Latest / Newcomer Pod / War on Reality: AI Voice and Video
Transcript
- 0:00This episode is brought to you by Forethought.
- 0:03Most companies build their customer experience in pieces.
- 0:06Sales in one system, support in another, and onboarding
- 0:09somewhere else. Forethought brings it all
- 0:11together. Forethought is an AI system
- 0:14made-up of advanced agents that handle sales, onboarding,
- 0:17support, and retention. Each team manages its own
- 0:21agents. The customer sees one unified
- 0:23experience. Forethought powers over a
- 0:26billion interactions every month for brands like Scale AI,
- 0:29Cohere, Air Table, and Upwork. Learn more at Forethought dot
- 0:34AI. Hey everybody, welcome to the
- 0:43Cerebral Valley Podcast on the Newcomer newsletter.
- 0:46We're getting ready for the Cerebral Valley AI Summit in
- 0:49London on June 25th. Thrilled to be back with Max
- 0:53Child. Hello, glad to be back.
- 0:55And James? Wilsterman.
- 0:57Hey, Eric. Good to see you.
- 0:59We are talking about Max and James's favorite topic, voice
- 1:03and the future of video. Though I guess you guys are
- 1:06getting in. Both.
- 1:08You got to get the branding right.
- 1:09It's voice AI. OK, now voice AI and then video.
- 1:14Eat it at you. I was like, I forget what we're
- 1:15fighting about, but I was like, I noticed you add the AI to the
- 1:19company. Yes, that's correct.
- 1:20We were pioneers in the field of voice AI, and I've added myself
- 1:25to the Wikipedia all right. We're in the Hayes Valley
- 1:29Wikipedia. Do you know that?
- 1:30I've seen this. Yes, yes.
- 1:32OK, what? About our first accomplishments
- 1:35clearly you know VO 3 is crazy like video creation has gone
- 1:40insane. We want to cover a couple things
- 1:42with this podcast where we think video is going and then we're
- 1:47going to review some videos. We just had a lot of technical
- 1:51difficulties trying to introduce AAI video Co host.
- 1:56It seems like the problem was that fundamentally ChatGPT is
- 2:00down because they introduce new models, which is honestly the
- 2:02most disappointing failure mode for having a AI video Co host.
- 2:08Since just ChatGPT working, it feels like it should be the
- 2:11minimum. What are they spending the money
- 2:12on? I don't know.
- 2:14Apparently not server uptime. This has happened to us pretty
- 2:17often by the way, we run a lot of our games on GPT API's and
- 2:20it's, you know, every every hundred days or so you get, you
- 2:23get nailed by this. They just dropped the prices,
- 2:25right, 80% on O3, yeah. Does that affect you guys?
- 2:29O3 is five times cheaper. I don't think we run any of our
- 2:31games on O3 right now, but we do run our games on the real time
- 2:34voice API, which is I think what we're here to talk about today.
- 2:38The Her, the Samantha, Her model, as we called it a year
- 2:42ago before Scarlett Johansson sued Open AI All.
- 2:46Right. So a direct question.
- 2:48Where do you think video is in five years?
- 2:52We love predictions on this podcast.
- 2:55And I think specifically I want you to get to, do you think
- 2:58TikTok or movie studios like loom larger in this world?
- 3:03And do you think AI is intertwined with human actors or
- 3:09do you think we're pure AI movies?
- 3:13There's a lot of layered questions there.
- 3:15I think I would first say that anyone who's seen this sort of
- 3:17viral Google VO3 model output, you know, on Twitter, which has
- 3:22been very popular over the last week or two.
- 3:23You know, there's a new viral sort of version of this video
- 3:27content where it's a day in the life of a Star Wars stormtrooper
- 3:31and they're making little joke logs of how they've crashed on
- 3:35Tatooine and that kind of thing. And you know it, it looks, I
- 3:39think as good or better than many scenes in certainly the
- 3:43television versions of the Star Wars universe.
- 3:45I guess you could argue maybe the film looks slightly better
- 3:47if you're on a 70 foot screen in a theater.
- 3:50But you know, it's it's close enough, in my opinion, to sort
- 3:53of cut to the chase where you're like, this could be a part of
- 3:56certainly like a Star Wars miniseries or something like
- 3:58this. And so it seems hard to believe
- 4:01that YouTube slash TikTok isn't going to take over this world.
- 4:05I don't know how much of this type of content you'll see on
- 4:07the movie theater screen. But I think what's more
- 4:09interesting is like, given that you now have kind of on demand
- 4:13video that's of the quality of, you know, Hollywood, Does
- 4:17everyone get a personalized TikTok feed of their own AI
- 4:20generated videos that are perfectly targeted at them and
- 4:24and perfectly dispensed like their little, you know, their
- 4:26little piece of candy that that they love.
- 4:29I feel like that's one of the key for you're betting on long
- 4:32tail custom video. I'm betting on ultra long tail.
- 4:35I'm betting on like the barbell curve or what, you know, EU
- 4:38curve, right Where on one end we'll have like monoculture,
- 4:41we'll still probably have like, you know, the Taylor Swifts and
- 4:44the sort of ultra popular films, you know, Barbie Oppenheimer or
- 4:47whatever. And then most content
- 4:50consumption will be super long tail, ultra personalized.
- 4:54Everyone gets their own TikTok feed that's like AI generated
- 4:56exactly for them. I think like, everything in the
- 4:59middle is going to have a lot of trouble.
- 5:00Like like our podcast, you know, until the middle.
- 5:04Is done. Yeah, we're, we're probably in
- 5:06the middle right now. So we need to get our AI Co
- 5:07host. Taking over.
- 5:08All right, James, what's your prediction here?
- 5:10Well, I think that maybe podcasts will be safe for a
- 5:13while. I think it seems hard to imagine
- 5:16that we will have a big market for people listening to podcasts
- 5:21purely hosted by AIS for a long time.
- 5:23Because the reasoning through line isn't there just because
- 5:26they're just gonna veer off. Yeah, I guess like I think
- 5:30people mainly listen for the the hosts and the the sort of
- 5:34personalities of the hosts and the relation to the real world.
- 5:38You're like, we're so charming, how could it?
- 5:40No, I don't know. I'm so charming.
- 5:42If we're talking five years, I mean, I guess any, anything's
- 5:45possible. Five years.
- 5:46Anything's possible. The Silicon Valley point of
- 5:48view, Yeah. I mean, I just think that
- 5:50podcasts in general are just not that big of a business.
- 5:52I mean, it's especially if we get self driving cars, I don't
- 5:54think people are going to listen to that many podcasts in
- 5:56general. I mean, I think you already see
- 5:58video eating podcasts. Like, Oh yeah, yeah.
- 6:00Well then I'm saying we're a. Video podcast.
- 6:02I sure can. I don't think there's a
- 6:04hardline. OK, well I guess I'm saying like
- 6:06the concept of a talk show or whatever, right?
- 6:08A long form talk show, which is sort of what James is talking
- 6:10about I think. I guess I really disagree.
- 6:12I think it will totally get eaten by AI generated content.
- 6:15I listen to the Bill Simmons podcast.
- 6:16I feel like the personal relationship I have with Bill
- 6:18Simmons is, is of course, like, you know, meaningful because I
- 6:21like that podcast more than other podcasts.
- 6:23But I totally believe there's an AI out there that could create a
- 6:27better show that I would like more.
- 6:28And then I wouldn't care if it was a human or not.
- 6:30Like there's nothing quintessentially human about the
- 6:33Bill Simmons podcast that's like keeping me around.
- 6:35I guess maybe maybe I'm underestimating Bill Simmons
- 6:37abilities. It's possible.
- 6:39Would you predict that the amount of content generated by
- 6:43AI podcasts makes it uncommon to ever listen to a human generated
- 6:47podcast in five years? That's what your sort of
- 6:49prediction is, or would you go that far?
- 6:50More airtime for human podcasts or AI podcasts in five years.
- 6:55I say human. I'll say AI if we're including
- 6:57all video like talk shows or whatever.
- 6:59James, you're, you're on the hardcore human side, right?
- 7:02You sure? My, my prediction is based on
- 7:04like podcast players as they are today, like, you know, whether
- 7:08that's YouTube or, or this. Is going with some kind of edge
- 7:13gears here I think in Apple Podcasts and Overcast.
- 7:21I do believe Max is on the right track here when we're thinking
- 7:24about short form video, when we're thinking of whether even
- 7:28Hollywood video will be mostly AI generated like anything
- 7:32fictional where you're already looking for kind of a fantasy
- 7:36type experience or world like, then obviously AI is just a
- 7:40faster path to generate that. You can still have human written
- 7:43scripts involved. I mean, sure, there will that
- 7:46will be AI impacted as well. But yeah, I think that would be
- 7:49where I would see this changing faster.
- 7:51All right, Eric, what's your thoughts on this podcast
- 7:53situation? I think what video hasn't had is
- 7:56fast iteration. And so I think the biggest, I
- 7:59mean, there are lots of things that are going to be interesting
- 8:01about AI video, but on this sort of like long tail of videos, but
- 8:04even popular videos like you get on TikTok, AI generates a short
- 8:08video, sees how it how it's doing and then sort of dedicates
- 8:12itself to videos that are getting heat.
- 8:14Like I think this sort of just AB testing a video and seeing
- 8:18what's working, what's interesting.
- 8:20I think that's going to be a path for just like novel video
- 8:23creation and is going to be really powerful.
- 8:26So to me, I think TikTok and YouTube are much more important
- 8:30than movie studios for video five years from now.
- 8:34I also think the movie business, it's going to be IP wars.
- 8:38You know, it's like we Studio Ghibli is like the top thing on
- 8:42opening high. Like I just think that if you
- 8:44actually try to make a movie that went to theaters about it,
- 8:47you'd have a lawsuit. And I think the movie studios
- 8:51with the actual IP have all these employees that are really
- 8:53hostile to using the AI to generate their IP.
- 8:58So I think largely YouTube, TikTok and this sort of user
- 9:03generated type stuff is just going to be AI generated.
- 9:06And so we're going to see a lot of the competition come there.
- 9:09Yeah, That sort of gets into a question I have about the
- 9:13future, which is given that the training data sets of all of
- 9:17these AI is sort of where we are today or where we are up through
- 9:212020. And also given that, you know,
- 9:24the only sort of rallying points for monoculture are sort of
- 9:27nostalgia and IP. To your point, like are we sort
- 9:30of in the like the end times for new content in some ways where
- 9:34like forever going forward, like just the content universe of the
- 9:3820 twenties will extend permanently into the future
- 9:41because it'll be so deeply embedded in the training data of
- 9:45the models that like sort of became, you know, became AGI or
- 9:49whatever you want to call it that like, you know, the 1990s
- 9:52and 2000s will be like weirdly important for the next 200
- 9:55years. Because it's already pre
- 9:57original. Content on the.
- 9:59Pre AI, that the Internet had killed sort of generations or
- 10:03that it was so self aware that you didn't have this.
- 10:06Like we all dress this way until we all, you know, Yeah.
- 10:09And so that banding was already sort of on the way out.
- 10:12We see with new technologies does open up the opportunity for
- 10:15new creative IP, right? I mean, Mr. Beast didn't exist
- 10:18before YouTube and you know, Mr. Beast, the ultimate derivative.
- 10:22He's doing The Hunger Games like.
- 10:26I've kind of I've got it with Eric here.
- 10:27I think Mr. Beast is gonna not not really last the test of
- 10:31time, but. Sure.
- 10:32But I think like obviously that that era of the Internet and
- 10:35YouTube like created new stars that wouldn't have existed
- 10:39without the Internet, right? So I guess I could imagine
- 10:42something similar happening with AI where there's new types of
- 10:45musicians, right, that are fully invested in AI.
- 10:48There's new types of film makers and directors that are like go
- 10:51100% in on AI, maybe new podcasts.
- 10:55I mean, right now we're constrained in a couple ways,
- 10:58like one VO3, what it's like 8 seconds at a time.
- 11:02So people and humans end up having to do a lot of the work
- 11:06because they're connecting all these different sequences if
- 11:09they want to do it. So we're not even really seeing
- 11:11true like AI creation right at scale.
- 11:16And then the other thing, Sam Altman just wrote a new blog
- 11:20post and I think he predicted that innovation or novel novelty
- 11:25was like an upcoming feature of AI.
- 11:28And so like in our five year window, I have a lot of optimism
- 11:31that novelty could be coming. I think it's the big question.
- 11:35And obviously if there's novelty, then there's new IP.
- 11:39And then we're like, wow, they came up with the new Disney.
- 11:41Like this is perfect. You came up with Pokémon in 20
- 11:44seconds. Like that would then blow us up.
- 11:46Then we're like every movie, it's like we get high on it.
- 11:49I mean, we could even have this period where it's like we're so
- 11:52in inundated with it and then we sort of pull back and we're
- 11:54like, oh, we're fatigued of the AI movies.
- 11:57That starts to ask the question of like, does it become a
- 12:00monoculture or is super hyper targeted and everyone has a
- 12:04different Pokémon that they're watching, right?
- 12:06I at the end of the day, I know this makes me an old, but like
- 12:10I, I, I watch TikTok, my wife watches TikTok.
- 12:12We're lying in bed. We think we're so unique.
- 12:14The TikTok comes up and it's like, oh, it understands me so
- 12:17well. Then she gets the same one.
- 12:18Like I, I think sometimes the Super niche tik Toks are like,
- 12:22actually, no, there's a lot of common humanity.
- 12:24Like we're all interested in the same things.
- 12:26Like I, I think some of the long tail stuff is overrated.
- 12:30Obviously we have hobbies and it's like, you know, chess or
- 12:33League of Legends or whatever, so I'm not denying that, but I
- 12:36think there's always going to be this urge for a shared culture.
- 12:39Yeah, I think you're probably right.
- 12:41I think the ultimate question is how many different human
- 12:44interests really are there? And like especially ones with
- 12:47any sort of density, like more than 1000 people share that
- 12:50interest or more than 10,000 people share that interest.
- 12:52And there's probably not that many, maybe on the order of
- 12:54hundreds or thousands. So ultimately there might be
- 12:57just a convergence around that and maybe the tail isn't that
- 13:00long. But I do think like Drake
- 13:02Apathy, who made like the original vibe coding tweet, had
- 13:05an interesting tweet this week about VO3 where he said this
- 13:09will be the first time in history that video is what he
- 13:12called like directly optimizable.
- 13:13Meaning like you could just run like millions and millions of
- 13:18iterations on the same video, you know, and keep tweaking it
- 13:22for almost no cost. Whereas historically tweaking
- 13:25video has been sort of prohibitively expensive to, to
- 13:28do more than a few variations or more than a few kind of
- 13:30iterations and tests. And so video is going to become
- 13:34a lot more like, you know, text or something like that, where
- 13:37you can iterate and test it so rapidly that you can converge on
- 13:41sort of the like optimal solution for a given problem,
- 13:44even if like the optimal solution is just like a very
- 13:47particular version of a meme or whatever.
- 13:49And so it'll be interesting to see what those shared pieces of
- 13:52content look like in the future and what the sort of hyper
- 13:55optimized version of something that that was, you know, merely
- 13:58popular when we were he's. Saying it's cheaper, not yet
- 14:01like there's going to be a lot of costs.
- 14:03Like I do think another reason to be bearish on hyper
- 14:07individualized video in the next 5 years is just everybody's
- 14:11like, oh, wave away costs, wave away costs.
- 14:13Actually, it's like make one expensive AI video for everybody
- 14:18in the beginning and then once we get costs down, you know?
- 14:21Yeah, for sure. I do think five years is such a
- 14:24long time in this AI era that we live through though, that it's
- 14:28really hard to predict where costs end up for anything.
- 14:31I mean, like we just said, Open AI reduced costs for O3 today,
- 14:3480% down from what it was yesterday.
- 14:37So I think things just change dramatically.
- 14:40I think there will be a successful Pixar style AI
- 14:45company. Like I think if we look at Mid
- 14:47Journey, part of why Mid Journey has been successful is it did
- 14:49sort of it's like, OK, we like this design aesthetic.
- 14:52They're sort of playing kingmaker on certain looks to
- 14:54make it work. And so I think there will be
- 14:57sort of like AI studios basically trying to make movies.
- 15:00Given that your your wife is in the filmmaking industry, like
- 15:04what is your sense on is Hollywood going to adopt these
- 15:09tools like for, you know, special effects or for like
- 15:12certain shots or to like edit certain things?
- 15:14Like how much blowback is there against it versus how much
- 15:17excitement? My life is in documentary.
- 15:20I think she's a little more open to it, given what I do.
- 15:23But certainly in those circles you can feel like they're way,
- 15:26you know, go to their parties. They're way more hostile to AI
- 15:29than be unfathomable to us. So yeah, there's definitely a
- 15:33lot of hostility in Hollywood. I mean, I, Tom Doton has been
- 15:37doing some reporting for newcomer and I don't know my my
- 15:40sense from his story is that it's like Hollywood secretly
- 15:44the, you know, the execs are gonna use it and it's gonna
- 15:46happen and and there will be this sort of clash with the rank
- 15:49and file. I mean, documentary, given it's
- 15:52about the real world, you know, has its own reasons to be sort
- 15:56of not as AI generated, but it'll continue to create sort of
- 15:59this blur between real and fiction.
- 16:01And I mean, we kind of talked about this earlier, but like, I
- 16:04already think a lot of this AI video looks more real than like
- 16:07an effect shot in a Marvel movie.
- 16:09And I think Hollywood obviously moved in the direction from
- 16:12practical effects in the, you know, seventies, 80s and 90s to
- 16:15100% CGI for the most popular movies today.
- 16:18And I just think that it's sort of inconceivable that you
- 16:21wouldn't take this technology that lets you create at least
- 16:25like effect shots for like 1-1 thousandth as much money, like
- 16:291100 thousandth as much money and like probably looks better
- 16:34already in 20. 25 to control the camera right?
- 16:37Like. There are people working on that
- 16:38for sure. We're really close, I guess.
- 16:41I have a question. I just think that we're dancing
- 16:42around that we haven't gotten to you, which is just like we're
- 16:46predicting that all this video is going to be generated with
- 16:49AI. We're predicting that Hollywood
- 16:50studios will adopt it for special effects.
- 16:54I'm still kind of unclear what will the human role in any video
- 16:58production be? Will it be writing?
- 17:00Will it be acting? You know, someone's likeness,
- 17:03their voice? Will there be anywhere?
- 17:05Humans part of the process. I think storyboarding will.
- 17:09I mean, I think the the script might hold on the longest or
- 17:14like ensuring coherence across the script.
- 17:18I think that's where I'm at. James and I have been making a
- 17:21game, and definitely today the script is the thing that is the
- 17:25hardest to get the AI to do a good job on.
- 17:27Like we're doing an opening cinematic scene for our game and
- 17:30James and our content team have devoted weeks and weeks and
- 17:33weeks to iterating on the script and reshooting the script and
- 17:35everything. So it does feel like the current
- 17:37thing that's not good. But you know, if you throw the
- 17:40script into Claude 4 or 03 and say, hey, like, clean up the
- 17:44script and like, give me a better version of it, like it
- 17:47gives pretty good notes and pretty good feedback.
- 17:49So maybe it can't hold a coherent script across 120 pages
- 17:53today, but it's doing pretty well at a six page script right
- 17:56now in terms of at least like editing and understanding the
- 17:59narrative arc and everything. So I just and does it have like
- 18:02a target offs and stuff later and like the sort of classic
- 18:05stuff? You can ask it for pay offs.
- 18:07I mean, yeah. It's kind of like you, You as a
- 18:09human have to know what you're looking for, and then it gives
- 18:12you good ideas for these things. But yeah, it can't do everything
- 18:15itself. But I do think that's key to it.
- 18:18Like you're like, oh, does it have a key, the key element of
- 18:20story that like there's some emotional payoff or something.
- 18:24You're like, oh, hand wave that away.
- 18:26The humans will make sure like that's that's what I'm saying.
- 18:28So which? Is why short form video maybe?
- 18:30You're saying it's like a script producer or something?
- 18:33Or it's like the editor of the script rather than the writing.
- 18:36I don't know, a year ago it was really, really effing bad at
- 18:39this. And now it's like getting kind
- 18:41of good at it. And so you're like, I see where
- 18:43the curve is headed on this. I don't know.
- 18:45I think, I think a year from today it could be pretty good.
- 18:48I have a prediction that within five years video games will be
- 18:53completely AI generated at the pixel level, effectively using
- 18:58video generation models that are interactive as opposed to
- 19:02current strategies to render in a game engine 3D environments.
- 19:06Just for context, the way most games are made today is inside a
- 19:09game engine. You put a bunch of shapes in an
- 19:11environment and walls and light sources, and then you try to
- 19:14figure out how the light bounces off various things and what
- 19:17camera the player's looking out of and such.
- 19:21You're saying the future is we're basically just gonna take
- 19:24VO 3 but like, it generates the whole environment in real time
- 19:28all the time. And so instead of watching a
- 19:30little 8 second clip of something funny, you can walk
- 19:34around the world that VO 3 has generated for you and it'll all
- 19:36just change as you're walking. Around doing this with Minecraft
- 19:40already right? I mean, I've seen weird demos of
- 19:42it, but it doesn't seem real yet.
- 19:43Yeah, they're, they're, they're doing demos where there's
- 19:47Minecraft levels being generated in front of you as.
- 19:50There's anything we know about demos that every tech demo that
- 19:53has ever been given has turned into a real product that someone
- 19:56can deliver. I don't think it'll replace
- 19:58everything. I think this is the classic
- 20:00like. All or nothing, false dichotomy.
- 20:03I mean, I think it'll be interesting.
- 20:04Doesn't seem like the Switch just came out first of all, like
- 20:07Nintendo is like building games on like 5 generations old
- 20:11technology. It seems like it's not even like
- 20:13OLED on the device. So I just think you're like
- 20:16overestimating how quickly an industry that's not AI will come
- 20:20for it like new. I think new companies
- 20:23fundamentally and new studios will have to create these games
- 20:25and they will have to disrupt the existing studios.
- 20:28I don't think the existing studios will necessarily embrace
- 20:31it quickly. And the existing studios have IP
- 20:34and so it's going to be a slow transition.
- 20:36Five years is really fast for this kind of thing.
- 20:38To Eric's point, just from like a business model and technology
- 20:41perspective, I also just think that I've seen all those demos
- 20:43that you've seen of people generating Minecraft with video,
- 20:45and like, it's not clear to me what they're better at, right?
- 20:48OK. They're more expensive, they're
- 20:50slower, the world looks like crap, the walls change all the
- 20:53time. Like, you know, there's no other
- 20:55players in there. There's no game mechanics like
- 20:57they're worse at everything, right?
- 20:59OK, that doesn't mean someday they won't be better at
- 21:00something, but I've never like heard the pitch for why it's a
- 21:05better way to build a game world in.
- 21:07You can do anything. You could do anything cheaper.
- 21:09You can prototype everything you know in seconds, you can have
- 21:13any game you want, right? But.
- 21:14You can do anything in a game engine.
- 21:15You can drop whatever shape you want in the game engine.
- 21:17You can make any character Sprite.
- 21:18I mean, maybe it's easier and faster someday, you know, but
- 21:21it's also extremely non deterministic, right?
- 21:24Like, you know you, you're like, hey, I want a wall there.
- 21:27And then sometimes the wall moves or whatever.
- 21:29I don't know. Again, I believe there is a
- 21:30world in which they're better at some things, but I've never
- 21:33heard which game you would want to build in this thing, even as
- 21:35sort of an imaginary awesome version of it.
- 21:38It might come back to this like long tail argument where you can
- 21:40make custom worlds for every player and somehow that's more
- 21:42interesting than, you know, a more structured game world that
- 21:45you know, everyone was in the same one, I mean.
- 21:47Fortnite is a huge game. Even if there is one game that
- 21:51is made this way, that is like, sure it's the everything game
- 21:54because it changes, you know, as you develop.
- 21:57As as you go through it through. It you know, but that's a much
- 22:00smaller claim. There will be 1 super intense
- 22:03game where you can do this. Or maybe there will be a whole
- 22:05category? I just think if it's technically
- 22:07possible, well, it seems like it will be better than waiting, you
- 22:11know, five years to build a new GTA or something, that it'll
- 22:15just be kind of laughable that you would invest in a studio to
- 22:19do that. I don't fully agree, but one
- 22:21thing I will say is like, and this isn't games purely, but if
- 22:25you like think about Harry Potter, right?
- 22:26And they're psyching themselves up for this like series on Harry
- 22:30Potter. But there's so many assets fans
- 22:33have already created so much. It feels like given a little bit
- 22:36of AI juice, like the fans are ready just to create their
- 22:40preferred version of Harry Potter and all the like.
- 22:42Fandom can fight with itself about exactly like how everybody
- 22:46should be portrayed. And I don't know, I think that's
- 22:49almost more exciting. I think this Harry Potter series
- 22:51is coming at the worst first time.
- 22:53Like this is the time where the fans are about to be empowered
- 22:56to really make a full length series themselves.
- 22:59And I'm super excited about that type of thing.
- 23:02Let's let's dig into some actual real videos instead of just
- 23:07imagining the world that could be, the world that is the insane
- 23:11world that is. Let's pull up some videos and
- 23:13get your reactions. All right, we're watching an AI
- 23:20video from seven years ago. You won't believe what Obama
- 23:25says in this video, Wink. We're entering an era in which
- 23:31our enemies can make it look like anyone is saying anything
- 23:34at any point in time, even if they would never say those
- 23:37things. Moving forward, we need to be
- 23:39more vigilant with what we trust from the Internet.
- 23:42That's a time when we need to rely on trusted news sources.
- 23:46My first reaction was this was better than I thought it would
- 23:48be. Better than I thought it would
- 23:49be from 2018. Yeah.
- 23:51It had that effect of the the mouth moving off to the side
- 23:54because it basically was like someone cut a hole in Obama's
- 23:56mouth and then just like put someone else's mouth in the
- 23:58video. I think that's kind.
- 24:00Of what you figure out, it's a voice actor.
- 24:02You're like, OK, that's not AI so.
- 24:04So the TTS voice wasn't there. I mean, it was, it was pretty
- 24:06good. It comes back to our point about
- 24:08the the monoculture versus the long tail here, which is Obama
- 24:12obviously very monoculture figure It was worth the
- 24:15investment to cut a little hole in Obama's face and have him
- 24:18make silly things because you knew that was going to go viral.
- 24:21Whereas in the modern era we can.
- 24:23Spend a ton of money on this video.
- 24:25More than more than you would think, probably.
- 24:28All right, Next. Oh, that's hot.
- 24:35That's hot. Uncle Phil, come try this fresh
- 24:40pasta of belly. Why do we think that this
- 24:47particular Will Smith eating pasta video went so viral in
- 24:502023? I know now it seems like who
- 24:53cares? I mean, this is definitely the
- 24:54argument for like. Wow, things have really
- 24:57improved. This is a horrifying and
- 24:58disgusting video and I think people all enjoyed sharing that.
- 25:02Look how terrible AI is. I feel like there might have
- 25:04been an element of dunking on how bad so.
- 25:07This Vadnais is what's useful about it, because it suggests
- 25:10OAI is not going anywhere. All right now we're moving on
- 25:14from old BuzzFeed videos in the Stable Diffusion era not too
- 25:18long ago to pretty recent history.
- 25:21Sora. Yeah, so there's two dogs
- 25:33podcasting on what looks like a an Alpine hillside or something
- 25:36like that. That one looks pretty realistic.
- 25:40I mean, I would say like, I don't think it passes the Turing
- 25:43test as it were, where I'm like, that looks like a real video.
- 25:46There's just enough issues with it.
- 25:48But you know, it's it's starting to get into the era of like, oh,
- 25:52this kind of looks like, you know, a realistic image, I
- 25:55think. Whereas obviously what we saw
- 25:57before is nowhere close to that. What's interesting to me else is
- 26:00that they had these watermarks on there, which I think now
- 26:03they've kind of allowed you to remove the watermark, but I
- 26:07think there there was this fear that these things would be so
- 26:10real that you really need a persistent watermark on all
- 26:14these videos. They're more real, we careless.
- 26:16Yes, that's what's surprising to me.
- 26:18It's just like, or we've become more comfortable with the idea
- 26:22that any video on the Internet could be AI generated and
- 26:26everyone kind of knows that, but I'm not sure that's true.
- 26:29I think it's a race to the bottom.
- 26:30I think it's just there's enough competition in AI that these
- 26:32companies know that they want to valuable.
- 26:34They need to provide stuff without their label on it.
- 26:37Be somebody else is all right. What's next?
- 26:40Welcome to a nonexistent car show.
- 26:42Let's see some opinions. I mean man, the acceleration is
- 26:47crazy, you look far, step on the pedal and you are there.
- 26:52I feel safe with him in an SUV and it seems to be like the
- 26:56right type of car for him. I think the range is only only
- 27:01going to get better. Sorry.
- 27:04We don't want to drive gas cars anymore, Yeah.
- 27:07No more gas cars. You can say I'm kind of a kind
- 27:12of a misfit here, but don't tell anyone.
- 27:15I've just bought an electric car.
- 27:17I think it's really great for families and for little babies
- 27:20with all the safety features that these SUV's have.
- 27:23But what you're really seeing is that technology is going to be
- 27:27very, very important in terms of how we go forward.
- 27:30It was great to come to the conference because my husband
- 27:36loves cars. I think I have to buy an EV now.
- 27:40I love my muscle cars but I try to stay as healthy as I can so I
- 27:46can make it to the next car show.
- 27:51I mean, as someone who's watched a lot of like random YouTube
- 27:54videos of people looking at cars, it's pretty damn good.
- 27:58I mean, I would say it's it's like trained on YouTube videos
- 28:02of people looking at we. Can tell it's fake.
- 28:04Yeah, boomers maybe can't. I don't.
- 28:07There's only some humans who'd get confused.
- 28:09I think if you presented this to me and I didn't know it was
- 28:11fake, I would have to think about it for a bit.
- 28:14No, you're. Like no, I would immediately
- 28:16know. Certain certain clips.
- 28:17Certain clips. Certain clips are a little more
- 28:20obvious than others. So if you if you presented the
- 28:22whole video, I might start to think it's fake.
- 28:24But if you presented anyone clip like I think I I'd be pretty hit
- 28:28or miss whether I know it's fake.
- 28:30Sure, the stills are a little harder.
- 28:32It's sometimes the movement that gives it away.
- 28:34I do think like something we haven't really reflected on is
- 28:37how Sora got announced, I think over a year ago and like barely
- 28:42launched and has just been obliterated by VO.
- 28:45I mean, isn't that like a, an interesting story that they're,
- 28:49you know, I think what we would all argue that GBT is basically
- 28:52best in class or, you know, very near best in class at all times.
- 28:55And I feel like Sora was supposed to be the best and
- 28:59Google has just come through with the steam.
- 29:02Trains. Dominant, but it's crazy like I
- 29:05mean I don't think any of us would have predicted a year ago
- 29:07oh, Google's video model will be by far the best video model a
- 29:11year from today like it won't even be close it's not like it's
- 29:13like oh head to head right now but.
- 29:15I was a defender of Google. But why do we think they're so
- 29:18particularly good at video? Is it because they own YouTube?
- 29:21Yeah, I mean, they're just like doing YouTube.
- 29:24Yeah, but like everybody else stole all of the YouTube videos
- 29:27to train their models, what I understand from the buzz in the
- 29:30industry. But like, why are they still so
- 29:33much better at this than everyone else?
- 29:34It's it's sort of an interesting question to me.
- 29:36Like does Sam not care? Does Dario not care?
- 29:39Or is there something else about it?
- 29:41My theory would partially be that everyone's a little compute
- 29:45constrained and they just have different priorities and video
- 29:49is not going to be open as priority.
- 29:52Like they are more worried about anthropic and coding models and
- 29:55competing and things like that. But I don't know, just a theory.
- 29:58So do you think they'll maintain their advantage, Google?
- 30:01I mean, that would sort of be the case for that.
- 30:03Yeah, I kind of think that Google will be a little bit less
- 30:06compute constrained and have all this YouTube data and seems to
- 30:10care more about this. So yeah, I think they'll
- 30:12maintain the advantage for the next year or more.
- 30:15All right, one more. Kangaroo the airport.
- 30:17All right, let's. I actually haven't seen it.
- 30:19All right, we're going to watch. I've never seen this before.
- 30:22This is sort of an infamous one, that kangaroo in the airport.
- 30:35I mean, the kangaroo is high quality, but maybe, you know, as
- 30:38human beings, I'm not as good at the animal, you know, But the
- 30:42lips on one of the women, you can sort of tell, but it's
- 30:45getting very picky. Well, there's a couple of things
- 30:47I noticed. There's this very good kind of
- 30:50comedic zoom in effect, right? And that was probably prompted,
- 30:55which is pretty interesting. And then I don't even know if
- 30:57this is a real language. I don't think it is is that.
- 31:00Do you guys the audio is weird? Yeah, did they strange?
- 31:04Did they prompt it to create a fake language also?
- 31:07I feel like this fits into it was a Maxis theory that just
- 31:11like this is the end of culture sort of thing.
- 31:13It's like just the cinematic language.
- 31:15It's like the zoom in is just like whatever it's like.
- 31:19Exactly. It's the exactly like that.
- 31:21And so it's just like the video language we had up until now.
- 31:25We're just gonna like. Roll with it.
- 31:26Other than me knowing that Kangaroos are very unlikely to
- 31:29be in the airport, I would be easily be fooled.
- 31:31I mean. And the language piece,
- 31:32obviously, but like, like if there was, you know, no audio,
- 31:35I'd be like, the only reason that doesn't seem believable to
- 31:38me is because I've no Kangaroos aren't supposed to be in
- 31:40airports. Like there's nothing about the
- 31:42video to me that. Stood oh man the comments yeah
- 31:44I'm definitely getting scammed at 80 like you're probably
- 31:50gonna. Get scammed at 40.
- 31:52Like, and some of them are just, he's just sitting there all
- 31:54polite. I mean, the, the other
- 31:56terrifying thing about this is like, there are plenty of people
- 31:59who don't really care whether it's real or not, right?
- 32:01That's been a constant Internet problem.
- 32:03We get obsessed with like, can they tell?
- 32:05But they, there are people, there are, there's a whole genre
- 32:08of fake videos where people are like, who cares if they're
- 32:10acting or not? It's a little nice sketch, you
- 32:12know, like this. Is why I'm saying from the
- 32:14beginning of this podcast that people are going to be totally
- 32:17cool with the vast majority of their tick tock feed being fake.
- 32:20I, I mean, you guys are well, like now, man, like there's all
- 32:23about the emotional connection or whatever and I just.
- 32:25Don't believe I I believe you when it comes to short form
- 32:28video. I when you're talking about like
- 32:302 hour long podcasts, I'm just a little more skeptical.
- 32:33And podcasts, not to bring this to a really dark tone, but like
- 32:37with the protests in LA, you're seeing like the creation of
- 32:40almost like two different stories about how serious it is
- 32:44with images. Now, obviously I don't think
- 32:46those are fake images. Some of them are images from
- 32:49different riots and protests, and not necessarily the current
- 32:52one, but we've already seen this tendency that you can have a lot
- 32:55of influence on the Internet by showing images that tell the
- 32:59story you want. And now, obviously with AI,
- 33:02we're headed down this world where people have a strong
- 33:04desire to see the world they want and we'll be able to
- 33:07deliver it to them. Yeah.
- 33:09I mean, it is an interesting question as to whether or not
- 33:11like traditional news brands, you know, gain or lose esteem or
- 33:15status because like the assumption would be anything you
- 33:18see on whatever CNN is real right now.
- 33:21Obviously some people would debate, you know, whether or not
- 33:23they framed it correctly or whether the hosts are all biased
- 33:25or whatever. But like most people believe the
- 33:27footage they see on CNN is real, myself and.
- 33:29Right. Interesting.
- 33:30There'll be no AI brands, brands that are like, we don't have it,
- 33:34no. AI exactly like like a news
- 33:36brand or even a movie brand as we discussed earlier, I believe
- 33:39there is probably room in the market for saying we nothing is
- 33:42ever AI that we show, we verify every clip.
- 33:44It's kind of like what you do with investigative journalism
- 33:46where your brand is I only write the truth, right?
- 33:49You know, I will never put rumors.
- 33:52I will never put things I haven't verified right.
- 33:54And I think that is compelling as as an angle I.
- 33:56Know it's exhausting it'd be so much more fun to be one of these
- 33:59people just says whatever the fuck like yeah trying to be true
- 34:03in a world where people can say whatever they want it's.
- 34:05I kind of think that hardware companies, like the phone
- 34:08companies are going to have to create new technology that says,
- 34:12hey, this video that you took on your phone is provably a real
- 34:17video. And we anyone can go check that
- 34:21it's a real video. And maybe the social networks
- 34:23will have some auto checking that it's a real video.
- 34:26But I think it almost has to go down to the hardware level to be
- 34:29able to prove that. We saw an interview with Demos
- 34:31to Sabas from who runs Google DeepMind and then also Sergey
- 34:35Brandon, which was interesting where Demos said, and I didn't
- 34:38know this if they'd revealed this before that there is a
- 34:40secret watermark in every VO3 video, you know, so they can
- 34:43tell every every video that's fake.
- 34:45He sort of alluded to that there was a collusion scheme that the
- 34:49major players all have watermarks in their videos and
- 34:51they all know each others watermarks kind of that.
- 34:53So they know what a sort of video looks like or they know
- 34:55what a, you know, whomever. And that was very interesting
- 34:58because it would make me wonder if you have like a deep mind
- 35:01situation where someone decides to like be the only AI video
- 35:04company that doesn't put a watermark in their video, you
- 35:06know? Right, well, xai would be the
- 35:07class. Yeah, they're, yeah, they're the
- 35:09only ones who can get through the filters.
- 35:10And so everything on YouTube is real.
- 35:12But then there's this sort of like vector attack from, you
- 35:15know, you know, XAI or DeepMind or whatever to there's to upload
- 35:19a bunch of. Yeah.
- 35:20Yeah, yeah, yeah. Anyway, it was just kind of
- 35:22interesting. I didn't realize that they were
- 35:23so serious about that. I mean, I think a recurring
- 35:26theme, I've been optimistic that Chachi, BT and other models are
- 35:32going to bring culture into more of a shared reality because
- 35:35instead of going to a bunch of different links, you have a
- 35:38model that's trying to get a vision of reality and it's at
- 35:40least delivering one reality. And if the model gets it wrong,
- 35:43maybe all of society is misled. But at least we like are getting
- 35:47back to a shared culture. I feel like this video, the
- 35:51experiment feels sort of the opposite.
- 35:53I mean, I think partially because it's going to be
- 35:55narrowly tailored. I don't know, I it feels like I
- 35:58worry about video being a path to manipulate the culture in a
- 36:02way that I feel like the text models, I feel all more
- 36:05optimistic about them and forming the culture.
- 36:09I agree with you, but maybe with a slightly different angle,
- 36:11which is that I think fundamentally video is like by
- 36:13far the most effective way to convey information to people,
- 36:16right? I mean, I think that like people
- 36:19take in, you know, the whole, you know, pictures worth 1000
- 36:22words, maybe a video's worth 10,000 words or whatever, like.
- 36:25Right. I guess it's dangerous, yeah.
- 36:27Yeah, exactly. They take in the information
- 36:29from video at such a higher bandwidth level than they do
- 36:32from books. You know that I think that it's
- 36:35a really, really powerful communication and teaching
- 36:39medium. And so in some sense, like
- 36:41everything that's on video, yeah, and propaganda, everything
- 36:44that's on video is just supercharged in terms of its
- 36:46ability to influence people and seem like it's reality, right.
- 36:50So I sort of think that, you know, the title of this episode,
- 36:52I think was like the war on reality.
- 36:54I think the war on video reality is sort of the most important
- 36:58one in some ways because it is the highest bandwidth, I think,
- 37:01communication medium that we have if.
- 37:03It's in video. People believe it.
- 37:04Well, I think your point, Eric, is that people go to Chachi BT
- 37:08to kind of learn about the reality of the world.
- 37:10They ask questions, questions about things, they ask for
- 37:13advice, life advice that they're actually going to put into
- 37:15practice. I think, you know, people go
- 37:18scroll TikTok for a different reason.
- 37:20It's really to just zone out, be entertained.
- 37:22People watch Netflix to be kind of in these fictional fantasy
- 37:25worlds. I just think these are very
- 37:27different use cases. Text, first video.
- 37:29And I think that's what you're you're noticing.
- 37:31Preparing you to go out into the world and sort of like, livid.
- 37:36Whereas I feel like video, you are like living the world.
- 37:39You're like, oh, it's as if you were like living.
- 37:41And you're like, take what you've seen in video and you're
- 37:43like, that's the world. I've seen it.
- 37:45You know, I don't know. It's, yeah.
- 37:48It seems dangerous, but very compelling.
- 37:51Anyway, that's our show. Thanks for sticking around.
- 37:54We're getting ready for this Rebel Valley AI Summit, London,
- 37:57June 25th. Thanks for joining us.
- 38:30Yeah, I'm still eating the spaghetti in 2025.