#36: Will AI replace artists?

2022 was the year for AI and every artist wants to know what this means for their job so as a special episode to close out the year I'm going to be giving my thoughts on the most controversial topic I can think of since nfts and then following that I'm going to give my seven predictions for what I think we can expect in 2023 AI concerns are understandable I got a lot of empathy because 99 of what we've been doing for the last 10 to 20 years seems to be exactly what AI is good at technical skill sets like creating a basic form and then rendering it very very very well at other one percent skill creating a compelling unique story that's the one that a lot of us have neglected myself included because technical skills are what the market has demanded thus far and now we're finding out in 2022 that that's the one that we'll probably need for the future so essentially this feels like we are starting our career over again the time investment that we've made learning these new skills sometimes over several decades now feels lost and you can find screenshots of AI artists relishing this fact which definitely pisses me and a lot of people off but do remember that social media will always you can always depend on it promoting the worst in humanity and you'll find one of these people in every community now the biggest argument against AI right now is that the data sets were trained on copyrighted images and I'm going to annoy a lot of artists by saying this but in the long run I don't actually think it matters because eventually someone is going to release an ethical AI product trained on non-copyrighted images that gets the same results with the amount of investment in this Tech and the concerns around copyright it's not maybe but when somebody does this it'll probably happen by as early as next year but we'll get more to that when I talk about my predictions at the end so whether we face it now or later eventually industry pipelines will change but to dive into the copyright debate for a moment we do kind of have to acknowledge a little bit of hypocrisy right because look if we're all being honest we know that if the police was to take our hard drive they would find thousands of copyrighted images inspiration folders puref boards Pinterest boards images from large IP like Blade Runner and Star Wars but often images rip from Independent Artists and photographers without their permission we literally call this reference reference for Styles reference for design choices and reference for ideas I've literally sat in the audience of talks at siggraph as they showed animal reference video that they rip from YouTube and use for dinosaur walk Cycles in Jurassic World but we accept this because it's not just fair use but because it's how every artist has learned since the dawn of time now some say you can't really compare the two because computer vision isn't the same as the human brain and I'm sure they're different but at the end of the day it's the same result if we're both inspired by Lord of the Rings and you draw the individual elements that make up the fantasy castles because that's how your brain sees it but then I do it with stippling starting with random noise that arrives at the same result is there really any difference now I will say though that there is a line there's always been a line artists can go too far by tracing and AI can overfit to training data essentially producing images that are not just inspired but almost an identical copy of the source image and this has happened and it is not good but no one knows this more than AI developers because their product is to produce Unique Images and if it fails users will lose trust in their product so you can expect that this will disappear over time as it's in their best interest to do so and I believe with Dolly the most mature of all the AI image generators that it already has as Shutterstock exclusively accepts them so we've all been inspiring each other over the years except now we've discovered we've inspired not just another slow artist but a very fast machine and it doesn't feel like a fair fight because it's not it is unfair to compete with AI and it's unfair to compete with a calculator because the calculator will always win but and this is the crucial Point only when the calculator knows what problem to solve is the most important point that I feel is the most overlooked AI image generators cannot tell you what solution best fits the problem that's the bit that's still in your control that's the bit you still need artists for when non-artists play with a tool like mid Journey they're usually very happy with the first generation because they don't know what to look for professionals do they know that say a token grassy landscape won't fit The Narrative of a culture in Conflict for a movie so they need to keep thinking through the problems from different perspectives infusing it with different cultural references trying different shapes different compositions which is why almost every artist that I've interviewed on this very podcast that has used AI in a serious capacity doesn't see it as a threat because they know there's so much more to good art than just rendering and that's why I think that reframing technology as an opportunity rather than a threat is often but not always definitely the healthiest mindset we cannot control the world we cannot stop Innovation long term but we can control how we react to it that's the one thing we've got control on and that's really my main point here mounting a crusade right now feels like dedicating an awful lot of energy to delay just the inevitable the best concept artists I know are leveraging AI not hiding from it leveraging it to level up their skills and create better Works faster which at the end of the day is what all artists want this does mean though that we'll have to let go of some things certain workflows will be replaced but again if we're creating beta Works faster most artists want this so as you can tell I think AI is a very fun very promising tool but that said I do not agree with everything that they do so for example they should never have let you cite a living artist's name in a prompt that was a very core idea at least in my opinion right because two artists that are living off commissions this affected their livelihoods immediately and narrowing a training data pool like that to such a small degree really blows the lines between fair use and infringement even if it legally isn't it really feels like it was so it was really the quickest way to turn artists against them and it could have been avoided so easily because they went to Great Lengths to stop you from referencing a person's likeness so why didn't they stop you from referencing an artist's IP I can't believe they did that it's so annoying but anyways so if they disallowed this yeah it would be a small thing but it would go a long way into winning artists over to the tools so I'm saying I'm not signing off on everything that AI does but I do think we need to acknowledge that AI isn't going anywhere and it's certainly not going to get any worse this is the first inning of the first game of the first series and each one of us will have to make the decision whether we adopt this new tool and level up or stick to honing a technical skill set that will likely be in less demand in the future so that's my thoughts let me know on Discord whether you agree or disagree we do have a Discord now it is at blenderguru.com forward slash Discord and in there you'll find an AI debate Channel where people are having fiery discussions and you can join so go to blenderguru.com forward slash Discord and I hope to see you there now on to my predictions first thing funny thing about predictions they're almost always wrong but they're very fun to make so I run a Weekly Newsletter called this week in 3D which summarizes the most interesting Tech and news as it relates to 3D artists and you can subscribe by the way blender.com forward slash this week in 3D and because of this it means that I get to see a lot of tech and I can start at least I feel like I can start to see the direction that things are headed and this is why I like making predictions as I said they're almost always wrong but they are fun to make so as the final podcast of the Year here is my top seven predictions for what I expect in 2023 starting with number one you could already call it an explosion of new AI not really a shocker this year was just the warm-up I think in the next five years we're going to see virtually every software package that we use being augmented with some degree of AI That's why I kind of think it's silly to try to like categorize different workflows as like this is AI allow not allow it this is not AI like it's like eventually everything is going to have some degree of AI like some of it will be generative like your mid Journeys your dollies creating a new image and then others will be like augmenting right like taking a tool or a software that we already use and then augmenting it with AI perfect example a lot of photoshop's tools now certainly all their new ones most of them have some degree of neural learning in the background like it used to be like you know lasso select now you've just got an object select button and it just like looks at the image and goes like that's a human that's a chair that's a whatever and you can just click on it and it'll just try to guess it for you so really this means we're going to be producing work at a faster pace which means production costs are going to fall but it also means that workflows are going to be changing at an even faster Pace than they were previously like it used to be you would you know like way back in the day you'd have to go to like a trade show or a conference to learn the new workflow right and then YouTube comes along and then the word gets out there faster right well now with like Ai and Twitter and like people are inventing things all the time it's like they just released like chat GPT and then someone figured out that you could uh ask chat GPT for a prompt which you could feed back into mid-journey and then other people start using that take that image then feed that back into the data set and then you can like iterate on that and it's like people are just coming up with ideas constantly so something that I mean if you want to get a job you want to stay employed the one thing you can do is just try lots of stuff be on Twitter groups be on my newsletter this week in 3D stay up up to date abreast of all of this stuff and um because your employee will want it they don't want to have basically that they don't want to waste money right you wouldn't want to waste money if you hired a plumber you wouldn't want them using the wrong tool and they don't want you using the wrong tool either so you don't want to be the person that doesn't know a new workflow I mean obviously there's always going to be stuff you don't know but you know it helps to stay abreast for example one of the guys we've got a polygon he loves he's like everything he does is AI he's tinkered with everything does a little bit of programming he's just he's a tinkerer so if there's anything to do with it he knows of all the little the pros and cons of each approach and it's like it's great that guy knows everything and we know for sure that the workflows that he's learning are going to be useful at some point in the future so there's a book The linchpin by um Seth Godin who talks more about this but like being a valuable employee is about that that kind of that tinkering and and trying to trying to get the most um trying to give your employer the most value essentially speaking and that's what learning some of these workflows are anyways we're not off on tangent number two ethical AI I already mentioned this but with so much controversy surrounding a lot of these AI models um I think some companies are probably going to hold off from using them um and Founders and investors know this so I think we'll likely see rival companies this year promoting their data sets that were built on copyright free images as I said I don't think it's a matter of if but a matter of when I think it will happen um and you know that's just going to be a a maybe a second market and maybe if it was regulated a lot that might be the only market and which leads me to point number three AI regulation so governments are very slow but they're just starting to wake up to the power of AI there's actually a website it's oecd.ai and that will actually show you like all the regulations around AI um with like a map of the world and you can click it and it's like man like the US has 66 policies regarding AI at the current moment and that I'm sure will Skyrocket in the coming year so um yeah you can expect regulation now some of it obviously regulation is not always good and it's not always bad sometimes if it's too heavy-handed it could stifle local Innovation but when it is done well it could lead to the Public's good in the long term but at the end of the day also I think that much like uh trying to stop the nuke after the first nuke has gone off and then stop other people from developing nukes it's a global issue so like it's kind of the point I was trying to get to like on Twitter I was like you can't stop China right like China if you put in a regulation that like made it so it had to be ethically trained data sources like you just have to go to a country that doesn't care about that and then like they're going to use it and like I know also like it's still like yes but U.S Industries will not and so therefore there's a win but anyways point is regulation is going to go up just yeah anyway that's my prediction number four staff count decreases volume of Studios increase repeat that staff counts at Studios I think will decrease the volume though of studios in the long term will increase particularly and a transition phase right when we go through this like new workflow where costs are plummeting to produce the same content I think we're going to see whilst they're replacing workflows they're still going to have the same budget so that's the thing I think it'll be a temporary measure like if a game's still they've got a budget for 200 million dollars it's like well now we've like producing so much more for like in such a shorter amount of time maybe the game can be longer maybe they can do more with it the goal posts move so you have to so it's a moving Target goal Postman I don't know whatever you want to call um but they've still got so the big guys will still have those budgets but I think at the start you will see like let's say concept art teams particularly Juniors within that some of those positions will reduce however I think with the lower cost you're going to see a lot more Studios and individual starting Productions that wouldn't have otherwise so I was actually talking with the the guy from lumalabs the other day just a phone call you know how you do it um it was just random um but he was saying like back in the day programming like I don't know like 60s 70s right it was so Niche there was only like a like a handful of people that were programmers in the world and there was a photo of this um I can't remember the name of him but he's like standing next to this ream of paper and it was just to compile code right well that was a skill set it was compiling code and then they released something called a compiler which does it for you so that skill if that was your job at a company compiling code it would no longer be in demand but it's not like that reduced the number of roles that helped the number of roles in the long term right because you reduce the cost you made something so tedious go to something automatic that it then made uses for programming go up right so think of for example like you look at indie games today it's very rare to find an indie game that actually like looks as good as a triple A game right they it like the cost of developing worlds is astronomical like games like Call of Duty you can't do it for anything less than at least like 100 200 million dollars right the cost of generating all those assets is just it's astronomical right imagine though if for example some of this generative AI this uh Point e open AI I don't know what you're doing with your names like fire your marketing guy Point e is what they called it and it's a 3D model generator so you type in a prompt like office chair and it gives you a 3D model it looks horrible right now horrible um we laugh at it now but we also laughed at Dolly right like a year ago and then it's like look at Dolly now it's like holy um so anyway imagine that right like if you were an artist in an indie game studio and you'd like to build a world if you just typed in city and it generated a city imagine all of the new applications for that right all of the businesses that might want to have their uh training course or whatever take place in a virtual VR world now because now they don't have to build you know for this everything like you could have like a teaching instructor like driving instructors do like courses I don't know I'm just trying to think like edu like teachers right they could do their next lesson in VR which they actually reckon is one of the biggest uses for VR of the future is education but like if the cost of generating those assets was like typing in a prompt it's now there's going to be so much more and you're going to need people to clean up that work you're going to need not just clean up I'm not saying like that's what you should aspire to do but you're going to any good people who understand taxal density poly counts all that stuff that other people don't in order to take it to the next level so yes there might be some of these like uh like would you call it teams right within a lot of large corporations that will some of those roles as the workflows change I think you'll see a reduction but overall you will now see like indie game developers that could be now like a thousand X of them and they could be producing work at AAA quality on a long enough time frame I'm not saying this is next year this is maybe five maybe ten I don't know to get to that phase um but you know what I mean like it just it opens it up anyways all right so that was number four rambling a bit number five Nerfs entering VFX so Nerfs stand for neural Radiance Fields if you've uh I don't know been on Twitter you might have seen people like it looks like a camera move and like it's like the world but it looks like a recorded video of camera but then the camera is like moving around basically I mean the way I sort of think of it it's like photogrammetry except instead of it building a more like it's just a different way of doing it instead of building a model it's building points and then depending on the view of the camera it's a color value for that point and then if you've got like the camera taken a photo here and then the camera taking a photo here it can then interpolate between that point so you could have an animation of a camera between two points when it's only got two photographs let's say right so it got widespread attention this year this is the first year that anyone really started to know what Nerfs were Ren from Corridor for example he started demonstrating how it could one day replace back plates so in VFX it's very common if you've got you know onset you've got a shot in a street but this other actor's not there or they need to have this other sequence take place on a green screen you shoot back plates so you shoot high-res photographs of back plates problem is once a back plate is shot it's locked into that location if the camera needs to be lower if it needs to be from a different angle you often can't do that right because it was just not photographed on set sometimes they have to rebuild it in CG just to change the camera shot right well if you've got a Nerf so instead of just taking one single photograph you go around with a basically like film the entire location right just walk around just constant movement of the camera getting every surface it will capture that and the interesting thing that's not that makes it so different to photogrammetry is that that because it is point clouds and it's like the value of that point changing it will capture the reflections of information so it's just natural right like the camera moves down and it gets the light reflecting off the puddle right um because it's uh yeah it it's Nerfs it's not photogrammetry it's different so I think next year we'll probably start to see it used in low budget Productions it still has an issue where it kind of all looks kind of blurry and choppy um but obviously that's going to improve and so I think you'll eventually start to see it in then mid and then high-end budget Productions I think it's very interesting there so that's Nerfs number six coming to our last two uh VR I think it will potentially Leap Forward next year in a very big way but this all depends on Apple so Apple for years think like 2016 the first rumors came out that Apple was working on a VR headset right um and it just it hasn't come out and like even like three years ago when I was in LA I was talking to some artists and they're like there's a studio in La that is hiring like 300 artists from like some of the top studios in the world they're working on something and they cannot say what right I don't know what it is maybe they're making their own IP their own games for VR so they can release with some of their own content that's actually good I don't know maybe the via content currently sucks I don't know what those artists are doing but the point is is like everybody knows apple is working on VR this guy on Twitter Robert scoble who's like in Silicon Valley he knows everyone and he's like it's coming this year is going to be big and it's not this year maybe next year um so really it's like it like it was really poised for this year so the fact that it wasn't this year I'm like it has to be next year but but anyways if it does right Apple isn't known for its like flash in the pan like here's a product and take it what you will you know we'll see they typically they do that only when it's like a real solid investment in a long-term strategy so if if they were to release a VI headset it would signal to the rest of the world that VR is a thing right and I think you would suddenly see like a 10 public increase in interest on the medium all right that was number six now on to number seven metaverse hype crashes so they say that the impact of tech is overestimated in the short term and then underestimated in the long term so an example social media in like 2008 it was the new buzzword like Web 2.0 right and it was this new thing and everyone talked about how it's all going to social media social media is going to be huge it's going to control everything and it didn't really for like the next few years television advertising people still watching TV and social media was just kind of this weird geeky thing on the side right it was overestimated its returns in the short term but long term it was underestimated every one of those Tech Bros that was hyping it up back then did not even imagine the impact that it would have on society today right then now it's being regulated and all this kind of stuff right so it's it's the same with it it's actually called like the Gartner hype cycle right so with the metaverse peak reaching the the hype of this year I think you can now expect it to go through what is called the trough of disillusionment in 2023. so this is basically where it's like now the hard stuff happens like everyone's heard about it everyone knew you didn't deliver and now you've really got to do the hard work for many years to slowly pull it up into something of value um and I think that's going to take like another five to ten years it's actually a book called The metaverse by Matthew ball and um he explained like basically he kind of puts it as like something is of like it's the key medium when you could ask it when your aunt is referencing something she saw in that Medium now your aunt is referencing things she saw on Facebook right or has been for the last five years uh he he reckons one day one day your aunt will actually reference something in VR she'll say that she saw something about that's when it's hit its moment and he puts that at like five to ten years out sorry not five it was actually ten years for that specific example so just to give you an idea on uh time frame metaverse not happening anytime soon but it's not just the zoc that is like nuts about the metaverse or thinks that the metaverse is anything a lot of companies do it's it's actually like it's kind of crazy if you look at the um the metaverse standards Forum which is this like form of companies that are trying to set the standards scroll through that and your finger will get tired so many companies are um investing in it in a huge way that's just I mean it's a free forum but it kind of gives you an idea of who is thinking about it um anyways so that's really it this is a I guess a little bit of a shorter podcast than usual a lot more controversial than usual that's for sure um but yeah thanks for listening and I hope you continue to listen in the New Year have a good Christmas I'll see you in the next podcast

Channel: The Andrew Price Podcast

Views: 45,707

Keywords: podcast, andrew price, blender guru podcast, andrew price podcast, advice, help, artist, art, career, learning, hacks, technique, habits, art habits, ai, midjourney, ai ethics, dalle, stable diffusion

Id: Ao5aV4I4Ivw

Length: 25min 8sec (1508 seconds)

Published: Fri Dec 23 2022