Vincent Valentine:
Art AI It's still very early right now. We've we've got SoundAI, which we're in the very early doors right now. So we're planning to create a tool that can create sound effects or create a mute, song lyrics, you name it. But we're still very, very early right now regarding that. Did you see the 11 Labs sound effect generator? Yes. Saw them. Yeah. So
[00:18:50] Matt Rouse:
these guys are on our radar. You guys gotta beat them to it. To be part of so, like, they're they're the top of the game, Rand. I don't know. Yeah. They're ahead of the game. I don't even know what their valuation is right now, but that's a company. As soon as I used 11 Labs for the first time for voice, I was like, this is a game changer. Like, this is absolutely completely different than anything we've seen before. Right? Yeah. I mean, I've used a lot of voice over tech over the years, you know, working in marketing. We have to make ads for people and stuff all the time and videos, and I've used every voice tech under the sun. And I saw that, it was like, you know, it's like magic. Right? You're just like it's like somebody shows you a magic trick that you've never seen before.
Yeah. And same with, like, Sono? Have you seen that Yeah. Music maker. Right, Sono?
[00:19:38] Vincent Valentine:
Yeah. I I think I used them last week. Actually, you I generate the lyrics in GPT, and then it was about my wife, and I put it into there. And then I just checked the prompts and put it in, and it was it's only on version 2. I was using the free model, but it was still very good. But the version 3 is amazing. I don't know if you've said I haven't seen 3 yet. Yeah. It's, under a paywall at the moment, but you can do, like, 2 to 3 minutes
[00:20:02] Matt Rouse:
of, audio, and some of this stuff is just mind blowing. Nice. I might have to get in on that. We use music for ads all the time, And, you know, I usually use stock, but brave new world.
[00:20:12] Vincent Valentine:
Right? But the question is what the as this tech gets bigger and bigger, I mean, the the jobs, that people normally pay for, they're like musicians. Now if you you normally hire a musician to write a jingle for your radio show or even your TV advert, now in a few years, they won't need them anymore.
[00:20:31] Matt Rouse:
In the book, will AI take my job? Yeah. Nice segue. Actually, no. I did I did talk about that in my book a bit. Here's prediction. You guys can hold me to this if you want. And and this is always held true, at least, you know, looking backwards to technology is that the pendulum kinda swings both ways. As it swings towards AI doing more and more of those types of jobs, you won't go look for a stock music of something. You'll just say, generate me something that sounds like x y z. Mhmm. The idea of kind of analog created, human created music, I think is also gonna be kind of the, like, the luxury item of music, you know, versus the commodity of anything that the system could make.
Mhmm. Because everybody says, well, what what what's gonna happen when AI can just make any kind of music? Well, it won't be able to make any kind of music. It'll be able to make lots of kinds of music, but there's still gonna be some things that it will have not thought of. No one will have thought to prompt it. Some combination of instruments and people and things that are just something where you gotta listen to it as a human to say, yeah. Okay. I get it. You know, that's gonna be something that's difficult
[00:21:48] Vincent Valentine:
to rep. I've got a I've got little prediction regarding music. And it's it's like with Michael Jackson, rest in peace. He's not gonna produce any more music for us. But with time, we could sample every single one of his tracks which hit number 1 worldwide. Right. Find an algorithm which why it made it number 1, and then generate a brand new generate 1, the lyrics, and then 2, the music, the hooks, and everything. And then before you know it, it'll be number 1 worldwide because the AI knows what made this song special in the first place. And if you already look at, like, the Beatles that I think it was about 4 months ago, they released a track which was generated with AI because they had to, like, replicate the voices of John Lennon because his samples were so low quality. They used AI to up it as well. So this is just the infancy, but I do reckon there's gonna be a time where you could just say, make me a Michael Jackson track, and then that'll be our playlist basically. I
[00:22:42] Matt Rouse:
completely agree with you, but one thing that makes music chart topping kind of music is timing. Right? Some of that is about, like, Michael if Michael Jackson's number one hit came out 10 years earlier or 10 years later, it wouldn't have been a hit. Right? So it's not the structure of the music that makes it number 1. Like, of course, it has the quality has to be there and everything has to be there And maybe it would still do well, but it wouldn't do as well because it wasn't timed appropriately. It's more of nostalgia, won't it? It will be. Yeah. I think there's a there's a there's a timing to culture when it comes to art and music that is very, very, very specific.
And I don't think it's it's immediately replicable. But I believe that there's also gonna be a big rush of as AI gets broadcast quality that both with TV, movies, music, pretty much anything where people are gonna use it to make stuff that's gonna be popular. You know, one person is gonna make a blockbuster type movie, like a movie that, you know, had a budget of $200 just to pay for the tools, you know, kind of thing, but makes $1,000,000 in the box office. Yeah. Someone is gonna generate a song from a prompt that makes $1,000,000. Right? Like, these things are gonna happen for sure.
[00:24:03] Vincent Valentine:
But Exactly. There's also gonna be a 1,000,000,000 other songs that don't. Mhmm. Exactly. It's just like you said, being at the right place at the right time and the same with Zohra with, OpenAI. It's my my son's currently in college. He's learning how to, like, make movies and directors. Right. And there's, I was explaining to him on the phone a few days ago that no, in the future, when the solar is released, your b roll that you can use is gonna say, you don't to buy any gimbals. You don't have to buy any drones to get aerial footage. It's just a prompt, and you can actually include that as part of your your movie. So I think you will see a higher quality in these, like, student movies. And then eventually with time, like, Hollywood will start using that because why pay for an expensive drone or an expensive crew when you can actually get it yourself, basically?
[00:24:50] Matt Rouse:
Yeah. Well, you could do something like Wonder Studio right now where you could just run around, shoot all the scenes with yourself in it, and then replace yourself with the 3 d model of whatever character that Yeah. The actual character and off you go. And that's something you can do now. And you can even look at something like the character mapping that they're using with the studio quality tools where it's not just remapping the character's face to say the words that they wanted, like changing the words around and and having the character's face redone. They can actually take, like, whatever the emotions that they were expressing and map the emotions to a new scene. So they could have you know, change the character from happy to sad even without reshooting.
Yeah. And, you know, that kind of tech crazy. Is crazy too. Right? As well as all the foreign films now are being redone in English, well, foreign for me. Right? For, you know, US and Canadian audiences and British audiences. All the foreign films or foreign language films are being redone in other languages, but then they're using Yeah. I
[00:25:59] Vincent Valentine:
Yeah. I saw I saw a lot of YouTube is actually show this tech where they can actually change their, you know, their videos into Hindi or Spanish, and it and it looks you tell it's not quite there, but it's still
[00:26:11] Matt Rouse:
pretty pretty impressive right now. So imagine in 6 months or a year's time what the technology is gonna be like. Yeah. I mean, we already transcribe our podcast into 3 languages with GPT. Wow. You know, so the next step is gonna be redoing it, the language and the audio in another language. I don't know if I need to get like a a contract or something for that, you know, like you Vincent, do you agree? Sign here for me to convert your voice into 10 languages or whatever.
[00:26:38] Vincent Valentine:
I don't know, but yeah. Let's let's see. Question
[00:26:41] Matt Rouse:
for the lawyers. Vincent, it's been great talking to you. I know we kinda went a little bit over time. We could probably talk all day about this, but, if somebody wants to reach out to you, what's the best way for them to do it? Best way is to hit the website right now, which is cognitive.ai.
[00:26:55] Vincent Valentine:
And if you wanna be a part of the beta for art.ai, it's the domain is art.ai. And what we're doing is regarding art is that you can preregister. And then once we release the first release, you'd be able to grab your handle. So it could be art.ai/mats. If you get there early, you're gonna get your nice little handle on there. And, yeah. So cognitive.ai, there's links in there to my LinkedIn. So if you wanna connect, you're more than welcome to as well. Perfect. Vincent, I appreciate you being on the show today. Nice to meet you as well, Matt.
[00:27:29] AI Narrator:
This voice over used to be done by a human but now it is synthetic. Oh la la. If you want to know if your job or business is safe from disruption read Matt's new book Will AI Take My Job? Predictions about AI in corporations, small business and the workplace Available now on Amazon. Trust me, it'll be worth it. Remember to tap like, subscribe, and follow to never miss a show.
Welcome to the Digital Marketing Masters podcast with your host, Matt Raus.
[00:00:07] Matt Rouse:
Hey. Welcome everybody back to Digital Marketing Masters. I'm your host, Matt Rouse. My guest today is Vincent Valentine. How are you doing, Vincent?
[00:00:15] Vincent Valentine:
I'm doing great. How's your day been? It has been kinda crazy. You got a, like, a flash snowstorm on the second day of spring here. But you're in Barcelona. Right? So no snow here. Yeah. Barcelona, Spain. Nice and sunny at the moment. Heading into spring. So, yes, beautiful Barcelona. Have you ever been to Barcelona?
[00:00:34] Matt Rouse:
I have not. I would like to. Well, if you ever come over Hit some of the cafes and stuff there. I hear they have good cafes, some good food, obviously. Nova Scotia, you know, all we got here is is lobsters the size of small dogs and, you know, scallops the size of baseballs. So that part is good also. But, hey, let's talk for a minute about Cognitive AI. Do you want to give us, like, just maybe a quick once over on what you do, what Cognitive AI is?
[00:01:03] Vincent Valentine:
Yeah. I'm Vincent Ballantyne. I'm the CEO of Cognitive AI. We're a holding company for develops Vance AI solutions, basically.
[00:01:13] Matt Rouse:
Nice. I think a good example for somebody who's who's, you know, like, what does that actually mean? A good example of that would be the art dotai. Do you wanna talk about that for a second? Yeah. Sure. We're we've we're developing
[00:01:25] Vincent Valentine:
multiple IPs at the moment, and the first one we're actually building is called arts.ai. Nice short domain. Mhmm. And, basically, it's a social media platform for social media for AI artists. So think of it a bit like Instagram. Right. But it's just focused on AI creation. So the art isn't just for art, it's art artificial, basically. So images, videos, music, anything 3 d, you know, as as AI evolves, now there's gonna be a lot more to to offer on there. Like, we've generated music, sound effects, voices. It's gonna expand as AI grows as well. So the idea is just basically you can create a profile, showcase your work, the people like what they say, they will follow you just like Instagram, and you'll be able to download free of charge as well the images to use on your website, you know, use on print materials. So it's a bit like a shutter stock as well, but only for AI generated content.
And with time, as we grow, we will plan to monetize it so to allow the AI creators to actually earn a living out of this as well. So they can start charging for prompts like micro transactions or even selling the art themselves.
[00:02:41] Matt Rouse:
Right. Is that more like a kind of a creator economy based thing like you have on YouTube where you get a Yeah. It's easy to use kind of thing.
[00:02:50] Vincent Valentine:
Kind of. Yeah. We're we're still working on on what we're gonna do with it, but we kinda see it as like a hybrid. So it's a bit like Instagram mixed with Patreon mixed with Shutterstock. So it's it's kind of like a mix of all 3 of them. We're still very early in the build right now, so we're still deciding which direction we're gonna go, but Like, you're gonna have to see what you're gonna do. Yeah. I've tried that. It
[00:03:16] Matt Rouse:
takes a lot of prompting. It'll it'll cheerfully tell you all the different things you should do, that each one is the best.
[00:03:22] Vincent Valentine:
Exactly. It's, I use AI a lot, and building the business plan was so much easier with GPT.
[00:03:29] Matt Rouse:
Yeah. Okay.
[00:03:30] Vincent Valentine:
It's insane what you can learn from this.
[00:03:33] Matt Rouse:
You know, one of my favorite things to to do with chat JPT is to somebody sends me a document. Right? They're like, oh, here's the agreement between our companies kind of thing, and I just pump it in there and say, tell me 5 things about this I need to know. Get the information I need and get out without having to skim a 30 page insurance document.
[00:03:52] Vincent Valentine:
Exactly. Yeah. Even with contracts, it's great. You can put I've I've done, used contracts in the past where you just throw it in there and say, what is the negative points to this document? And it will just list them for you instead of going through like 40 pages of a document, which you don't have time for. It's just brilliant.
[00:04:08] Matt Rouse:
I don't I've stopped using GPT altogether now. I'm actually using perplexity. I don't know if you've used that before. I have, actually. I kinda find that I like I like perplexity a lot. I built, like, an entire kind of custom instruction set for GPT 4 that seems to make it better at following instructions, and it seems to work well for the uses that I have. If I need large context, you know, I use Claude and, yeah, for other stuff I use perplexity. My daughter really likes pie, but, unfortunately, I don't know how long pie is gonna be around for. They just had a bit of a shuffle. Yeah.
[00:04:41] Vincent Valentine:
That is kind of what inspired us to build Girlfriend was Py, because it's, because it was, is it DeepMind in is involved with Py or one of the developers?
[00:04:53] Matt Rouse:
The lead developer came from DeepMind, but he just went to open the new division from Microsoft. Right. Okay. Bunch of their staff jumped ship and went with them. So Mhmm. That's something. Because,
[00:05:05] Vincent Valentine:
because PIE seems great, but I just don't see It's really good. Did it?
[00:05:10] Matt Rouse:
They just don't have any way to monetize it, I think, is the problem. Mhmm. Like Exactly.
[00:05:15] Vincent Valentine:
It's, it's the voice, when you speak to it as well, there's always a little bit of a delay, I've noticed.
[00:05:21] Matt Rouse:
Yeah. And I think that's pretty much solved. Like, if you look at something like groq, like, groq, the one that's software, not the Elon Musk's b s The LPUs,
[00:05:34] Vincent Valentine:
basically, where it's super quick. Yeah. I've been researching that. Yeah. So fast.
[00:05:39] Matt Rouse:
If you guys have not seen that Grok, look for Grok. It's a you can see a demo of it online. You can go on to their site, and you can use it. It is instantaneous. Like, you ask it a question, and the answer's already there. Like Mhmm. It seems like, you know, like, it's it's not quite to the point where it knows what you're gonna ask before you ask it, but that's probably the only way it would be faster.
[00:06:02] Vincent Valentine:
Yeah. I did a test with it with because you could use Mistral. You can use, Llama 2 on there. And with the demo, you just have to ask it one single question, and it's like paragraph after paragraph in 1 in less than one second. It's, it's pretty remarkable.
[00:06:18] Matt Rouse:
Yeah. I don't know. I mean, I see the the, you know, the idea of how they're doing it. But the other thing that's really interesting is that it's using different hardware, right, than
[00:06:30] Vincent Valentine:
It looks like the the very when I've done my research into them, they're very cryptic of how they're actually doing it. So is it an algorithm they're using just with hardware acceleration? I have no idea, but you're obviously keeping it under wraps.
[00:06:46] Matt Rouse:
Yeah. I think somebody I heard talking about it was saying that there's something about the way that it pipelines the information, and so they were using chips that are specifically for that task, but they didn't say what those chips were. Yeah. But, I mean, they're probably just Intel server, you know, whatever the newest version of the Xeon is or something. Mhmm. I can't remember now. I used to work at Intel. I can't remember. I should probably know that. But, anyway, so you you mentioned girlfriend dotai. Do you wanna explain what that is?
[00:07:18] Vincent Valentine:
Girlfriend dotai is, yeah, very black mirror right now. It's, we're we've bought the domain Girlfriend dotai. Mhmm. And we're building it's safer work for 1. It's no adult content like most, but but if you do research for AI girlfriends, they're very seedy, very adult like. Right. We're totally opposite from this. We're we're we're looking at ethics. We want AI to help the individual. So we're literally creating, like, the, the the best AI girlfriend slash companion for any male that is single trying to bet themselves. So basically we're, we're using, using it to be a companion, but at the same time to push you to improve yourself. Now if you improve your confidence and speaking to women to going out and actually doing things basically.
[00:08:07] Matt Rouse:
And when when you say it makes people better, I mean, is it like I I know that you're under talking about, like, mental state kind of thing, but is that I mean, I guess the question is how are you making them better? Or is it like just encouragement? Is it helping make good decisions? Or
[00:08:27] Vincent Valentine:
we have to create a first thing, we need to get to know the individual because everyone's different. Mhmm. So the onboarding procedure will be very personal and it's only data that we store is very, very private. It's never leaked to anybody. So the first phase, what was what we're currently working on right now is basically getting to know the individual. No asking about them, about literally everything about the person. So it understands them how to speak to you and in terms of like, emotions as well. Like we've found AI tech that we can use where we can detect in your voice of your current state as well. So if you're stressed, we will we should be able to detect that you're stressed, and then try and de stress you with, you know, based based on what we're training the the model with right now. Same way if you're happy to keep you elevated, if you're sad, no, it's, we're using a lot of like therapy methods to to actually help the individual better become a better person, basically.
[00:09:27] Matt Rouse:
Right. How do you keep something like that from I don't know. I guess, like, from going off the rails a bit, you know, because it's
[00:09:36] Vincent Valentine:
LMLs are kinda known for doing it. Yes. We're we're we're we're protecting the LLM with we're we're basically wrapping each request. So we can't. And there's one thing what we can't do is instruct somebody to do something like we can't say, you don't know if the L and M is gonna say, jump off a building and someone dies and then we get sued for it. So we have to be very non direct. We have to basically act like a therapist would do. Right. Where we'll suggest, like, no, 80 percent of men would find a partner if they go to a date night or something like that. We've given suggestions, but it's up to you as the individual to follow that or or ignore it, basically.
[00:10:15] Matt Rouse:
Right. Is it some sort of, like,
[00:10:17] Vincent Valentine:
mixture of experts model or something where you're using an LLM to check the LLM kind of thing? Or Well, we're gonna be using, like, reinforcement training over with the first phase. So, we've got lists of questions that we ask the individual. And we're just basically, we're gonna just see because the first time we we we're gonna launch this, it's not gonna be good. It's we need to train the model. Now every AI developer that I spoke to said, look, don't expect it to be the best straight away. So we're just experimenting with ideas how to, like, refine the model to become better. And then once we get it to a certain level, that's when we will start now releasing it to the public. We just don't know how long it's gonna take. And is this something where you guys are training your own
[00:10:58] Matt Rouse:
specific model? Or is it more like you're using another model and using, like, I don't know, RAG or something to help back up the data? We're gonna be using a lot of experts in this. Like, like, relationship
[00:11:10] Vincent Valentine:
experts know know to basically help us, assist us based on what would you do if you were to, like, relationship therapist. Right. How would you go from a to b? And this is what we're learning a lot with the professionals at the moment. So it's just finding out what what no. How people are and how they react. And there's there's certain ways that you can actually map a person's personality, which I didn't even understand until, like, 2 weeks ago. And there's I think it's called can't remember the exact name of it, but every human has like 16 different types of personality traits. And based on a set of questions, they know exactly what your personality is. And then obviously with that personality, then you can develop on top of it. So if you're an introvert or extrovert or no, there's just so much psychology into this and it's, for me, it's I'm learning something new every single day regarding it.
[00:12:03] Matt Rouse:
Did you ever see that movie Her?
[00:12:05] Vincent Valentine:
I was about to come to I was gonna say that I've got a system gal. I've got an interesting story with us. My business partly, partner, Andy Booth. We started in January, and the first project was Arts. And then I was telling them about her. I've only watched halfway through it, but I got excited from it. So I was telling him, no. Watch it. Watch it. Watch it. So he did eventually watch it, and he just was like, I want to buy the domain right now. So he went away, bought girlfriend. Io and just went, right, Vincent, here's the name.
Let's develop something. And I was like, okay. So we are taking a deep dive into it. Obviously, with the movie, it's a bit negative because
[00:12:43] Matt Rouse:
he falls in love with the the AI operating system. This is not like designed to be a girlfriend. It's just, an AI.
[00:12:51] Vincent Valentine:
Right? Exactly. But technically an AGI. Right. But exactly. Yeah. So, with the movie, she kinda gets to know him. And then I've noticed with certain parts of the movie when he's sad, because he's going through a breakup, she's comfort in him, like sending him messages, sending him pictures, writing songs for him. And this was like 11 years ago, this movie was made. And we currently have this technology today to do exactly the same thing if we wanted to. Obviously, not at that level, but we're getting very, very close to it.
[00:13:19] Matt Rouse:
Yeah. I thought that was really if you guys haven't seen that movie, definitely go check it out. It's called Her. I've seen some things. Like, you see a lot of people using chat gbt kind of out of the box as their, like, therapist, which I would not recommend. I mean, it's yeah. It's quote somebody to talk to, but it's not really a somebody. Right? So then you've got, like, character AI. Mostly people making AIs that seem slightly like celebrities or Yeah. Elon Musk and yeah. I've seen it. Yeah. Yeah. Yeah. Glad we don't need any more of those. I think one's enough for the planet. No. Exactly. Yeah. But The biggest part of this, though, is making the LLM
[00:14:01] Vincent Valentine:
sound realistic. Yeah. So because when you're speaking to an LLM, you'd know you're speaking to a bot, basically. And this is what we're working on is the the human element to it, the responsiveness from the user. And what we're doing with Girlfriend is we're trying to make it feel as real as possible. We want the user to be immersed into the tech, and this is where we're going to find the secret sauce in this. And once we nail this, this is when it's gonna feel like a real person. And I think once you get once you blur the lines between that, this is when it goes very sci fi, basically. Right. Definitely some interesting stuff going on in that space when you look at, like, customer service bots and things like that Mhmm. Where you can interrupt the bot, and it'll, you know, wait and then, like, respond differently
[00:14:44] Matt Rouse:
using systems to kind of really, really cut down on the response time. If you look at something like Sym Theory where you could have, you know, use their bot to call I forgot what the name of the program they're using in it is, but you could have this bot that will go out and call, you know, make you a dinner appointment or reservation or something like that. Mhmm. But it has a big delay, especially on introductions and stuff. Like the person answers the phone, they're like, hello.
[00:15:10] Vincent Valentine:
Hi. I'm Rachel. You know, like it's just too long. The latency is the key with this. And this is one of the biggest thing I spoke to the AI developers about. I said the first thing is latency. We can't be delayed. Now you, if you're speaking to somebody, you need them to actually be processing as you're speaking. And it's,
[00:15:29] Matt Rouse:
we've solved this type of thing, but it does require a lot of compute. That's that's your new trouble. Yeah. Well, you know what? The other thing with that I mean, you guys obviously will know this already, but the idea is, like, if you're building an application right now and there is a lot of latency in it, there's not gonna be a lot of latency 12 months from now. Right? Because the the software will be faster,
[00:15:50] Vincent Valentine:
the chips will be faster, like, making sure that your connection to the cloud and all those kind of things are optimized so that Mhmm. You can cut down on that side of the latency latency because the processors will be able to do it. Yeah. Yeah. The the biggest problem is you're joining loads of dots together. So you got, like, text voice and then processing that and then and then it's sending it to the LLM, grabbing the response back, but then grabbing the response and turning it into a voice. And then also with this automatic voice gen, detection, there's a lot of roadblocks that you have to try and, avoid with this. But look at NVIDIA this week with the new hardware that they've released. Yeah. They say 30 30 times faster.
There's some is it something like 4 trillion parameters it can support? It's over it's it's crazy.
[00:16:39] Matt Rouse:
Yeah. I mean, the models that are gonna come out like your GPT 6 level model is gonna be just off the charts. Right? Mhmm. The other thing that's really interesting is people use a lot of things. Like, you can hear what I'm thinking about what I'm gonna ask you. I'm slowing down my speech to give myself more time to think about it before I ask you the question. Or people will use ums and ahs and and filler words and things like that. Mhmm. You know, those can be worked in to help, you know, with lag time and stuff like that as well. But
[00:17:08] Vincent Valentine:
Exactly. This is the thing. It's like, when you're talking to somebody, you hear mhmm. And then this this could be a simple MP 3 file. Right. So as you're speaking, but the thing is, what you could do to reduce latency even more is that we know the questions we're gonna ask you. So if it's a yes or no, no Boolean request, like, is it yes or no? We can pre render that if you didn't understand. So we know if you say yes, we can instantly give you the response. And this is where we're we're if we ask the right questions in the right, way, we know there's only 2 responses. We can Right. Literally cache the response with this.
[00:17:45] Matt Rouse:
Yeah. And there's also that you know, the idea from video games and stuff where it's it's only caching the, assets it knows it's gonna use in advance. So Yeah. If it knows it's going to ask 1 of 3 questions coming up, you can process all those in advance and then have them kind of waiting in the wings ready to go.
[00:18:05] Vincent Valentine:
Exactly. But they have to be personalized for that person as well because that response will be, no. You'd be they'll be speaking back, no. Hey, Derek or hey, Vincent.
[00:18:14] Matt Rouse:
So it's it still requires a lot of compute to actually do this, but, I'm sure we'll figure it out. Yeah. So do you guys have anything else kinda waiting in the wings there? Are those your kinda two main projects right now?
[00:18:27] Vincent Valentine:
Art AI It's still very early right now. We've we've got SoundAI, which we're in the very early doors right now. So we're planning to create a tool that can create sound effects or create a mute, song lyrics, you name it. But we're still very, very early right now regarding that. Did you see the 11 Labs sound effect generator? Yes. Saw them. Yeah. So
[00:18:50] Matt Rouse:
these guys are on our radar. You guys gotta beat them to it. To be part of so, like, they're they're the top of the game, Rand. I don't know. Yeah. They're ahead of the game. I don't even know what their valuation is right now, but that's a company. As soon as I used 11 Labs for the first time for voice, I was like, this is a game changer. Like, this is absolutely completely different than anything we've seen before. Right? Yeah. I mean, I've used a lot of voice over tech over the years, you know, working in marketing. We have to make ads for people and stuff all the time and videos, and I've used every voice tech under the sun. And I saw that, it was like, you know, it's like magic. Right? You're just like it's like somebody shows you a magic trick that you've never seen before.
Yeah. And same with, like, Sono? Have you seen that Yeah. Music maker. Right, Sono?
[00:19:38] Vincent Valentine:
Yeah. I I think I used them last week. Actually, you I generate the lyrics in GPT, and then it was about my wife, and I put it into there. And then I just checked the prompts and put it in, and it was it's only on version 2. I was using the free model, but it was still very good. But the version 3 is amazing. I don't know if you've said I haven't seen 3 yet. Yeah. It's, under a paywall at the moment, but you can do, like, 2 to 3 minutes
[00:20:02] Matt Rouse:
of, audio, and some of this stuff is just mind blowing. Nice. I might have to get in on that. We use music for ads all the time, And, you know, I usually use stock, but brave new world.
[00:20:12] Vincent Valentine:
Right? But the question is what the as this tech gets bigger and bigger, I mean, the the jobs, that people normally pay for, they're like musicians. Now if you you normally hire a musician to write a jingle for your radio show or even your TV advert, now in a few years, they won't need them anymore.
[00:20:31] Matt Rouse:
In the book, will AI take my job? Yeah. Nice segue. Actually, no. I did I did talk about that in my book a bit. Here's prediction. You guys can hold me to this if you want. And and this is always held true, at least, you know, looking backwards to technology is that the pendulum kinda swings both ways. As it swings towards AI doing more and more of those types of jobs, you won't go look for a stock music of something. You'll just say, generate me something that sounds like x y z. Mhmm. The idea of kind of analog created, human created music, I think is also gonna be kind of the, like, the luxury item of music, you know, versus the commodity of anything that the system could make.
Mhmm. Because everybody says, well, what what what's gonna happen when AI can just make any kind of music? Well, it won't be able to make any kind of music. It'll be able to make lots of kinds of music, but there's still gonna be some things that it will have not thought of. No one will have thought to prompt it. Some combination of instruments and people and things that are just something where you gotta listen to it as a human to say, yeah. Okay. I get it. You know, that's gonna be something that's difficult
[00:21:48] Vincent Valentine:
to rep. I've got a I've got little prediction regarding music. And it's it's like with Michael Jackson, rest in peace. He's not gonna produce any more music for us. But with time, we could sample every single one of his tracks which hit number 1 worldwide. Right. Find an algorithm which why it made it number 1, and then generate a brand new generate 1, the lyrics, and then 2, the music, the hooks, and everything. And then before you know it, it'll be number 1 worldwide because the AI knows what made this song special in the first place. And if you already look at, like, the Beatles that I think it was about 4 months ago, they released a track which was generated with AI because they had to, like, replicate the voices of John Lennon because his samples were so low quality. They used AI to up it as well. So this is just the infancy, but I do reckon there's gonna be a time where you could just say, make me a Michael Jackson track, and then that'll be our playlist basically. I
[00:22:42] Matt Rouse:
completely agree with you, but one thing that makes music chart topping kind of music is timing. Right? Some of that is about, like, Michael if Michael Jackson's number one hit came out 10 years earlier or 10 years later, it wouldn't have been a hit. Right? So it's not the structure of the music that makes it number 1. Like, of course, it has the quality has to be there and everything has to be there And maybe it would still do well, but it wouldn't do as well because it wasn't timed appropriately. It's more of nostalgia, won't it? It will be. Yeah. I think there's a there's a there's a timing to culture when it comes to art and music that is very, very, very specific.
And I don't think it's it's immediately replicable. But I believe that there's also gonna be a big rush of as AI gets broadcast quality that both with TV, movies, music, pretty much anything where people are gonna use it to make stuff that's gonna be popular. You know, one person is gonna make a blockbuster type movie, like a movie that, you know, had a budget of $200 just to pay for the tools, you know, kind of thing, but makes $1,000,000 in the box office. Yeah. Someone is gonna generate a song from a prompt that makes $1,000,000. Right? Like, these things are gonna happen for sure.
[00:24:03] Vincent Valentine:
But Exactly. There's also gonna be a 1,000,000,000 other songs that don't. Mhmm. Exactly. It's just like you said, being at the right place at the right time and the same with Zohra with, OpenAI. It's my my son's currently in college. He's learning how to, like, make movies and directors. Right. And there's, I was explaining to him on the phone a few days ago that no, in the future, when the solar is released, your b roll that you can use is gonna say, you don't to buy any gimbals. You don't have to buy any drones to get aerial footage. It's just a prompt, and you can actually include that as part of your your movie. So I think you will see a higher quality in these, like, student movies. And then eventually with time, like, Hollywood will start using that because why pay for an expensive drone or an expensive crew when you can actually get it yourself, basically?
[00:24:50] Matt Rouse:
Yeah. Well, you could do something like Wonder Studio right now where you could just run around, shoot all the scenes with yourself in it, and then replace yourself with the 3 d model of whatever character that Yeah. The actual character and off you go. And that's something you can do now. And you can even look at something like the character mapping that they're using with the studio quality tools where it's not just remapping the character's face to say the words that they wanted, like changing the words around and and having the character's face redone. They can actually take, like, whatever the emotions that they were expressing and map the emotions to a new scene. So they could have you know, change the character from happy to sad even without reshooting.
Yeah. And, you know, that kind of tech crazy. Is crazy too. Right? As well as all the foreign films now are being redone in English, well, foreign for me. Right? For, you know, US and Canadian audiences and British audiences. All the foreign films or foreign language films are being redone in other languages, but then they're using Yeah. I
[00:25:59] Vincent Valentine:
Yeah. I saw I saw a lot of YouTube is actually show this tech where they can actually change their, you know, their videos into Hindi or Spanish, and it and it looks you tell it's not quite there, but it's still
[00:26:11] Matt Rouse:
pretty pretty impressive right now. So imagine in 6 months or a year's time what the technology is gonna be like. Yeah. I mean, we already transcribe our podcast into 3 languages with GPT. Wow. You know, so the next step is gonna be redoing it, the language and the audio in another language. I don't know if I need to get like a a contract or something for that, you know, like you Vincent, do you agree? Sign here for me to convert your voice into 10 languages or whatever.
[00:26:38] Vincent Valentine:
I don't know, but yeah. Let's let's see. Question
[00:26:41] Matt Rouse:
for the lawyers. Vincent, it's been great talking to you. I know we kinda went a little bit over time. We could probably talk all day about this, but, if somebody wants to reach out to you, what's the best way for them to do it? Best way is to hit the website right now, which is cognitive.ai.
[00:26:55] Vincent Valentine:
And if you wanna be a part of the beta for art.ai, it's the domain is art.ai. And what we're doing is regarding art is that you can preregister. And then once we release the first release, you'd be able to grab your handle. So it could be art.ai/mats. If you get there early, you're gonna get your nice little handle on there. And, yeah. So cognitive.ai, there's links in there to my LinkedIn. So if you wanna connect, you're more than welcome to as well. Perfect. Vincent, I appreciate you being on the show today. Nice to meet you as well, Matt.
[00:27:29] AI Narrator:
This voice over used to be done by a human but now it is synthetic. Oh la la. If you want to know if your job or business is safe from disruption read Matt's new book Will AI Take My Job? Predictions about AI in corporations, small business and the workplace Available now on Amazon. Trust me, it'll be worth it. Remember to tap like, subscribe, and follow to never miss a show.
Introduction and discussion about Barcelona and Nova Scotia.
Overview of Cognitive AI and its applications.
Exploration of Girlfriend.ai and its ethical approach to AI companionship.
Comparison to the movie 'Her' and the development of AI companions.
Discussion on the impact of AI on creative industries like music and movies.
