Episode 80
AI Disrupting the Arts
About this episode
Welcome to the latest episode of the Untangling Web 3 podcast! Today, we delve into the fascinating intersection of AI and the creative arts, including Hollywood, music, and beyond.
Artificial Intelligence has made significant inroads into these industries, creating tools that both empower creators and spark concerns about job displacement and authenticity.
From generating lifelike voices to creating immersive film sequences, AI is changing how art is made and consumed. Join us as we explore the transformative impact of AI on the creative world.
Key Highlights
- AI in Film and Animation. The episode explores how AI is revolutionizing filmmaking and animation with tools like Runway and Synthesia. These technologies allow creators to produce hyper-realistic video content and animations in a fraction of the time and cost. However, the discussion also highlights ethical concerns, such as the potential misuse of AI-generated actors and the implications for authenticity and copyright.
- AI in Music Production. We discuss AI's ability to generate custom music tailored to user preferences, from Snoop Dogg-inspired tracks to electronic beats. Examples like Grimes’ open-sourced AI voice model and Blau’s AI-generated music demonstrate how AI can help artists expand their creative capabilities. But as AI blurs the lines between original and derivative work, questions arise about ownership and royalties in the music industry.
- The Future of AI in Creative Arts. The conversation extends to the integration of AI in visual arts, from creating viral digital images to producing full-length films. AI is lowering the barrier to entry for aspiring creators by providing tools that enhance productivity and creativity. Yet, its potential for misuse—such as generating deepfakes or unauthorized representations—raises critical issues about trust and regulation in the arts.
AI is reshaping the creative industries by offering unprecedented tools for expression and innovation. While these advancements democratize access to artistic creation and streamline workflows, they also challenge traditional notions of originality, authenticity, and copyright.
As AI continues to evolve, balancing its benefits with ethical considerations will be key to ensuring its positive impact on the arts.
Transcript
Auto-generated from the episode audio.
welcome to another episode of the untangling web 3 podcast if you hadn't heard AI is actually booming right now and surprise surprise that's what we're going to be talking about today if you'd asked someone five to 10 years ago they probably wouldn't have been able to tell you that AI would have been so impactful on the Arts specifically and that's what we're going to talk about is how impactful it's been for the creative industry both in a negative way replacing jobs but also in a positive way lowering the barrier to entry so it's not all doom and gloom and there is a lot of things to be hopeful for so yeah let's kick this one off let's see how AI is impacting the Arts today this is something that is close to a lot of people's hearts and I think when people you know everyone's kind of slightly worried in the back of their minds at least about AI coming for our jobs right and I think nowhere is that more prevalent than in the
creative Industries where we've seen already lots of use of AI which we'll kind of get into some specific examples but that's an area where you know from the early days of the AI boom you know you see mid Journey creating images things like open AI being able to generate techs you know write scripts for you for example so yeah I think there's lots of ways in which this Tech can be used for good and for ill for the creative industry so yeah why don't we start as you said with one that's a little bit closer to home and look at how it's uh affecting the podcast industry in particular yeah I think like you're you're you're right like my background is in robotics and I think 30 to 40 years ago people were having a similar kind of conversation to what were having around AI right now is oh my God this is going to disrupt you know manufacturing lab about all this kind of stuff how are people going to transition and that was
really like the main kind of thought process around the world was that it was going to be the the white collor jobs no wrong one it was going to be the blue collor jobs that were going to be most disrupted by automation right and it's only been the last you know few years now with the kind of the Advent of chat GPT and people understanding how disruptive and how easy to use AI can be that this is really started to come for white color jobs started to come for you know the what we' consider professional jobs um and like the last Bastion in a lot of people's minds that would be safe from automation AI all these things was the Arts that was kind of like you know when you kind of think of this utopian future you'd say well humans you know we're not so good at the manufacturing anymore robots can do that we're not so good at the you know the kind of the the white color jobs you know create doing
accounting creating scripts all this kind of stuff we're going to be good at the Arts that's what we're all going to do that's our last Endeavor we just do podcasts or write scripts and all this kind of generate you know mosaic but now we're starting to see that AI can actually do that as well and it's actually starting to replace that field so it's becoming more and more interesting and like the space that we are better at AI at is becoming smaller and smaller and you know like Jack said we're going to be talking about bod podcast this is M and Jack's artistic Endeavor and now we found that there's an AI replacement that actually does it much better than we can that's quite concerning so let's jump into notebook llm yeah exactly this is something that definitely scared me a little bit when I was uh when I when I first came across it so this is from Google right the notebook llm and I think this has been kind of doing the rounds
recently has gone very viral um is essentially from what I can see is it lets you generate a podcast from any kind of input media so like a text document for example and it creates a full audio podcast with speakers right so AI generated speakers it kind of puts together a few different things right it's the generative AI for you know text to X text to something else um models and then it it has these really super realistic AI uh hosts right of the podcast that can can make virtually anything interesting so this is this is why it's a little bit close to our hearts is because you know 18 months ago we started this podcast we think you know if anyone can make web three interesting maybe it'll be us um and then actually come along these AI voices that you can toggle and change and things they actually if anything doing a little bit of a better job in in some cases so yeah it's a little bit scary right yeah I mean
I found out about this in like some of the people I work with know about my podcast they sometimes listen very few of them maybe and someone just like atted me in the general conversation saying Al like have you seen this your podcast's over and I was like what the hell is he talking about and I went on to it I was like no way played around with it for a little bit and was like this is incredible like scary obviously I think the big thing is there's been these kind of VoiceOver Styles you put into a script and they read them out you know think of Siri sty and all this kind of stuff for for quite a while but the kind of the the nuance around being able to make something make something conversational like they're putting in so an example I saw that I put in was you put in the Bitcoin white paper right very techy very heavy not many people are reading from top to bottom apart from Jack who's
got it on his wall um and they turn it into a very liart conversation that's very interesting there's lots of jokes there lots of gags like they kind of they just nail it down it's so enjoyable that you actually really can't tell the difference between that and like a top tier podcast so as soon as I did this I was like what the hell this is our job's over I thought this would be our last Endeavor but should I play a little bit from it yeah go on give us give at home A Taste of how how much better this be the one and only yeah the document that started it all right pretty much I mean everybody knows about Bitcoin now but I'm curious how many people have actually gone back and read the original white paper written by the mysterious Nakamoto so that's what we're gonna do today we're gonna unpack the core ideas directly from the source figure out you know what problem Bitcoin was trying to solve and see how this
whole thing is supposed to work and don't worry we're going to break down all the technical stuff so you can really grasp the Brilliance of this thing yeah it's uh it's really incredible how this paper it's not even that long but it really changed how we think about money about tech it really did so the white paper jumps right in and it starts by laying out a problem that probably feels familiar to a lot of people which is that you know online transactions always rely on a middleman you know Banks payment processors those guys right always damn like it's so good isn't it yeah I mean we used to joke on on the podcast especially in the early days that I was a bit robotic and that I was a bit like an AI you know this is back when chat GPT first came out and and was only you know gpt3 and things like that I'm used to joke that it's I could sometimes you know if I wasn't on the ball I would
look come across a little bit like an I model and not have the the social cues and things and now wow I mean they're doing arguably a better job even when I'm trying my hardest right so it's and it's one thing that you definitely note there is that it's it's scaled that uncanny valley so if anyone hasn't heard of that effect it's when uncanny valley is you know when something is sufficiently close to human but it's not quite good enough that it gives you this strange effect right it feels wrong and to me now these have got so Advanced that I don't have that when I listen it does just sound genuinely like you're listening to two highly experienced skilled podcasters yeah and and the thing there is like the difference between what's been inputed which is a very technical blunt to the point paper very like purely academic paper and what they're producing there I even said there like what the AI algorithm sorry is producing there is just miles apart like the fact
that they're breaking it down making it easy to understand bringing you in there like I thought this was a very kind of intuitive human only skill to do that to kind of connect with people and to create empathy whatever it is but clearly they've done that I put the Bitcoin white paper in nothing else and 30 seconds later they have a 15 minute podcast that goes into it and there's several jokes it's funny it's engaging and most importantly it's quite scary because this might be the end of the podcast but we've kind of said that we talk about it often that podcasts are the new goto medium new goto media because they're much more engaging people want to connect and they feel like podcasts are a way to do that whether they connect with the individual or well they hope they're connected with us anyway and also learn something but you know if that's going to be automated what's the what's the point I'm joking yeah and and something I find interesting about notebook LM
specifically is that when we'll talk about some of the other ways that that geni is affecting creative Industries I think in lots of those cases there's almost a direct competition or replacement competition for people doing these things I think it's slightly different with notebook llm because I think it's a completely different model where PE where people will just be generating their own media for themselves right it's like it's not compete it's not like there's a new podcaster competing with us and trying to gain market share it's almost reducing the the market size overall because people can just go and generate a podcast about something they're interested in so it's a slightly different way of looking at it which is again interesting just because of that inversion right the idea of people self curating content and being able to you know find out things for themselves much the way people use chat GPT now um rather than having podcasters like you and I Alec deciding you know what do we think is interesting and then you
that might not always be interesting for everyone episode to episode so yeah it's interesting how that's going to change the industry I think yeah and you know you kind of you talked there about our opinion putting our subjectivity and like our personality into things whatever small personality we have and I saw that a few AI a I experts that are saying that one of the last things one of the last things that you know Humanity will be able to latch on to that won't be you know AI won't be able to take over very easily was taste so it's still that the ability to have an opinion that maybe is not the mainstream they're slightly out there slightly different and that's one of the last things that kind of will Define us as human so maybe a you know slightly weird style is a taste in some sense of the word and we can hold on to that but it's not all Doom and Gloom for podcasters like that you said that that might you
know take up some of the the small market share that we're after but there are a lot of tools out there that actually do help podcasters so one of the ones that we use or Jack uses quite regular is descript so this is like a kind of a Production Tool basically where it can like autof filter out pauses gaps you know we have a hell of a lot of issues with our podcast it wasn't for the Jack and D script I think it would be 10 times worse yeah that it's it's really interesting so when we first started the podcast you know a little bit of behind the scenes scenes here but I did all the editing very manually doing all the cuts manually and things like that and then stumbled across descript as an example of of one of these platforms for editing and it uses AI in so many different ways whether it's from you know just the pure transcribing so going from uh video input that we're giving it to to text
and it lets you edit the text directly which is kind of amazing and a lifesaver but then I really I noticed there were some really powerful features that I haven't had to use all that often but you know things like replacing audio so when one of our Wi-Fi cuts out or something and you miss a word you can actually uh have one of us say that word because you know we don't want to go back and re reccord the whole episode right just to get this one word and what it does is it builds an AI model based on our voices right so once or twice in the history of the podcast you will have heard one of us say a word that actually was based on an AI model of us which is amazing and you would never notice you know I mean sometimes okay it takes a couple of generations to get it right but normally it's not too hard to find Once yeah that kind of thing is impressive you've also got
like the ey tracking stuff okay we don't actually use that feature but it's it's kind of spooky to see how good it is to make sure you're actually you know looking towards the camera and not kind of looking around the room like sometimes we do but yeah so powerful we have to have a thing where it's like tell the difference between you real you speaking and you AI you speaking and see like a little Capt style model see if people can tell the difference web three test it's funny this stuff but like it kind of gets into the realm which is obviously some that we're constantly talking about we talk on the topic of AI is authenticity originality and you know copyright and all this kind of stuff if it can duplicate our voices in such a way that you can't tell the difference like you know when I call Sant andere their new thing on on the banking is like you know you saying like a sentence and that proves who you are our
voes everywhere like obviously that would be the easiest thing to spoof in the world and like the facial stuff is getting more and more out you saying the ey placement you can start to put avatars on there people might have seen with the Apple vision Pro and stuff like it's going to be really tricky to start to distinguish people in like the digital space I'm not when you know we are we start as a web web 3 project we are a web 3 project and I think more and more this is kind of getting into the realm of authenticity and we're going to need blockchain solutions to underpin this stuff to actually make it safe secure and guarantee that someone is who they say they are yeah for all the good use cases of this and you know aside from whether the fact that's going to replace people's jobs it will like any technology can be used for for good or ill right and and people are you know on one hand they'll say this
is really great for uh this kind of text to to audio and AI voice based tools you have that's great for doing like customer service Bots over the phone right so if people can get the same information based on internal knowledge base that you train it on that's going to actually save a lot of time it's going to be able to serve more people rather than having to wait on a phone for an hour to get to a real person but then yeah you have all these other things where you're going to have super realistic uh scam calling it's it's always you know it's it's like when they invented the car that helped the the cops and the robbers right so it's it's one of those things and just on you know on the replacement point you mentioned I think it is important you know to reinforce that that idea of taste right I think that for me when I hear you say about taste being the last Bastion of you know what will be
unique and and worthwhile in the creative industry um that makes me worry more about our podcast in general because because we're quite informative right I I I think we tend to make things informative and that's where things like notebook LM will be very successful because they're for getting information whereas you know the talk ters those kind of podcast of the world where you're going there for the individual for the person I think they're the ones that'll be the least affected so yeah I don't they wear out the words potent potentially but who knows yeah I do wonder I think like on the topic of um copywriting if there will be some infringement issues that will kind of limit the ability for these algorithms to just kind of go completely um mainstream right like it's very hard for a human to say this opinion that you have is copyright what's in your brain is copyright all this kind of stuff but it's very easy with an algorithm say I've put in this paper exactly this is
how it's built like you are built on my opinions if I make a Joe Rogan style Pro podcast tomorrow that is just built on Joe Rogan like there's an issue there um but yeah let's see let's see what happens maybe we need to become a bit more um kind of influency K style start to have more strong opinions and be less informative sell our souls yeah I think uh we'll see how that goes I'm not sure we'll have the most success for it but yeah so then moving on from the audio medium into kind of Visual Arts right and things to do with photo and image generation I think this is you know also one of the first places people saw as we mentioned AI being used mid Journey was one of the big breakout examples it was probably I think I actually used mid Journey before I used um chat gbt because I I was really interested in you know seeing it goes viral very easily right you see the images that people have
made you go wow what can I do with this so that was one of my first touch points and then you know we saw very quickly with the rise in 2022 2023 the big dogs in the kind of Visual Arts world like the creative Studios Adobe for example were some of the fastest to integrate these AI components and enhancements to their core product right you know even things like canva that we use a lot also embeds this now so that's really taken off I think in in the kind of visual medium of creatives yeah I remember using these early on and they were kind of quite cutesy I guess like the images were very cartoony you could kind of see like where the inspiration came from if you were to Google the exact same thing that you're asking for from the prompt on on the actual AI algorithm it would look kind of similar they would very much like copy paste with a bit of editing in between um but they're so Advanced now some
of like the hyper real humanik um images that they create are indistinguishable sometimes from reality it really is mindboggling like how easy it is to use these things and again there's an auth dentiy think are they going to replace artists but we use it on the podcast and it really justes lower the barrier to entry for us to just spin up content have like thumbnails very quickly without us having to give too much of our soul like neither of us are particularly artistic but we don't have to be we just need to put a sentence in it adds a lot of value to a lot of the things you're creating I think one of the things the jumps that I've really seen in the last like few months on this is the hardware that's pushing into the space like every new phone every new laptop is now ai enabled they have like a an AI powered GPU and the whole point of that is to just enable AI processes on that device and the most
obvious thing they push for in a lot of this stuff is for the images it be like the Adobe style editing stuff where I can you know start to create a picture and say you know make this background Mordor or something like that instead of the pretty landscape I have or it's on your camera you take a photo some of the things you see demoed quite a lot is you take two photos and if some people are blinking in one and not blinking the other then you have one photo of no one blinking or one I've seen recently you take a photo with three friends the fourth friend takes the photo and they swap around and then they can merge those two images together it seems really gimmicky but this is kind of like a lot of these devices slowly dipping their toe into this and all the new phones are now you know Samsung 24 plus AI you know Plus+ or something like that it's kind of a bit gimmicky but also it's kind
of the new Norm yeah I I agree and it's it's interesting because I think the hardware angle is is is actually fascinating because it's been something that's been going on for ages if you look back at Apple releases for the last kind of five six seven years they have all you know the whole meme about Apple iPhones kind of tailing off because they're so good and you can't do much in each release well they started really pushing the the computational engines the the photographic computation they do so that you know you're not actually getting more out of the camera the lenses themselves because they're already like 40 megapixels or something you know you can't get much more resolution than that but this this computational layer on top was being really pushed for a long time so the hardware has been worked on but now with AI being you know the big buzzword now then yeah everything's trying to keep Pace with that um one other thing I want to say about the visual arts is
I I it's it's an area where I think it can become quite uncomfortable very quickly for me so as you said we use it for the podcast for you know generating themed backgrounds for each episode so you can see a different theme in it but it's not like a core thing when you get to the kind of face swapping and the Perfect Image stuff that that I find a little bit kind of unnerving right I think there's the advert you see for is it the Google pixel with um the England football team and you have them you know literally changing their whole facial angle or having a better a better profile picture and I find it a little bit I I don't like that lack of authenticity that it's bringing into things even though you know millions of people are going to use it it does make me a little bit worried as well yeah and I think think it's like a lot of the negatives that we have from social media being so pervasive
in our daily lives just on steroids basically you know social media you see the no no no no one% best of someone's life so you're like oh my God is this what everyone's life is like should my life be as perfect as that and now were're taking it to this extreme where it's like you no longer have blemishes there's no longer in photos like everything is just perfect it's kind of like that's not real it's so fake that it's kind of going past what is you know simulating reality simulating something that's actually better than reality which is I I don't really like as well I can see why this is an issue um but yeah it's quite scary I think another thing on the hardware is like you say there's so much there's so much focus on just increasing the computational capacity and I think that is going to align perfectly with you know this new idea of self- sovereign AI models where rather than we have a server that's run by a Google style
scenario offshore somewhere or something like that takes all the data and spits it out I could have a personalized model that is run on my device that you know is completely customed to me the data only goes there doesn't feed anything else and it's kind of like a web 3 version of AI and we had a whole conversation about this with um Alex Kahana and Heather flanner um a few months ago so yeah check that out if you're a bit interested in self- sovereing AI um yeah lots to do there I think the next one so everyone's kind of familiar with you know these these chat prompts making images but some of the things that they're getting more advanced in is music and Film Production I I think you spoke when we first talked about this stuff and you you spoke about like making songs with Snoop Dog style of like some kind of classical music right do you remember that yeah I mean this is probably was around the start of the podcast time
right about a year and a half ago and there were all these things on social media right of people training you know um models based on people's voices know Kanye West was doing the rounds and they got super popular but there are some kind of specific examples that have happened more recently that I think I'd like us to get into as well but yeah it's it's crazy right yeah so one that I've used is um AER music do you want my friend as we were on a long road and got clearly very bored was like do you want me to make you a song with some like details on you so we put it into this uh this Amper music and they create a little song about my life do you want to hear it Jack oh really okay yeah sure I mean is it gonna depress our audience or is is this a good ideaa it's quite funny I think it's quite funny Let's Have It lons and Dragons nerd shows on the he's
a Krypto Bo no more of that what do you think of that H where do I start I mean you I think you do quite a good job of hiding your your nerdiness sometimes but this is not one of those times you know we recently you know got to the the Milestone of 50,000 subscribers on YouTube I think we may have just lost about 45,000 of those well you didn't like it I mean it's it's country I think everyone says that country is the simplest music to replicate but that was again generated in like 10 to 15 seconds or something like that and I know it's not to your particular taste but I'm a big country bumpkin and I like it like it's just it's crazy how easy is for these algorithms to make like very melodic you know simple but enjoyable songs like with just a couple of prompts I don't know it's again quite scary yeah and you know all jokes aside it is it's hugely impressive I me I'd still rather hear
you sing the song to be honest but um it's impressive that it can generate it especially I think the musical element because again to to talk about our own experience when we started the podcast we thought well we need a jingle every podcast needs a jingle and we were completely stuck on how to do it right we needed to get someone else a friend to help us help us out and do it but now you have these tools that could help someone who doesn't have that friend in their lives to kind of you know again reduce that barrier to entry and there's a really interesting example that I I'd like to just you know highlight here on the music industry so basically there's a couple of things that went on but around uh I think around June 2023 so Grimes who is now the ex-wife I believe of Elon Musk um and she's also a music artist kind of famously so she became pretty much the first person to open source her voice so she
basically allowed this AI model to be made of her voice I think it's called elf. Tech and anyone can use that that voice now in song they make um with a with a kind of royalty split I think it's like a 50/50 royalty split between Grimes and whichever artist uses it so that was one thing and then based on that so an artist that I have kind of liked for a long long time called Blau it's spelled three Lao L but I've always said it like that I think it's actually pronounced Blau and they did something really really interesting right so they created the first kind of AI X AI song so they trained a model for the beats in the melody side cuz it's kind of an he's he's an electronic artist he makes kind of DMB electronic Star music so he created the Beats for this track using uh a model trained on his own library of Music stuff he made before stuff he hadn't you know released before necessarily and then used
that for the beats in the backing track and then Ed the Grimes AI model for the vocals and I actually came across this song organically first time so I didn't know it was an AI song I was like oh amazing blow's got a new song I listened to it I loved it and then I found out this whole backstory and now that he's you know creating a a platform around this to do royalty sharing and generate music and tokenize of royalties but it got me thinking because have you ever had that experience of thinking I love this artist I love this music especially maybe around electronic stuff and I've thought I wish I could just have more music like this right and not in the way that Spotify recommends similar music but I want new music that's like this right that I haven't heard before so I think it's it's really fascinating that you could do that now yeah like I would say I'm not really that surprised now with everything that's happening that that's
possible I remember like listening to a TED Talk not May a few years ago now and the point was that um music is very mathematical which I didn't know I'm not very musical but it's very mathematical like that certain notes harmonize and the way that certain go together is it's very like yeah it's very easy to simulate with equations and very easy to know you know through a through a function basically what's going to sound nice and on what won't and this whole there's this company built around it where we started to have like I think it was it R The Edge shearing debate and all this kind of stuff where people were infringing on other songs and saying they were too Sim similar and trying to copyright Melodies and songs and note combinations and all this kind of stuff and this company's whole premise was actually there's only so many Melodies that go together so many note combinations that sound nice that can ever exist so their point was because it's a limited space
and time goes on forever and there's more artists like you shouldn't be able to copyright these things there's only so many combinations so they basically created this algorithm that would go through and get every single good Harmony possible every good single good Melody and try and copyright the whole thing so they could open source it and say any artist is allowed to use these and know how to infringe on this because this is you know the limited search space available so yeah I'm not surprised that they can do this and like kind of say this is the pattern I see from this person's Melodies let's mix it around slightly and recreate it because there's so many times now when I hear oh that kind of reminds me of this song and this kind of reminds me of that song and there's obviously a bit of inspiration there but it is cool the ability to kind of just say I like these five songs create a song that's are the best bits of all of them
but yeah I think it's very smart by these two people to jump in and say you can create any song and I take 50% royalties like that's a pyramid scheme If Ever I heard one well yeah when you say music being mathematical I you make me think of uh just as a side note you know tenet the uh the Chris film there interesting I don't know if you're aware of this but you know love VI Goron score he's got all these paland dromes right in the music that could be played back forth so cool um but you know it was I I agree it's kind of one of the things music you should be expecting the mathematical stuff to be able to replicate it and manipulated in these ways but for me it it was a kind of watershed moment because it was the first time I'd enjoyed a piece of media that was fully AI generated or or kind of near fully gen AI generated and it had been a completely seamless experience right
it was it was as if it had come from the artist itself which kind of and the fact that I didn't even know first time it really genuinely kind of shook me I was thinking about it for days right it was one of those moments where yeah we have we reached the singularity of music maybe potentially so yeah that is quite scary I think everyone's going to have that experience soon and maybe like people that are not in the space will have it more readily because they won't be expecting it but where we first interact with some kind of AI and don't realize I mean there's the boring AI that are underneath the hood that you know the search algorithm and optimization all that kind of stuff but when something mimicking a human so much so and you can't tell like a song or an artist um I think one of the big ones that's related to all this is obviously film making and animation all this kind of stuff that's the big area that's
kind of come up for debate and you know it's very relevant to the grime stuff because there was this whole writers kind of debacle where they were saying that they were worried that you know production agencies were trying to record their identity trying to record them in certain scenes get them to play script and then open source them so they could just reuse that actor you know in perpetuity forever and put them into whatever film make them say whatever script and they're like no this will end our careers end our profession this went on for a year or two years or something absolutely crazy but I can understand why like even the animations I'm seeing online on Twitter through like Sora or Runway especially seen a lot through Runway and Synthesia the other ones youve kind of heard about um they are incredible now they're getting so like so intuitive like it just looks like someone has spent two years on these animations and this but it's just a prompt again and it's made in
like minutes or something like this it's absolutely crazy yeah I mean this is It's been happening for a long time again you know as with a lot of these things so we've seen a lot of controversy around this you know people like uh Grand moth Tarin right in the Star Wars franchise that was one of the first times that people had had seen a fully kind of AI Rend or I'm not even sure if it was AI it was just an artificial version of a a person you know who was no longer around and then had died oh yeah put back into a film and now we're seeing it all the time right it's happened in Star Wars happened to the aliens franchise and how you manage the rights of someone's likeness you know especially as you get older and older especially as you get older and older and you know the the rights might pass through generations and generation generations and then it becomes someone else's decision you know that that's a scary thought
that you could not consent to it while you're alive but some someone one of your grandkids something could decide to to open source your likeness or something like that it's a very scary thing but yeah it's it's happening right that's the headline here all these tools are making it a lot easier now even than when when it first happened eight nine years ago now it's the barrier to entry to that is all reduced a lot and you mentioned Sora that's generating new video content from scratch or Runway as you said generating new stuff from scratch it's you know it it's really mindboggling how good it is and now you're seeing have you seen the memes and stuff like when they've got new versions of memes right which used to be static images that everyone knows they've been around for 10 12 years and now they're kind of morphing into each other they're changing elements it's very entertaining um and impressive but again like opens up a whole Pandora's box of of things to worry about
potentially yeah they are very entertained I saw the um the CEO of Runway actually saying that he predicts that within a year or two years be an Oscar award for an entirely AI generated and produced um movie which is a bit Bonkers like I don't know how the Arts industry would feel about that given everything with the the writer strike but it just it is mindboggling but some of these things I've seen just pure ads on this stuff it's been like a three minut sequence that's been made just advertising a film that's every single part from conception to production to you everything after is has been generated by Ai and they are incredible they are like kind of Dune two levels of production and it looks hyper real and I'm like I want to watch that film like it is scary and I think in my mind I don't really feel uncomfortable like watching these films it's not because I think we've we're kind of used to films not always portraying humans right like animations
all this kind of stuff where it's just voiceovers and that so I don't think that's bigger step to say you know we're just taking the human voice out of this animation um but yeah it it I don't know what it means for the industry itself like I imagine they'll be kind of more CR giv to theater and people going live performances that will be the thing but even when you think about like ABA Voyage which is this kind of VR hologram thing where where you know Abba basically dancing around stage as Holograms doing this whole concert is that going to be the future of like even live theater like I know that stuff cost a lot but everyone that's been to it said it's one of the most incredible performances they've ever seen so I don't know what it's for Live Theater yeah I had agree I that that is gives me the same feeling that the the perfect photo stuff gives me the ab Voyage like even if it's great I haven't seen it
but it makes me slightly scared I think it was with happened with Tupac as well right like years and years ago before all this but yeah again it's demonstrating huge potential and I don't know about the Oscars one because I think the Oscars like to be very traditional and I think they'll maybe be the last to honor this stuff but maybe a Golden Globe or something I can definitely see and you know on the a iing side and stuff as well I think there's there's a really really interesting to use Marvel as a case study because they one of the biggest you know Brands uh franchises in cinema right now I mean they're kind of going off the boil a little bit but they're still one of the biggest names out there and they've into a couple of problems around this right so two to name in you know 2023 there was the secret Invasion TV series one the one on Disney plus and they actually so at this point they were using AI in
a very open way way so they said look guys we've made title credits using AI how cool is this and they got a huge backlash for that because they were like well this is work that should have gone to you know actual artists es and especially because Marvel is uh you know building stuff around comic book IP and comic books are you know famously handd drawn painstakingly done um you know animation has always been a painstakingly handdrawn thing in the past as well people got very angry about that and they faced a lot of backlash and then you know literally last month so you know a year and a bit into the future after that now it seems like they're still using AI people are pointing out but they're hiding it now so they're trying to get away with it so I don't know if you saw this but the the Thunderbolts movie that's coming out early next year people spotted that there was a character in the poster with five fingers so five uh
five digits and a thumb basically on their hand and they were like they're doing it again they're using generative AI to make their posters that's taking work from someone else and we've spotted you right but it's interesting that they're still trying to do it maybe cut costs but you know they're trying to hide it now I mean that's like a really poor old school AI model as well I remember like the when we were using it to create images when like on the surface at a high level looked great and then when you zoom in like people would have three eyes or two teeth and 10 fingers like this you didn't see immediately you look back and be like this is some there's something weird about this phot I don't know what then you zoom in be like oh my God this is truly terrifying like there's some AI images that we used generate would haunt my dreams afterwards well that's the that's the uncanny valley thing again right yeah these older models I I
remember what in particular you know back to behind the scenes there was one we did for our end of year episode where we were predicting what was going to happen this year in 2024 and I was trying to do one where we were both crowded around a crystal ball and it just could not do one where we both had a hand with five fingers right four fingers in a thumb on it it just couldn't do it it was one it was like those uh optical illusions where you know is it four legs or five legs kind of thing it was horrible and I was getting creaked out by it but yeah it's come a long way since then I saw one I saw earlier I don't one of my friends got it up when they were on the website and I think it was Christmas special or something like this and I didn't even notice but there was like it was both of us in like Christmas outfits or something and then in the backround
there was two like little elf children or something and I actually didn't notice them at first and they was zoomed in the bottom right corner and I looked I was like I didn't see those before it looks really creepy yeah we need we need to standy check this a bit better maybe get a real artist to start doing this stuff well and it's and they're definitely not perfect right that's something we should also mention they aren't perfect yet or at least a lot of the ones that the US consumers can use um you know a lot of the time I'll generate an image for one of our backgrounds and I'll think this is perfect this is exactly what I want but there's something wrong that was added in a little extra icon somewhere that I just really don't want and I have to crop it or you know sometimes I end up just going with it because I try and get a better one and it doesn't get better than the original even though it
has that blemish so yeah it's still it's still a bit of a pain at times to use but it's a lot better than us trying to you know to us to hand draw it because we're not professionals and that kind of stuff and this is one of the things about this stuff is like it doesn't really understand what it's doing and that's quite like a philosophical thing to say like what is understanding but when you ask it to do something like create an image of me and Jack In a Christmas setting what it does to achieve that is rather than like just drawing it it try down into like the kind of component patterns of that request and it's kind of assembling these component patterns and filtering it through to create something in the end so to then say something like okay this thing has five fingers if it's basing it on these underlying tricky for then to edit but this is something that is really starting to improve and even like chat PT the
models now you have a specific edit for an image to say okay this little part here is actually wrong can you do it and they are getting much better at this stuff they really starting to break down what is understand and like yeah it really has impr a lot in like the past year on these things and and when you I think it's just really important that you highlighted the philosophical side right and one thing you know as we're coming to the end here that and that goes through line through everything else we've said is that all of these AI generators stuff are by definition they are derivative because they're all generating new things but based on set a set of inputs that we've given it right it's deriving new things from those inputs and re jumbling them and making something and new out of them but ultimately yeah it's not necessarily it's not thinking in the same way that we say it now there's a whole other philosophical debate about well what do we
as humans do when we create things aren't we just taking our inputs and our environment and splitting something out but I think still that there's that link that that is missing what the AI is doing and that that is again back to the copyright question you were saying earlier is where all that stuff gets very very murky right if this is all derivative by definition then you know how the hell do we we account for all stuff in in the royalties right yeah and I think maybe we line different sides of the fense of this I think that it's fine because I think humans do the exact same we just don't understand brains well enough to actually prove that they do importantly but I think if you break something down so much so that it's unrecognizable from the original thing if I break down the Mona Lisa into these patterns underneath that are so kind of like disconnected from the original so the kind of more Bas and fundamental like goes well with this color
these two triangles look well together then is that and then re jumble like you said is that really infringe on copyright I not but then like there's obvious threshold there where it's like if it does get so complicated that the face pattern with this pattern looks so similar you know there is a threshold of area in between and I don't think right now that I wouldn't on the side that I don't think AI copyrights a lot of things but there are some things if it's like 100% identical then obviously there's a copy there and there's some gray area in between but that's above our pay grade yeah I I don't have a strong answer or opinion on this right because again I I agree we don't understand I don't think we understand most of us enough about how the AI models are actually working and also how the human brain works to kind of come down on this hard either way so yeah I think the the monisa of thesis thing is also an interesting
argument like if it's if I've changed all the elements in some way is it still the same piece of art I think that also comes into play yeah but I mean the important thing here is it's not all doom and gloom AI is is obviously incredibly impactful on the Arts right now and will continue to be but there are some pros as well like what we've mentioned even for our podcast like it really is democratizing access lowering the barrier to entry like we couldn't run this podcast very effectively without all the AI tools that we've talked about it really does reduce costs allow us to do things faster allow me and Jack to concentrate on the things we enjoy which is actually recording planning the episodes and all this stuff and worrying less about the production the art and all the images that go out but obviously there are a lot of challenges with job displacement with authenticity content rights all this kind of stuff and this is the legal ease we're hoping will be
addressed over the coming years but it's obviously creating a lot of fear right now where we don't even know have a podcast next year because these tools are moving so quickly and so impressive right now that it's hard to know the future of this space yeah I think that is such an important note to end on that you said there right that I I hadn't kind of fully crystallized until you said but yeah it's it's linked to that lowering the barrier to entry for us as podcasters okay so you're right we couldn't focus on the important stuff we couldn't focus on curating the right what we think are the right ideas creating episodes kind of thinking and planning for episodes and and all this stuff because you know it's incredibly time intensive we're doing this in our in our spare time if we didn't have all these tools at our disposal to make the kind of the grunt work of bringing an episode from idea to getting it on on YouTube and published and everything
then yeah we probably couldn't be doing this so ultimately it's the reason the podcast exists in some ways right yeah I think it's a bit of a gateway drug in some ways where you kind of you start using chaty PT to help like formalize some of the episode content then you're using descript to help with production then all of a sudden you know you start changing words with it then all of a sudden Jack's ill for a couple of months I just use descript to do all his part and I'm recording on my own then all of a sudden I'm like you know let's just automate it all it is a slippery slope and I can see how you like the big companies are going down that route and maybe it doesn't lead to a good place but at the at the moment in my mind anyway we're in a good place where AI is mostly just enhancing a lot of the Arts but let's see where it goes I mean I would encourage everyone
to stay ahead of the trends because this technology is moving so quickly and it's just advancing so fast that it's really you have no idea what's going to happen next week but a lot of it's very exciting and enjoyable so yeah have fun with it and thank you for listening wherever you may be and to Le a little more of web 3
Topics: AI & Robotics · Blockchain & Crypto