EP#26 | The Importance of Critical Thinking in the 2020s (Bosstown Dynamics, Deepfake Technology)

[Applause] with your host tiny tiny chair say hi we are back with episode number 26 of the corridor cast it's good to be here I'm here with guests okay our real names there are real names should be reenter this so we don't have people's eardrums I think people like to have their eardrums destructs for your I'm Jake what's your name Sam it's Jake Sam Wren what's your name what's today Orbis if you guys are listening only and we might confuse the resume only you cannot hear the sound of the waves the same our opinions will be different different brains yeah well put these t-shirts because today's sponsor is not the corridor digital merch store its plug in everything that's right guys we're gonna tell you blent more about them later but for now let's jump into our first topic hey the Boston Dynamics video speaking of plugins so recently we released a video called bus I don't even know what it's called Boston Dynamics robots fight back yes new robot fights back you know it was a viral hit it blew up over the weekend yeah and a lot of people saw it a lot of news articles were written about it and it was a big deal for us talk about that a little bit here having experience yeah with me and Jake what was it we were in Portland we were walking down old let's go all the way back yeah let's let's roll this all the way sound effect episode inception is basically we had been talking with this company xns who makes the motion capture suit they were like you want to mess with this you want to get it on loan I said yes yes yes and then basically we had in studio we played with it we figured it out you guys made the Transformers piece and after seeing how that turned out and how cool it looked it all kind of put our heads together and I specifically remember ujq words like you're like this is so cool we have to do something else with it before we give it back and we're like yeah we do yeah we definitely have to do that and we put our heads together we came up with a few ideas that we shot down but then in a stroke of genius Clint was like what if it was me and Jake in Portland and we were like we were like just riffing on the porch you know as you do yeah and we were like yo Boston Dynamics meets the xn suit and it's the scientists beating up the robots we had a few other ideas before that yeah cuz they will do the Amazon warehouse well yeah but that was the one that I remember thinking alright this is a solid idea the idea being that like in Amazon robot is taking over our jobs because they Sam sampled at that article it was like new it's like an Amazon robot that fires people basically an algorithm they'll fire people so we're gonna literal eyes that you have a robot just like this Boston Dynamics guy roll rolling around firing people yeah so funny but yeah it didn't have that magic touch to it yeah yeah a lot of lot of stuff needed to come together to shoot it but yeah this this one kind of clicked and it's funny because it was just like a parody at first but who knew what it would blow up in become then after the Amazon video we went to geriatric transformers which I don't know maybe that's more of a fun idea to talk about that it is – see I like Oldsmobiles [Laughter] yeah that one we had a good time riffing on that but then you know nothing nothing felt quite right on those ideas until and we were just walking on the street and you're like well we could always go to the Boston Dynamics idea and then like what what do you mean go back to it where did this come from this is this is the perfect iteration of this concept yeah so I think the one thing is is that we didn't actually set out to fool people with this no no what about to make the most interesting accurate parody blossom dynamics that we could yeah yeah that in fact it was try to make it as a realistic-looking as possible but we didn't specifically set out with the goal of tricking people into thinking they're watching mainly cuz we didn't think that was an achievable goal exactly this is awesome it's icy hot it looks 100% believable but in this day and age who I don't know who sees a video on the internet and believes it I guess actually the answer is actually everyone but optimists we were talking about how like the difference between people watching it on the coridor Channel versus people who saw just the snippet on Twitter or Instagram yeah because they didn't see the full video where we had like hey we're selling shirts too and like it's just me goofing off in a green-screen kind of thing to clarify that basically what had happened is that for our for everyone listening watching it who don't know the our video ended up getting ripped pretty much immediately and re-uploaded to every different media platform mainly me me my platforms mean platforms you know like aggregates yeah mainly Twitter was the main place where the views started taking off there is one video that ended up with like 17 million views in two days in several other videos with multiple millions of views all on Twitter from different users re uploading the video and so when in the process of re uploading it they chopped out the whole part at the end where it's like hey you know check out our merchandise where it's like obviously the illusion is broken at this point so without any of that people are just scrolling through Twitter and they see this robot getting abused and then the video ends and you're like dang it went crazy and you're like okay that Boston Dynamics is up to some shady stuff these days yeah next tweet yeah exactly I actually even I think the most effective cut down was the one that starts in the middle of the test with it like there's no one walking in the building it's just boom robots being tested and some of them that ended before it actually goes rogue yeah because of Twitter's like video like limit some of them were were trimmed to be Batwoman is yet to yeah 20 seconds on Twitter so I found that really interesting because what that started to do is create this weird illusion that this could actually be a real video and no and and and no one read the text in the bottom we argued about this for so long it's like do we write Boston I am dynamics or do we not do this and then obviously the there was this you know you were very clear Jake I like it we can always change the title if we write Boston Dynamics there but he cannot do it if it's baked in the video so we ended up going boss town dynamics yet no one read it anyways so didn't matter cuz over they cbos just like people nor like that just like oh you just don't even think about it especially if you're already familiar with Boston Dynamics yeah cuz you're also staring at a robot getting kicked and shot in the face with a gun and stuff like that so it's like you don't even look at it but other people is the first thing they see mm-hmm like I showed a friend she was like last town dynamics I was like that was literally the first thing you saw the one guy who though who was like I know this wasn't real because you guys have California plates it obviously you're not in Boston obviously this is staged and then yeah yeah that one post he shared is so funny it's like obviously this was staged and like no it's all CG it's like what it was even just like what it was like I don't think you're right I'm pretty sure it's not CGI in that they only staged it yeah it's like no they just got the role the old robot out of the closet to do some training simulations dude one sent me a message on Instagram that was like I they had to have been trolling me but it was like this one guy it was like CEO of mega core of industries and like hi I'm an arms dealer and I would like to purchase this robot that you use I would like to you know the statistics and whatnot we were looking into something like this for the future thank you goodbye and I like you find this again so we can look at this so here's the thing because that is that was like an absolute joke until you go to his profile and you're all of a sudden kind of like hesitating like wait a minute what is he like actually serious cuz he actually has a bunch of posts that are like no if we want to call it out but I don't don't don't read the exact name of the guy because but I'm really curious cuz that's the best that is fascinating but I'm like you're you went to a film makers Instagram page it should be incredibly obvious just by judging at the other eight posts surrounding them one this is is this a sign of the times guys because you you take something that you like think about all the layers going on here it's like okay you've got corridor digital we're a little studio here in LA we make like high-end concept videos okay we got this great motion capture suit which is this really sort of cutting-edge technology in the way that it functioned and allowed us as a small independent studio to access that technology and take advantage of it right which isn't something that traditionally it's always only been like big movie studios or big game studios right the equipment is expensive so that's that's Tier one can right that's tier 1 and then tier 2 is let me know you take all this experience get more the regular audience that we have the experience we have making these videos and we post it on to our primary social media platform corridor digital on YouTube and it gets this initial amount of green screens hits and views and then obviously people with these so that's Tier two we post it on a platform that has followers and then people that have meme compilation platforms like on Facebook and Instagram well there's all these meme profiles to your three they take that they chopped it up right they don't you know take it out of context you reduce the resolution of it so you know and you put it on onto like a Twitter post so you go from like this 4k upload I don't know if it was 4k or not anyway yeah 4k down to like 720 then cut out cut the heads and tails on it do just like 60 seconds or 2 minutes upload that back on to Twitter condenses it even further well yeah it's the photocopy effect just like you're photocopying a photocopy because you're downloading an h.264 video right we uploading it it gets reprocessed then other people are downloading that alright many people were downloading the one Russian Twitter video as opposed to our YouTube videos I'm so downloading that one so everything's just been like so now there's no context the cott the concept is still there like right like the initial artistic concept that we went for is still there but now there's no context you're photocopying the work and now it's like is this real where did this come from what is this and I think that's the interesting element here it's it's kind of the nature of virality has changed since we started YouTube whereas previously the nature of virality was simply sharing the content directly from the source like there was certainly reuploads from time to time but I'd say like the true original virality was you create a YouTube video and then you have a direct link to it because it's easier to share a URL then just straight-up download a video and re-upload it on your own yes my email change that was the original viral yeah but yeah exactly yeah but you know you have Facebook you do on Facebook or you put a youtube link on Twitter or something like that but but it's crazy watching this go viral cuz the last video that truly went viral for us was the real GTA video yeah sure and Nico would make an argument for nerf John wick but I don't think that's nearly the same sort of caliber is now a real GT I was totally I mean real GT is still the most views in one day we've ever gotten on the video but but um but it's really interesting because the thing is there's like a it's kind of like a little paradox or like catch-22 kind of situation going on here where it's like we don't want people re uploading and profiting off of our content but without that people can't share it across platforms and it can't go viral mm-hmm you know cuz yeah because those other media platforms literally uh suppressed outside links Facebook I know for a fact does this if you share a faith YouTube video on Facebook it's not gonna get the same share ability whereas if you're just directly upload to Facebook it's your set because they're not profiting they're there they're taking people out of their sight yeah yeah yeah I mean thankfully I think on Twitter you can when you throw a youtube link on there it's still embeds and it still plays you can on a mobile app so it's not like but it doesn't autoplay in the tweet itself and that's the key difference I guess that is the key difference but you know it's so they're always prioritizing their own content over external links so without without trying I think we we introduced not without trying without effort but without trying to do this end we've we've we found ourselves in a position where we we caught a little bit of light being shed into what it's like to like create fake news and all these other kinds of things that are a little bit more sinister with social media yeah because you can very easily see how you can go from like viral video made by a regular studio regular guys just like us to somebody being like hey I'm going to incept this story or this idea into the universe yeah and with like malicious intent yeah with malicious intent understanding that this is how things work you know someone takes it they reapply you photocopy effect you were uploaded to these different platforms it's not enough context you don't understand what it is you don't understand where the source material came from and all this sudden there's there's a piece of information out there but I think yeah and it's funny because for me like we make the video we put a lot of work into it when I see it reloaded on someone else's page when they don't credit us and they're profiting off of it in some form or another the only the only like satisfaction I get when I see that is knowing that I tricked people thinking robot has come to life and it's fighting back and they legit got freaked out and like I I takes like it's so much satisfaction like I love freaking people out like that ah sit all day and someone hops in the room like a feeling I've seen in had they make an entire article clarifying to people that know this video is fake exactly it's okay yeah which makes me like kind of feel a little bit better that like if if we really tried to fake people out right yeah like well I got caught on day three maybe even day one or two there was there was other site saying it's fake before CNN right yeah but what if we put it out without the ending where we were like selling shirts what if we put it out without any context and maybe even minus the shot words getting shot a face like what people just be like this is real is real is real or would there just be still people would be like zoom in you can clearly see here that it's not real you have to take a few more steps than that but like to just to just take that and extrapolate it like if we were we were a fake news organization we had some agenda we were hired by somebody to be like they put this out into you know some backdoor alley like briefcase deal like okay so you you put it on first of all put it on Facebook and that's the first place right and then Twitter and Instagram and and do it from a profile that create a few posts beforehand so like the throw away profile yeah throw away profile so that the platform thinks that you're a legitimate profile and you put something like this out there and I mean yeah it goes back to like trying to fool people wasn't our number one intent it was like second or third in the list of things were like number one was like we're just trying to make a fun video you know it's like something funny yeah and two is like just try to maintain a caliber of quality that we've always tried to do which is just try to make it look good and a side effect of that is when it looks good enough it actually looks real and you can start fooling people but trying specifically to fool people wasn't art I mean we have boss town dynamics in the frame of every single shot if we were actually trying to fool people we would have just said Boston right yeah next time it'll calm slash door ya know so it's been really interesting seeing that blow up I mean like Wiz Khalifa saw it Stephen Fry was straight-up fooled by it is it's like I feel like anyone who actually sees the video and tries to figure out if it's fake or not we'll be able to conclude pretty easily and quickly that it's not a real video yeah people were falling are the ones who didn't put any effort into discerning it like they're just like they just see it on their Twitter timeline or Facebook feed and they scroll by and they just there's an inherent trust that what you're seeing as a video is real right that's the craziest part cuz like people you know I mean I do it too I'll just read a headline and like oh that sucks and I'll keep going you know I don't read the full thing if I don't care that much but I read the headline like ah interesting you know now I know that it's like I'm not gonna just judge the book by its cover yeah and believe it this conversation though is just leading me to like what we have to do for the next one I know I know I'm not I'm gonna say what it is but we've talked about it already dude are you saying you want to try and make this as realistic you not do that full proof is pieman because I'm in firearms training indoor-outdoor oh yeah every time I add it like a kill house you know with like little walls a nice ambience walks through the realize someone tries to shoot targets like three guns we can't even talk about like this I know someone's gonna do it I know we've already shot it that's right we have everybody uploaded its person [Laughter] scheduled I want to hear about the sound design process on this video because a lot about the visual breakdown between Clinton I but you did a lot of work on sound design yes and I think that helps fool people quite a bit because I usually I tried to keep it like really rich and like meaty I guess but still keep it very bare-bones because I started by doing a pass of like funny sci-fi gizmo gadget servo sounds and that helps like accentuate things and bring a little bit of like a emotion to it I guess because because basically I would go through all the key moments where it turns or does a sharp movement or something like that and and I put specific sounds in for that so it's like constantly I guess it's like ear ear candy you know what I mean where you're just like you know and it sounds really cool but that wasn't what made it sound realistic and what I did to make it sound realistic is I literally just took parts that were metallic or like metal beams and rods things like that and I went back to the loading loading dock where we filmed it and just got raw Foley of just like shit like around like all the footsteps all that stuff and I did it in the same exact space that the scene was filmed yeah that way it sounded exactly like it was happening in front of us all the metal clanking is you're mixing that sound with the actual real sound that was captured of us like shuffling around the ground actually hitting exactly the back with the kaki stick yeah so having all that spatial like reverb and all that stuff matched exactly makes it just very convincing well I remember and at least initially one of your first or second passes of sound design was like everything was basically there I felt it was still a little too sci-fi because I remembered there is some sound effects that sounded like straight out of the Transformers movies like that there's there's a very unique like electrical mechanical sound so you helped some of that in there and it was like every time I heard I was like this is just breaking the illusion for me just sounds like that little friggin mini robot from the second produced after that in the final upload you you did something differently because I member listening to the final thing to me like that's that stuff so real that's great I didn't have a chance to tell you that thank you yeah I key with to making it sound like real in that last stage like getting that fake Transformers element was basically taking all those produced sounds I had and making them about half as loud as everything else hadn't seen like each time a box is dropped or smacked or the robots hit or a footstep like the impacts in the scene are are loud and everything else like well if you had a robot like that and it's in the scene it's not gonna be just like you know it's not gonna it's not gonna be that loud so to speak so I made sure to keep it as subtle as possible and let's just like the footsteps I can make the impacts actually hit reverberate plus you also used my my drone for some of it and then that was the little key yeah a little startup beep there's a start a beep when you plug in Oh is like Italy yeah exactly so I kept that chopped it up messed with the pitch so that it didn't seem like it was the drones sound India just did like little beepers but we tried actually doing a thing where it's like I actually would like try to like move the motors around and have it like do things in correlation yeah the video but was it like there's a dog barking or an airplane or something like it didn't work one who's ended up being like really subtle actually it didn't have like the size it felt too small I've been yeah the second part was that the high-pitched whine of the servos made the dog that was like above us and like the loft above us start barking so it's just this sound I just couldn't clean it up my favorite thing you did in the sound design Sam was the end went right before I kick right in the nuts this sound you put it was like oh yeah as the beeps amazing have it yet you basically have to make it talk by like doing musical notes or something like that yeah that's what I that's what I Meghan but you so you mentioned like kicking me in the nuts that was that was an interesting solution that we had on set so initially you're like alright I'm just gonna try to kick you in the nuts but not actually kick you I'm just gonna get really really close you have to trust me and like you did one test kick and I was like hell no yeah I can't I can't handle this I was like just eat you may have been like maybe a foot away that there's still too much just mess with my house like side-by-side comparison videos that you just you did the full kick with your thigh but you didn't extend your knee and then when you did the same thing to to jake behind you you did a full-on kick but he didn't fully extend your leg because we can actually go through and animate that leg to be extended cg knows we have that control if we can do what you have met him you did it and you did it really quickly and easily – oh man hey guys before we jump into our next topic which is gonna be about deep fakes that's right I wanted to tell you a little bit about today's sponsor that's right plug in everything plug in everything has a bunch of awesome After Effects plugins now we use them over here they've also been used by companies all over the world and booting Google and Warner Brothers one that they wanted us to tell you about today is deep glow deep glow takes your organic lighting from your footage and it makes it way way way cooler it's for After Effects so if you guys are looking to do anything in that program consider plug in everything there's a link in the description below you can also go to plug in everything dot-com and peruse have a good time and I think after this Ren is going to tell us about one called coffee break which apparently can extend your render times more on that later but anyways a huge thanks to plug in everything guys go and check them out link in the description now back to the next topic now – so did you talk about the coffee break so they have a whole list of plugins but there's this one effect that is called extended coffee break and mmin fact that you apply when it comes to rendering that will actually slow down your computer so instead of like a five-minute render it'll take half an hour so as its rendering you know typically like oh my computer's rendering I can't do anything about is no game a smash yeah so it's like you can actually now delay that if you want to it's it's it's kind of messed up it's shady and it's malicious frames and you want to like extend that out so that you can take an extended lunchbreak is an effect that you can do you guys were to get it plugging everything there great yeah if you want to wants to buy you a fast new computer just use that plug-in every day thousand dollar machine yeah if you want to of corporates with 17 graphics gpgpu thousand cores it's about that sharpness inflection at the end that anger like sharp anger and flex and your brothers are still taking ours your special isit keeps you in this office you to unique we were talking about how the Boston Dynamics video led to this realism that we didn't think it would happen and and just to link up our next topic deep fakes depth faking the the the the 26 just I'm gonna bring in a couple world events here okay call cultural zeitgeist events the 2016 presidential election in the United States of America was fraught with all these stories and conditions of fake news and then later it came out and the Congressional Intelligence Committee and multiple other sources of highest authority have it have stated and admitted look there were hundreds if not thousands of stories introduced on social media to affect public opinion now with deep fakes right right now it's 2019 we have an election coming up next year okay there's already been some deep faking recently with reeling with yeah with Nancy Pelosi that wasn't a deep fake though sorry wasn't a deep fake but it was a I don't know the clip you're talking about it was it was a thing where somebody had basically made her say something she didn't say and for about eight hours or so which was a story should watch the clip because I thought I heard it was something did you watch the clip no I did not I read about it in an article after I would I might either we watch it or I would maybe put that one on the side to be totally honest I think he just literally slowed the video down that is just a is just sort of a stepping stone to what I'm really getting to which is like our deep fakes going to be good enough next year to actually affect like people's opinion to fake people out I think so I know you're gonna bring Nikko in here to talk about more defects cuz out of everyone here at the studio he's the biggest pro on it but like he just gave us this whole like hour-long tutorial on how to do defects mhm and now I know each one of us can do it like it's not a hard skill to learn and that's what's scary is that like right it's low barrier to entry all it it's like literally you just download this free folder of files and you once you learn how to like use which file it suddenly you can now do deep fakes you don't have to know how to do after if I just have a computer kaien yes I so I think deep fakes in the realm of like personal attacks and like that kind of scandal stuff like someone being you know having affair things like okay so first off like I think Trump has shown us right now that no one cares anymore about that like oh no Hillary Clinton had enough I'm just saying like I'm not Sara Lee saying like obviously they'll be used for all kinds of things what I'm saying like across the board there's so many different situations you could use them for to like create make up a story you know not necessarily about someone I mean yeah in particular but about but about like a situation that's representative of something going on between yeah party politics like it could be anything and that's what yeah and I have a very strong feeling that late next summer like once we get into like August of 2020 that's when we'll actually start seeing some of these things pop up cuz all it takes is like like you're saying just someone with our skill set and malicious intent right specifically focused on trying to fool people we're gonna start seeing those videos pop up over a long period of time like if I think that if if we don't know if it's real or not that's when it's the issue you know because it's like we just put out a bunch of stuff right now we as a company know deep fake is realistic enough to let me let me just and also I just want to point another thing out with the deep fake thing a deep fake requires an actor and a set ran hand like a cameraman like you have to stage the scene ahead of time like the moon landing or something like that for example and then you have to bend you skin deep i joking you know saying like you need it's not just about like it's that it's not as easy as you think it's literally a production you have to film scene with a cures and and a visual effects supervisor and stuff like that basically – let me pose it and then to use and put the face on there so it's like there's so many fake things in the scene already that it might be easy to point out that I feel like it's still gonna be tough I agree right now that's where it's at I think obviously the technology is gonna continue to get better but if you take cameras you can you pull up this news article just just just right there the image that you have that Amir Putin where there's there's two two of them side by side technically there are different kinds of deep fakes the 1d fake we've been doing is basically taking a performance and putting it on someone else those are just changing initial performance that's a something that's what I'm mostly concerned about is the ability to take the same person and change their performance and change the things that they say and I think that's gonna be harder because I think that's yeah that's like the performance mapping thing where you are literally yeah I feel like that's a whole no technically that's not a deep fake right but I feel like that 24 about saying like what technically is or is not a deep fake but I think we're very quickly approaching de fake is now a term that applies to any sort of manipulation of faces yeah just a broad stroke it's a deep fake if you try to fool someone into thinking someone said as a color here's my prediction here Sam's official prediction check this out okay deep fakes will be used in the next election but they will not change people's minds they will only embolden to their pre-existing opinions so imagine there's someone who likes to make fun of a democratic or republican candidate they'll they'll get an actor put a wig on they'll do their little spoof though deep fake the politicians face onto it and then they'll send it right back to their user base right who already believes everything they're saying and already believe it's it's it's all a deep furthering them yeah it's like it's like imagine we're doing a Hillary or Trump mean let's say you you you're a right-wing site and you make fun of Hillary every single day deep fake her on to another person's body send it back out the audience same it's like it's not gonna actually affect anyone's opinions because the only people who are gonna believe them or share these things are gonna be the people who already are in line with that's still dangerous I would say it's just as dangerous is what we already saw where it's like it's this internet meme culture that affects politics yeah making fun of people yeah I don't know if it's more dangerous I'm just saying it's another tool in the malicious man's toolbox yeah you know yeah yeah it could be like a bunch of really tough right wingers either going after like Bernie or whatever going like oh look at how much of an idiot Bernie is because he said all these dumb things and you're gonna have all bunch of lefties on the other hand probably some big Trump videos going oh look out dumb Trump is he said and everyone's good it's the world's could be exactly the same that is what happened right before the 2016 election not with deep fakes but with news yeah yeah yeah that's what I'm saying it's just it's it's so they go back to what Clint was saying just about how you know you're scrolling a feed you read an article you see something you take those two seconds like people did with the Boston Dynamics video you move on mm-hmm at that point you know you're not gonna fool everyone obviously but if you fool if you're fool success rate is 25% that's you you know and in your video got 50 million video it's like that's that's real damage so yeah that's true what what can be done yeah now I'm thinking about it like oh it's harmless in the sense that like we're already fucked thinking yeah because in this world like we're in it's like we're gonna be dealing with this now for the rest of eternity I think it's gonna require a couple of things on the part of like the major media platforms out there first of all I mean there are companies that are already making a software that specifically identifies something that is or is not a deep fake originated from all sort of like markers that it should exist we have classic videos what they Adobe released some things on that I think I think as this ramps up raw will so start seeing more security ramped up but before we change the topics that's super cool I think you guys I don't want to change topics that's what I want to say change tone is a YouTube I want to see Facebook I wanna see Twitter bullet just like they scan this stuff for profanity or content like that they need to scan for fake stuff and have they'll have little markers in plan I was gonna say yes Sam is the AI you know that the AI technology that they're using currently is not being used nearly to the extent that it could be you know well that's a smarter every month because yeah that and also verified profiles posting like to say say say the algorithm and I know that this makes it harder for individuals and this is where the catch-22 comes in because am the the algorithms should literally be only promoting news and content to from verified from verified channels yeah but then at the same time it's very hard to become a verified Channel if you're a start out person start out creator start out contributor so you know that's the catch-22 but I feel like no matter what the the talent will find its way through you know like so I'm not as concerned about that but like I think that's the only way to you need you need like some human manual review is say like okay verify this channel manually then only verified channels get bumped and then that helps curb some of it but isn't there a difference between a news channel and an entertainment channel because like what's to stop the this a the algorithm from deciding between like a deep fake on the news and a deep fake for our Tom Cruise thing totally you know so like where's the line on that dude I think you just accidentally hit on a really deep question because it's like all news the state these days are not they're innocent they're not news networks they're entertainment channels that's what they are these days yeah that's a good question you're right you're right sorry red no but moving on sorry I'm trying to segue into getting Niko in here because I'm in here then we should bring once ones this conversation sort of tones down then we should call in Niko and have them come in here too okay I think I think I think sounds good to me I think one of the biggest my biggest concerns are reading about the people who have to do the manual reviews at YouTube and Facebook and Twitter who are taking down PTSD all the content that we don't see oh god yeah and I've been reading about this stuff and it is it is the most horrifying thing in the entire world eaten what do you mean Facebook it's okay so first off these people are the people at Facebook anyways who have to review all the content that is like child pornography and like gore stuff like that do they to combat that they've they open up these giant offices of people who just have to do the manual review of all posts basically getting flagged and reported and stuff like that make the decisions whether to take it down or not these are jobs that didn't exist years ago but they do now and what from what I'm reading it's like the working conditions are terrible they're like super super stressful and then on top of all that basically have to watch child porn and gore all day cool and so what are we turning these people into it's like what is going on I'm sure it is the most horrifying thing in the world and the faster we can get AI to take that place yeah these people the perfect solution for AI to take care of in the future well I mean hopefully it can but the problem is is is we have these companies now that like are in this position where they have to do this now they they have to do that so yeah there's no other way than by sending hundreds of people into these little cubicles to watch the world's worst images for like eight hour eight to twelve hours a day and it is like and it's like you think drone operators in Vegas have it bad dropping bombs and stuff it's like no way man like no Act that's probably way easier to come home from than these guys at Facebook so crazy right then you do like we saw like 24 hours a day yeah and then go to Facebook and then he just like watch just like because that at like I always I always think that like you know probably cops have really hard jobs cuz they deal with the worst of humanity on a regular basis but this is is deliberately dealing with the worst of humanity yeah it's not just like oh this person did this it came up now here's all the other people that are normal know here's here's intentionally we need you look at this yeah they have to watch every single New Zealand shooter video every single jihad beheading every single like god knows what but anyways I'm gonna stop talking about this okay stressing me out we're already doing this videos already demonetised all right well should we bring Nico in here to talk about his take on deep Oaks yeah no I'd be really curious to hear it so I will watch this podcast later and three two one I'm Nico what a cut demonize dynamics video that led into like the sense of like how things get photocopied and things get taken out of context how that if with done with malicious intent that can create fake news and then now what's upcoming a new a new thing that's sort of similar what the possibilities are with deep fakes and so we kind of went through that now bringing you in we can just start with like okay Nico you know a lot more about deep fakes than a few of us since you just did a whole thing on it so tell us like what the process was like maybe creatively and then like we can get into the some some sort of the bigger picture stuff after that great okay cool and three actually Dean I don't want you to cut this part three two hi Dean one hi Dean hi Nico you didn't live at Joe Rogan the style of countdown I don't know snip their word he leaves a new dimension we don't we know incision Joe don't say what happened Joseph rubbish yeah we're talking about deep fakes we're talking about the implications of deep fakes but I fear I think first before we get into that it would be cool to hear from you like creatively what it was like creating this first video that we just put out and then sort of where you think the technology is at right for sure so I'm something that most people listening to this podcast have probably seen on the court or crew channel and most of our Instagram so we have this deep take please subscribe guys Cruz yeah yeah here I subscribed consider doing that but um yeah so there's an idea I've had for a long time they've been wanting to do which is a hire a celebrity impersonator and put a deep faked face on them because it's one thing to do effectively a mask put a mask on somebody a face swab it's another thing to have the voice and the mannerisms in the costume and really the voice at the end of the day that's the hard part I know there's like some early stages like voice synthesizing tools but even then like you're not gonna have nuance over performance you know like the moment where even the impersonators telling me to untuck my shirt like you can't fake that yeah could you please untuck your shirt yeah exactly um not to mention you know being able to play off each other's the same reason you recorded Boston Dynamics with Clint to actually in the shot you know I need to have human beings acting off each other yeah unless like you want zero spontaneity in your scenes anyways so yeah I had this idea to do this celebrity impersonator deep fake for a long time now two years ago or so is when like the first deep fake stuff came out like before that there's an academic paper where somebody basically wrote you know here's I can train an AI to recognize facial patterns and if you've ever seen like Google's deep dream or other things like that you know basically somebody boiled down like how our how our visual processing works which is you see you know black and white you see color you see shapes you see line you see form and then these eventually get turned into symbols in your brain and the symbols basically I used for communication everything from as simple as oh that's a phone or a cup to everything that's complicated as oh the car is moving towards me at this speed you know that there's mythologist coming down from shape recognition and tonal recognition and that kind of stuff and so if you step away and you abstract it all you can do the same thing with computer it can recognize tones and shapes and forms and patterns so here's a neural network and you can kind of train that and so just like how we can recognize faces people you know basically these people out of university kind of figured out the same thing and that's when the first like deep fakes were invented now these papers are published the algorithms are published and you know it's not like somebody's just sitting there with a nice ready to go program you know you just have an algorithm now like somebody out there has to sit down and do the work of turning it into a program yeah and so somebody actually did that back in the day and that's where like the term deep fakes comes from just from like this early a soft excuse-me software that people had made I think it's just a guy the problem is it was like the it was the roughest like most basic program ever seen like you had to have a very specific driver version installed for your GPU you did they have a very specific CUDA sit like environment installed a very specific visual basic environment stalled so like basically you had to have your computer frozen at that point in time to ever use the program and so he didn't have that specific GPU with that specific driver in that specific visual but our you know microsoft.net framework all installed in that machine in a very specific way you couldn't run the program so needless to say very few people made deep fix and just recently another guy I think his name is Igor could be wrong I came out with a program called deep face lab which is using the same algorithms people had worked on the algorithms since then people like you know instead of doing a half face which is from like the bottom of the lips to the top of the eyebrows and they do you know full face now which goes all the way down to the chin you know that was a new algorithm for example lets me just basically made it bigger than the other one yeah you know yeah there's a little tweaks and quirks and you know somebody's added you know color matching and somebody else matt added like edge blurring and you know all these little things have been added in to the program there's other programs out there too I think it's one called like face to face I guess deep face is another one but deep face lab is in my opinion currently the easiest to use the most accessible and what it does well is you don't need to have a very specific driver and dotnet framework and stuff and solid you can you can run that on any nvidia card and even if you don't have an nvidia card those versions you can run so the way all this technology works and I guess the reason it needs your GPU is it's basically you're going to feed your computer a bunch of images of a human face ideally of one human face and your computer is then going to basically break these apart in GPU and to like tonal regions and shape regions like that and just try to start understanding the pattern not really like doesn't know what a face is it's just going to learn patterns and the nice thing is is let's say so we did Tom Cruise you know I give it five thousand pictures of Tom Cruise's face that's how much you have to give it because you need to smile you need a frown you just smile from the left smile from the right you know all those different poses you can make with your face need that from all angles so you end up with a lot of images and you feed it into the computer and the computer starts looking at these images trying to learn them it's kind of like as if I were to look at Clint like just look at him and then turn away and try to draw a picture of him and at first glance I might look at him back okay he has two eyes one nose a mouth and be there not a beard I look back at Clinton like that was correct but I drew his eyes too big I drew his nose the wrong way I drew his mouth the wrong way or someone go back and like refine this a little bit okay let me look at Clint again by drawing again okay these changes i made made it closer to clint and that's one of the key things here is the computer needs to be able to take these synthesized images compared to the real thing and see how good it's doing and so like with the Tom Cruise faces you know look it takes one of the faces it's been going through them and kind of studying them and then tries to replicate it and it looks at the replicated face and it compares it to the ground truth as it's called the actual original face and it says well how close are they the things I did today does it does it match more but it's just doing a basic difference calculation right and if the answer is yes the things you did made it match more good keep doing those things if the things you did make it match less stop doing those things and it just does that over and over and over and over hundreds of thousand times users do like computers ooh and little by little it learns it learns what made it look more accurate versus what made it look less accurate so you do that for your actors face the face you're gonna create the mask out of you need to do a similar thing for your destination face the face are going to put the mask on just because the computer is going to need to understand okay the destination face smiling and I know what the actor smiling faces looking looking like now I can recognize what the destination face is looking like in terms of like oh it's smiling or it's looking to the left I know how to recreate that already so now I can put the two together and generate a new synthesized face and at the end it's like it's not actually like that crazy technology it's just an application of a thing that computers can do now these days which is Eva GPU which can do billions of calculations very quickly basic calculations but it can do those basic calculations so if you can break down a task into billions of basic calculations well then you can generate for example a face but you could also do with other things you can make a tree generator you can make a cloud generator you can make a image denoiser all these things kind of boil down to the same thing which is pattern recognition like for example someone did a image D noise or where all you had to do was taking it an image from a spot with the same camera take one without grain so like say do a long exposure at night on the street and then do the same picture with grain and then give that to computer the computer looks at it goes okay so this is what it looks like without grain and this is what it looks like with grain and I can use pattern recognition to understand that this grainy part should actually look like this not grainy part then you can grab a new grainy noisy picture from the same camera via the computer and goes oh yeah I I know it's supposed to look like without grain now you mean it's like so many plugins it's a little bit different in that you're training at first like a lot of plugins we use like Rd noises that we use are basically looking at the same frame or like a same spot of a frame over a couple different points in time to kind of figure out what it's supposed to be but what we're getting with AI now is that you're teaching the computer it's learning I mean that's why it's called artificial intelligence it's not just a plug-and-play math function it's an actual thing that changes based on what it's learned but you know we have to understand that learning to us to a human being is a vastly more complicated system of learning than it is to a computer you know learning to a computer is basic things like I'll do this routine I'll measure the output it has to be simple measureable outputs and then you can have it lean one way or the other I actually took so when I went to college I studied a little bit computer science and there's one class I really want to take which was introduction to artificial intelligence to take that class you had to take two semesters of calculus by that point I wasn't a computer science major anymore as an art major but I really wanted to take that AI class so I took two semesters in calculus so I could take that AI class and it's actually really really useful and really interesting at the end of the semester you had to do a project we had to make some sort of AI based thing some people made like a image recognition thing some guy made like a game where you could play Mario just by like moving your body you know it's all basic stuff works like the computer learns that you're facing it or you're leaning left or you're leaning right you know basic basic image functions and what we did is we made made the game Mont kala Mancala is kinda like an African version of like chess or checkers basically you have pockets with beads and then you're trying to lift up the number of beads and you place one in each pocket as you move towards the end pocket and you want to drop your last bead in the end pocket to get a point where we're ads like the most points the end of the game basically wins the nice thing about Mancala is that it's got the term here but it's basically it's a game that has very specific ranched functions or it's like you can either move this or you can move this and that leaves you with two new states and then from those new states you can do this or you can do this and that leaves you with two more new states so it's almost like a tree of decisions it's not like suddenly halfway through we like somebody drops a bomb and the board blows up and all the pieces get randomized you know it's a very specific decision tree and chess is the same way the problem with like trying to teach a computer chess we're like so the most basic form for an AI for that the computer does pro it just runs the decision trees deep soak in it's like okay if I did this move what happens okay if I did this what happens if I did this move what happens okay what if I did this move and then I did this move afterwards what happens the problem was like chess for example is you have like forget how many pieces you have it something like 24 pieces or something like that so you know what's 24 times 24 times 24 times 24 like suddenly it becomes an impossible number to calculate like three moves into chess for every piece it suddenly becomes something like a computer can't handle it's just too many decisions so you just start shaping it where it's like look you don't need to measure dumb decisions hey does this decision to get you're like Queen taken yeah don't continue down that decision tree so you start pruning your decision tree and these things you start I think the term for them is heuristics I could be slightly wrong but you basically start influencing it with kind of shaped things so this mahn call the game we made so I'm going with this is that we had a grad student on our team most most projects and in like school you don't get lucky enough to have like a grad student on your side but we had a grad student on our side so we built a basic mount caller game right where it's like you run the decision tree as deep as you can it was like you could go six decisions in it's good and you can add some things like Oh does this at the end of these six decisions you may not be at the end of you know you may not know who wins or loses they're less points you know so there's a way to measure a decision is it good or is it not good do you have more points or less points or are you set up now to do a couple more moves that will give you points like so you can measure these things but you don't quite know how to weight it right like how important is it to be strategically set up versus how important is it to have more points and so in this grad student being extra smart and knowing way more about this than any of us he sets up a an option here where we had this game and he hasn't run this game thousands of times on all the computers in the computer science building at night when no one's using them on this and basically what he did is he had it like a variable to each one of these values like so you have your do I end up with more points and he had a variable to it right that would weight it so the final thing would have a weight of like how important is it have more points how important is it be more set up how important is it to have like less pieces in these parts you know it's like and he added just like an XYZ variable to that and that would be randomly shifted randomly weighted and this is where the learning comes in because the computers not learning with it just does a decision tree that's basic logic anybody can do that you can do it with pen and paper the learning comes in where the computer looks at and goes like okay I'm gonna apply a value of 5% two points at some point to have points okay now I'm going to apply a value of 10% I'm gonna play a value of 15% which one of those is now going to give giving me the most victories Oh 5% is okay now I've learned it's only 5% important to consider points and that's that's what this guy did is he basically this is for all the variables for all the different things that you could possibly have in the program ran it on thousands of computers formula honed in the most optimized AI to play Mon call it and so a curly I think I still have it probably the best a I'm on kala ever made it can be any other boncala game we could find like I went online I tried any flash game we could find it could beat everything because I don't think anyone else has ever wasted as much time but that's when I understood learning in AI learning and that's not so much like how we view the world is a much more simple version of just repeating a task and then measuring the outcome and repeating the task and measuring out competing the task measured an outcome and you know that's kind of how it is with deep fakes you're just learning little patterns you're putting them down and you're repeating the task and measuring the outcome until you end up with something that looks like your ground truth so yeah it's it's crazy it's like at one hand it's complicated but in another hand it's so natural like we're so familiar with this world that like it's not complicated so how far are we from being able to do this like with the with have it be built into your OS on your phone I think well that's a good question because you have to train it first yeah now once you've trained it you could have the model built in like silks and right now we have a very good Tom Cruise deep fake model and I can grab a video view Jake and I get slapped Tom Cruise's facing you and it would look pretty good in about an hour versus like it took about four or five days to train the model on just in general but now that it's trained I could put it on you pretty quickly I would need to train it on your face so it knows where your eyes are and what your mouth looks like because maybe it thinks your moustache is your lips it doesn't know until it practices on your face a little bit right so theoretically it could just be built into your phone or your OS but it would need to be trained first I think that kind of stuff actually happens with Siri on your phone like Siri starts to tune herself I'm pretty sure I could be wrong but this is based on my experience series starts to attune herself or himself if you have it set to the guy to your voice so I can be next to my wife and I can say hey Siri and only my phone turns on and hers doesn't realize it's been trained over time little by little to understand the difference in our voices I mean we could try it right now that's got to be the case because when they first introduced Google home and stuff and when they first introduced they weren't that good no they weren't remember when we first got the it would like the the original goo home yeah barely do anything and then like a three months later we tried it again and it was way more advanced em hmm like wait a minute that's crazy let me do a little test here I have Sam's phone I have my phone do you say any of your hey Siri turned on mm-hmm okay hey Siri there it is my phone turned on Sam's didn't Siri turned on him I didn't so Siri has actually been trained to your voice a little by little it's just happening in the background but don't think that's something that happens on launch though I thought you give it like a little bit of a voiceprint right when you maybe I think I'm think so okay anytime you're doing it you're giving it again try and sound like Sam Clint say it will hold up both our phones you say it let's see if either referenced or not I don't know H or no no you just say to our phones I'll see if hey Siri hey Siri that's probably just what's worth might mean say we're saying two feet away from you we're Christians my room hey Siri hey Siri there goes Christians again jakester hey Siri hey Siri here's how I say I go I have a downward inflection hey Siri so try like that okay with that inflection hey Siri hey Siri just got mine that's crazy you might be reading it from an initial OS setup I never I never did a pattern ization no no I assignment yeah it's if you train you see I think you basically just go hey Siri hey Siri hey Siri like two or three times right when you started up and it takes that as the initial voice print probably I would imagine that it builds off of that but I'm okay I don't think it's automatic that's crazy there's definitely thurr way I think it's pretty clear that there's definitely deep learning yeah going on it's great I love it yeah it's gonna be very we were talking about how a perfect application for stuff like that is these these take down videos mm-hmm because right now Sam was saying how there's just so many people that are just physically watching these things and how draining it is to watch all this horrible content I mean the better that dpi gets that's that's a perfect application for it right there yes just I mean nope no nope did you watch the look on valley yeah yeah i've seen every episo– do you remember that scene where the guy makes the hot dog recognizer yeah like recognize hot dog hot dog look at all these pictures and yeah so I'm pretty excited about deep picnic ecology because at the same time people like act like deep fakes like it's gonna change the world politicians they're gonna like be framed for things it's like that's not quite how it works that's the dramatic version the framed for things the automatic version they're like slightly tweaking people's thoughts is the is that real that's what we were talking about is yes like you it can be used to embolden people's current beliefs rather than introduce them to new views I think I saw this video that reaffirms everything I've already believed about this person but it's fit but it's fake I don't care that it is I didn't take the time to figure that out I'm one of the 25% of people who just scrolled past it made my opinion and then moved on and now it's reinforcing my belief which is perpetuating me down a road of truth exactly and I think I think that's the key here actually I've never really thought of it this way but I I think it something just clicked for me I think it's that incremental change that in that the small nearly invisible incremental like steering of one's thoughts yeah it's it's almost not even it's like the grain of sand thing or it's like Chinese water torture version or droplet at a time ya know one droplet is it's the end of the world but yeah it's not like so for example you know you have a deep fake of President Trump saying we declare war on North Korea it's like you know everyone doesn't just take off in their planes you know what I mean that's not that dangerous to be honest like those huge big ones but it's that idea of small subtle changes you know the things it's the same thing that like all social networks right now we're facing when it comes to like how they handle content how they handle censorship how they handle who gets demonetized or flagged or not like that and like the whole social culture surrounding it so I think that's that's that's definitely the weirdest scariest part because it's a little it's kind of invisible still yeah it's like this invisible critical mass that it's sort of leading well eating to the you're at your example of we declare war right and then people from North Korea see that and they're like why what what and they're like maybe for a split second they're gearing their minds to like about time let's go and and then they find out Ono's fake its fake and they're like oh I was ready the thing is defects don't let you do that like deep faith I know I know but a mask I know they just the idea of that though it's like we're you in your mind you're ready to go for a second but then you realize it's faking like oh oh it's like yeah it's like you say but for something you're gonna buy a new car you save up and then it's like I was ready to buy it like oh but I'm not gonna buy a new car came out but like I was ready to buy it I might as well just do it I was ready I'm just I'm seeing so many people use deep fakes it's just a catch-all term for like faking somebody else yeah right I think it's it's just worth I like I'm a technical guy I like to be specific I wanted to get technical here deep fakes are only when you take somebody else's face basically generate a mask and put it onto a new person so like for example there's a you're going around of Obama saying basically they took a he was a key appeal I think it's Jordan Peele yeah but he doesn't know about my impression one of them it doesn't like an Obama impression and they had Obama actually saying it and I was like whoa look at this deep fake like that's not a deep fake as performance you're targeting if it was a deep fake you'd either put Jordan peels face on Obama or you put Obama's face on Jordan Peele it was easy still be Jordan peels balmy body or Obama's body was what the other person's face so this is not a deep fake that we're looking at here because that's per that's this called performance retargeting or they're making basically doing a two-dimensional morph so Obama so yeah that's but this is this is arguably going to be the thing that emboldened people more to believe what they currently believe even if it's wrong then actual deep fakes yes and this in so what's going on here is it's similar thing that it's like pattern recognition but you have this source of Obama and he's an always have imposes with all these different like mouth shapes right and as Jordan peel is doing his performance the computer is basically going in fighting the little snippets from Obama that matched that it's like okay here he's smiling or here his mouth as an au shape for example mm-hmm it's taking that just that little bit and using that little puzzle piece for that little spot and like and okay now his eyebrow he blinks here so it goes to the video looks for where Obama has blinked and some other thing uses that as a little puzzle piece for that little spot and it fakes are similar but this is this is literally just taking footage of Obama breaking it out into little puzzle pieces of everything and then reassembling those puzzle pieces based on a new pattern which is different than a deep fake which is basically taking out somebody's face effectively turning into puzzle pieces but then putting that on top of a new face after learning the structure of it there's another little clip those like these girls tearing up a gun gun range target and then someone's just like comp the Constitution on to it they're tearing that apart so that's one of the girls from the Florida shooting yeah so but that's what I'm saying though like how many people believed that that was the thing instead of look real I don't know how I did that but this is just after-effects compositing like this technology has been around for 20 years yeah it's like this is the thing where it's like this idea of incremental or it's like incremental control and incremental like steering so to speak where it's like you you already are far left or far right and you see something that supports your views and you you just kind of get that one extra bump in that direction it's not gonna it's not like gonna completely go oh my god president Trump laying around naked no way it's just never voting for him most important things he could be teaching in school now is critical thinking right yeah exactly like this this is the thing a captain disillusioned always talks about I know it like he's like he acknowledges that his videos are often too like nitpicky sometimes on VFX but he's really pointing out it's about skepticism and critical thought and I'm I'm honestly shocked and kind of saddened by how many people thought the Boston Dynamics video is real because it does not take that much critical thinking to figure out that it's not it might look a hundred percent convincing but the mole and some dude takes out a gun and start shooting it in the room if you still think it's fake by that time and now we're sorry if you still think it's real by the time after that scene finishes and you still think it's real you should really stop for a second and think about how how willing you are just take the world at face value because the world what you see is you really not what's going on and you need to be skeptical with your things you need to take time and apply critical thought because anybody can say anything just cuz it's written down doesn't mean it's true and if you're just seeing somebody shooting a robot oh yeah that's real you need to pause and think about the rest of your life and just suppose we all make that that all happens to us from time to time you know it's maybe it's like Saturday morning and you're hungover maybe it's like whatever you know like you're maybe you're at a baseball game maybe you're like at a party maybe you know there's so many different circumstances it's not like it's not like every moment of every day you're your most self-conscious like it's not like you're your best self every time you see a piece of information that's true you know and so there is a there is a responsibility I think on the part of like the major social media platforms to continue to improve the way that information gets presented to people right now clearly it's there's some there's some gaping holes and if if if it can be improved through like verified channels and the algorithm learning more and over time I think that's the only direction that we can go in I agree and once again YouTube's leading the way on that well I mean the thing is is that you go to a small but potentially a counterpoint to what you said of like oh you see a guy shooting a gun at a robot you should know this is fake like yes it's ezel if well it's it's also that is the way we designed the video actually which was to basically hold the viewers hands for every step of the way from when we start out trying to keep it unassuming just a robot next shot it's just a robot doing basic Boston Dynamics things like basically there's one extra smack with a hockey stick but that's already in Boston Dynamics videos and so each shot tries to like do this it because like the original idea was a the robot goes rogue and like shot – and the whole video is about it being Rogen oh we were like wait a second no like that should be like that the cherry on top is is like fighting back and going robe the rest of it completely unassuming and building to that just like with each piece perfectly interlocking it's like okay you can hit it with the boxy and hit with the chair you're gonna hit it with a bottle but at that point we've already smacked it with like a half dozen totally average things that when you break the bottle over it like maybe yeah if you're if you are paying attention that's the joke that is the joke mm-hmm but at the same time for us to keep it quote/unquote realistic it's just that puzzle piece over that building block so that by the time the gun comes out you're like like I yeah every step but there's a dove made sense and so now a belief retention that slowly goes down yeah over time yeah ideally ideal like that at least that's what the thought was when we wrote it was like this is funny like this is super funny the chair out of the corner [Laughter] with the chair yeah so I mean I'm yeah so I there's the thing is that we were kind of working against people's like I guess suspension of disbelief there's no not suspension of we're working towards their suspension of this below you know what I mean when you can it break like we're saying like yeah we're just trying not to go like think like too loudly you know and as we say fake we're just like thank you like the frog in boiling water starts off room-temperature terms of heat and the Frog doesn't realize it's boiling until it's too late which by the way actually doesn't happen frogs no it's hot jump out before they boil else have anything else to say about this before we close these matters before we adjourn critical thinking there we go thinking I think that's it that's a dope points a good point a dope last poll I just say if you guys are following the corridor channel they're going to see an exploration of this ideally in the next month or so and all these other videos were kind of like build up other things they should teach in school personal finance oh my god business and taxes business how do I communicate communications taxes personal finance do you guys think they should teach gun safety in school absolutely why not be more aware like do ya everything just be aware of everything why not I don't know if you can like give them guns right away but like start with like guns what do you know and maybe just like just basic guns they have a gun yeah I would admit there's things that are far more important than that and I wouldn't say like hey let's pass legislation to do that right over-over way more important things but if you're talking about an an ideal situation where you can teach anything and everything and it's it could be an elective mm-hmm like well I see we're saying I think the thing is not to interrupt but I think also like when you look at all right we're looking at a physical education we're looking at math or history or all these variants of that you know like math is the precursor to learning to do taxes still you know obviously there's more advanced versions and reading and English and things like that in physical education being a precursor to learning how to shoot guns I guess this more of a machine shop kind of class that's more that's better that's a better precursor that in chemistry are no physics and chemistry are precursors to gun safety one of the greatest threats to the world is fear and I think that absolutely Will Smith one of the greatest one of the greatest problems just as Will Smith what I mean we're friends now I've gone through wills Instagram yesterday like oh well I said I said wills and then I think I like restarted Will Smith yeah do you think his friends call him bill hey Bill yeah billion billion but anyway what you're saying is one of the greatest threats is fear and I think that absolutely applies in a number of circumstances and I think firearms is just another category in that yeah just be aware of everything just you where the other thing too is like we're this wasn't like it this when we were in high school but at this point now the internet has changed in a way where everybody's everybody's just constantly regurgitating information because that's how you yes how you're ranked on Google like if you're are news articles new we're gonna get views versus if it's old so what happens is every just in this churn of constantly rewriting stuff and things aren't really researched anymore it's just it's literally some 20 year old kid intern grabs an old article bangs out another one by adding some extra words and then post it and that happens here again later it happens a year later happens you later and we're at this point now where there's like there's a constant churn of fluff online where people are just constantly writing stuff like I seem like when people post about our videos for example they've never called us they never did any research they just like flesh out two or three just fluff paragraphs before before CNN 24-hour news didn't exist not a single wasn't it was a concept it wasn't like hey where can I constantly find news information it's like well before that people would go well there's no new information to talk about so wait until this time yeah that's true not a single person not us and not a single article that was written about the Boston Dynamics video which there's a couple dozen of them at this point none of them reached out for any further information or anything actually from the source which is very interesting so basically everything they said was basically pure speculation they just thought it was that Russian guy pure speculation or an obvious like lot just obvious facts that are then like there are no word it out to sound more robust I think appointed making that was like there's no authority and information anymore the only authority is your own critical thinking Wow alright that's it guys information but you're right it's not the way that it used to be for sure I mean we say there's a thorough information you're right this summer three but only if like only if the actual sources and data are there like right Spanish like the main Spanish news networks thought the Boston Dynamics video was real the universe other no not not Mexican like Spain Spain like didn't even bother to read boss town dynamics in the corner that even catch that they thought was a Boston the behind the scenes was on the news they put they posted the actual video and then you're talking about in the middle the day saying like Boston Dynamics is abusing robots in their laboratory something you're just sitting there like I got a bang on a store you'll quick talk about this and like drum up some anti robots and that's those were like the guardians and information twenty years ago and that's what you're getting now is people just it's just a YouTube channel at this point people just churning out content as fast as they can like it's funny that's not funny that's hilarious it's funny but it's also sad do you guys want to have a positive takeaway from this podcast it's that the word heuristic means enabling a person to discover or learn something for themselves Hey you are i FDIC are signing off [Laughter]

41 thoughts on “EP#26 | The Importance of Critical Thinking in the 2020s (Bosstown Dynamics, Deepfake Technology)”

  1. About deep fakes and whether it is more difficult to puppet the face of a famous person than transplant a famous person's face onto an actor. In terms of the machine learning algorithms, I think it is actually easier to puppet a face. Also voice synthesis is totally a thing people can do.

  2. I know it jeapordizes the overall immersion and artistic vision of a piece, but you guys should REALLY put baked in watermarks in all your stuff from now on. I think it's at the point now where you're too talented and too hardworking to not get the recognition you deserve

  3. If you guys ever put the robot-shooting range video out, please put a thick corridor logo on it, for your own sake

  4. Great episode. It blows my mind someone can think your video was real. I watched it on your channel and I am following you for years but still…it's not like you have tried to hide the fact it's made up.

  5. People who repost other peoples content to make a buck like that honestly deserve to have their router ripped out the wall. Its one thing that pisses me off about the internet.

  6. Deepfakes become more scary when you arent a celebrity. When you have a chance to explain yourself to millions, you will probably get your voice heard in the end. Imagine if you had 250 facebook friends and thats it, your porn comes out and gets published on your open wall as post. Do those 250 people care to give a opinion or are they just move away because they dont know enough about you. How do you explain yourself? there isnt going to a newsource in the end which you can quote to say thats not me, when you are a nobody, you dont get that, you just get ruined

  7. Another great Cast you guys. You and your talks seem to always reflect a part of myself. Knowledge is Power, many animals are stronger than us, yet we dominate the planet. Why? Knowledge! That is why it's scary to know that data mining by AI controlled by evil intended companies is here and has been for a long time. But again as you said, we have a heuristic ability to gain more knowledge ourselves. The poison is also the antidote. Teaching more people basic things can have a profound impact (aka balancing a checkbook, compounding interest, churning, gun safety, critical thinking…). Everyone needs to embrace the scientific method of thinking and acting. Can't wait for the next cast!

  8. You should have forwarded that email to the dude behind the camera to blur the sensitive info or something. Would have like to see it lol

  9. I mentioned to a friend of mine that it says "Bosstown" but he just thought that it is a parody by Boston Dynamics themselves

  10. Oh geez, the talk about Siri made my iPad constantly open Siri (apparently she’s not trained to my voice yet).

    Also its kinda interesting that all the german articles about the Bosstown dynamics video all say that its a video made by Corridor Crew. Apparently the first one who wrote the article wrote that and every other site just copied that from them

  11. My new favorite podcast. I really enjoy the ones without the guest and just talk about issues pertaining to vfx. I hope you guys keep doing it!

  12. Things are almost getting to the point where someone could fake landing on the moon or launching a car into the earth's orbit.

  13. 1:12:46 Jake, I completely agree with Will's point here, so we needn't be afraid to learn anything, including gun safety and usage. But Guns themselves are a product of fear. They were invented because people were scared, so they made this dangerous weapons so they didn't have to be scared anymore. I think anything which has spawned from fear, never leads to anything good. Makes sense?

  14. You guys need to do clips of this podcast! I’d love to listen to stuff but I don’t ha e the time for full podcast

  15. Yes! I always thought it was important for people to critically think instead being like mindless zombros, and they made a video. XD

  16. I'm really with Niko on the point he made about it not being very difficult to discern whether things are fake in videos if one uses critical thinking. I might be biased because I am a film student and scrutinizing video and audio is a LOT of what I do, but even before becoming a film student I felt like I was able to easily figure out whether something seemed too fishy to be true. If something seems WILDLY questionable or too good to be true then it isn't difficult to pause, take a moment and analyze whether it is real or not. The fact that A LOT of people completely missed the "Bosstown Dynamics" and legitimately thought it was from Boston Dynamics are the perfect example of this.

  17. The thing about "the russian video" is from the fact, that almost every Boston Dynamics video gets reuploaded with google translate text complaining about "skin bags", or something like that. And I do find them pretty fun, for example. Way funnier than people just beating the robot and, basically, obstracting him from his goal. o,o

  18. That video looked absolutely amazing. I can never get a 3d object to match my lighting like you did. However, I could tell it was fake in a millisecond. Watch it on a TV and you can see that it's a render, same as the Tom Cruise video. But if you lower the quality to something under 720p, it get extremely realistic.

  19. Reminds me of the promotion for Planet of the Apes, where the CGI chimpanzee stole an AK47 and started firing back at people bullying him 😀
    People shared it like "yeah good thing fuck those people" ahahah

  20. Y'all should really check out the series that Destin from SmarterEveryDay did on algorithm manipulation. Really eye opening stuff and incredibly well done.

  21. To Jake's question about whether it's possible to have machine learning on our phones: anything is possible with cloud computing. Someone (google for example) provides a bunch of GPU's as a service, and now you can send your face from your phone and some Google server merges it with Tom Cruise.

  22. Funny you guys are talking about deepfakes, I made a short film about a rising musician who is a victim of a deepfake'd video of him cheating on his fiance. And now that this could literally happen to anyone, and that people can be fooled by it so easily (see the reception of the Bosstown Dynamics vid). …Next year is going to be terrifying.

  23. The real question is if Boston Dynamics has invited yall to come out and actually see the real robot and maybe do a video with them, that would be great!

Leave a Reply

Your email address will not be published. Required fields are marked *