This is the nail in the coffin for me, I think. I can't see myself ever truly engaging with new digital media without some kind of proof of authenticity, or trusting any digital communication...
This is the nail in the coffin for me, I think. I can't see myself ever truly engaging with new digital media without some kind of proof of authenticity, or trusting any digital communication without a way to vet it in real life.
we really do need to heavily regulate (if not, outright ban) all deepfakes and invoke heavy punishments on anything that falls short of good faith. Deception in this day and age is every bit as...
we really do need to heavily regulate (if not, outright ban) all deepfakes and invoke heavy punishments on anything that falls short of good faith. Deception in this day and age is every bit as dangerous as murder.
Do we need to ban Photoshopped images as well? I fail to see a fundamental difference. If you know that any form of media, whether it's written text, still images or video is easily manipulated or...
Do we need to ban Photoshopped images as well? I fail to see a fundamental difference. If you know that any form of media, whether it's written text, still images or video is easily manipulated or faked, you simply discount the validity of anything in that form of media that doesn't come with some independent verification.
"anything that falls short of good faith" Is the determining factor here. But yes, I do think you should be criminally liable for maliciously making a bad Photoshop of MS-13 on someone's hands....
"anything that falls short of good faith" Is the determining factor here. But yes, I do think you should be criminally liable for maliciously making a bad Photoshop of MS-13 on someone's hands.
Parody would be good faith, as well as anything explicitly labeling itself as fake.
you simply discount the validity of anything in that form of media that doesn't come with some independent verification.
As seen with modern discoree, society has clearly lost its ability to properly verify, and it's only getting harder to distinguish betwene real and fake. When man falls to their own devices, that's when government steps in to regulate. So this seems appropriate.
I would indeed put disinformation on the same platform as factors like gambling, alcohol, and sexual intimacy. Natural factors of life that government chooses to regulate.
This is from ChatGPT and I didn't verify its accuracy because I'm just illustrating what kind of ballpark our exposure to various forms of media is in: Your mind is not capable of discounting the...
you simply discount the validity of anything in that form of media that doesn't come with some independent verification.
This is from ChatGPT and I didn't verify its accuracy because I'm just illustrating what kind of ballpark our exposure to various forms of media is in:
According to industry and academic sources (e.g., Nielsen, Statista, academic studies in media psychology):
Average daily screen time in Western countries (e.g., US, UK, Canada) ranges from 7 to 10 hours/day, split across devices and platforms.
Within that time, people are typically exposed to several hundred to several thousand discrete pieces of visual media content. For instance:
Ads alone: 4,000–10,000 impressions per day, mostly visual (source: Yankelovich Inc. estimates, cited widely).
Short-form content (e.g., TikTok, Instagram Stories): dozens to hundreds of clips per day for active users.
Streaming or TV: several episodes or hours per day, each with dozens of shots or scenes.
Your mind is not capable of discounting the validity of this kind of avalanche of content. You can say to yourself "it may not be real" but your lizard brain will still process it as if it were, to an extent. The less attention you pay to what you absorb, the less scrutiny you will be able to apply and the more directly the marketing messages get embedded. If you try to scrutinise everything you see, all the time, you will tire out really fast. Each of us only has so much bandwidth for such high level cognitive processing a day.
Did you know that Netflix is now deliberately producing slop content? They call it "optimising for second screenability", meaning it should be brainless enough for you to be able to follow along while you're actually engaged with a handheld device (your "first screen"). That's from some likely searchable online source (I forgot where, might remember later). My own interpretation is that they're investing in it due to its massive subliminal marketing potential.
Subliminal messages in marketing are banned due to their stunning effectiveness. A subliminal message is something you don't consciously pay attention to. When a Netflix customer is "second screening" some slop content, they are exposing themselves to such messaging in a way that is currently completely legal.
Official announcement: https://blog.google/technology/ai/google-flow-veo-ai-filmmaking-tool/ I'm seeing a bunch of Veo 3 stuff being posted to various Slack and Discord channels I'm part of, and...
I'm seeing a bunch of Veo 3 stuff being posted to various Slack and Discord channels I'm part of, and it's honestly crazy how realistic they are. The lip syncing feels a tiny bit off on most of them, but unlike a lot of other AI generated human imagery, I am not experiencing the uncanny valley effect when looking at them... which is creepy in its own way.
I honestly was still seeing some uncanny valley in a lot of it but I thought it looked great for videogame cutscene stuff. Then I saw this other one on twitter this morning and it’s the first time...
I honestly was still seeing some uncanny valley in a lot of it but I thought it looked great for videogame cutscene stuff. Then I saw this other one on twitter this morning and it’s the first time it feels real to me that some movies and media we connect with will (perhaps soon) be generated with AI:
It's like 95% there. Weird teeth, exaggerated emotions like we're playing LA Noire, and text telling me to "keep rell". I suspect the last 5% are the keys to the kingdom and exceedingly difficult...
It's like 95% there. Weird teeth, exaggerated emotions like we're playing LA Noire, and text telling me to "keep rell". I suspect the last 5% are the keys to the kingdom and exceedingly difficult to attain.
I've seen a couple commercial-likes with someone running and that's when it looks indistinguishable. As soon as they start talking it's... off.
There will be commercials completely built with AI very soon. The ones that do not really need in depth acting, but a person walking by or showing off their shoes.
For actual deep fake stuff we need to go deeper. Although, your video will absolutely kill on Facebook. FB is already cooked when it comes to AI.
I was actually just thinking about this myself not long after Google I/O. I apologize; I'm going to jynx this but what I came up with is what if Google updated the YouTube ad platform? Brands...
There will be commercials completely built with AI very soon. The ones that do not really need in depth acting, but a person walking by or showing off their shoes.
I was actually just thinking about this myself not long after Google I/O. I apologize; I'm going to jynx this but what I came up with is what if Google updated the YouTube ad platform? Brands could pay for a targeted campaign with maybe a general theme and then just let Google run with it. Google could take that, plus their user targeting/personalization infrastructure and dump some kind of "make an ad script" prompt into Gemini for a user or groups of users. Then take that output and dump it into Veo. Then take that output and put it into your YouTube videos, per their last "put ads into the best locations" development. Boom. Hyper-personalized ads at the perfect time for maximum click-through..
Cost-wise I don't know how this pans out, but I would be shocked (SHOCKED!) if this wasn't being worked on.
That is almost precisely what Zuck has said he intends to do with Facebook. Though that also includes fake AI accounts 'posting' submarine advertising for products and other astroturfing tools.
That is almost precisely what Zuck has said he intends to do with Facebook. Though that also includes fake AI accounts 'posting' submarine advertising for products and other astroturfing tools.
Someone already made a proof of concept that to my eyes is already there. The lip sync is a bit off and there are a few other things not quite right. But lip sync being off is something I have...
There will be commercials completely built with AI very soon. The ones that do not really need in depth acting, but a person walking by or showing off their shoes.
The lip sync is a bit off and there are a few other things not quite right. But lip sync being off is something I have already seen with commercials for decades (here in the Netherlands anyway) as a lot of them are made in one country and then just dubbed over for other countries. In some ways it made it more of a commercial to me.
This is an extension of the 80/20 rule. You can get 80% of the result for 20% of the cost. It applies to so many things in so many ways and is another way of saying you can have good or cheap....
I suspect the last 5% are the keys to the kingdom and exceedingly difficult to attain.
This is an extension of the 80/20 rule. You can get 80% of the result for 20% of the cost. It applies to so many things in so many ways and is another way of saying you can have good or cheap. Applies in metrology too - cheap to measure an centimeter, expensive to measure a nanometer. Applied to generative AI: (currently) easy to have a generated video, hard/expensive to have a believable generated video. Except it's getting tech-generationally better on the timescale of months.
With all that said, the speed of progression of AI is.... beyond breakneck. We've had people working on AI SINCE ALAN TURING. We've had computers + people working on AI since we've had computers. But these past few years? It's as if we're seeing decades in days.
Utterly convincing. Like, yes, the subtitles and text on the signs are obviously wrong, but I had a moment where I wondered whether those were added in to actual footage to make it look...
Utterly convincing.
Like, yes, the subtitles and text on the signs are obviously wrong, but I had a moment where I wondered whether those were added in to actual footage to make it look AI-generated. Was the actual dupe that this was real footage, and someone just edited it to masquerade as something AI-generated? Did they direct some of the actors to exaggerate just a little bit to come across as a little bit less than real?
I've since gone through the video multiple times and can spot several small "tells" throughout, but if you're not looking for them then this is, at a casual glance, a professionally made video of actual people. Also the fact that I genuinely can't tell what's real anymore, even in a simple video like this, is incredibly disorienting.
With how rapidly AI is advancing, we're likely only months away from the obvious "tells" being iterated out, making future videos absolutely unclockable.
I don't want to be pessimistic, but when I think about how powerful this tool is, and then I pair it with information like this, it feels like we're in an informational free-fall right now. The term "post-truth" has been around for a while, but it's never felt as urgent as it does in our present moment.
The last barrier to fall will be the speed. There's still a delay after a prompt before the response in text, and one can't have an AI render a live stream in real time reacting to the chat. That...
The last barrier to fall will be the speed. There's still a delay after a prompt before the response in text, and one can't have an AI render a live stream in real time reacting to the chat. That delay is the only hurdle keeping this from feeling 'alive' and fooling everyone. When it falls...
I'll just keep my fingers crossed for a mega solar flare that scores a direct hit on Earth. That'll end everything electronic forever with extreme prejudice. The internet needs a total do-over and that's the only way I can imagine it happening. :P
I plan on taking the Murderbot approach to the near future: (no, not killing people) Turning down my emotional response to things and holing up to consume the massive archive of media I've...
I plan on taking the Murderbot approach to the near future:
(no, not killing people)
Turning down my emotional response to things and holing up to consume the massive archive of media I've accumulated.
My next SAN project is going to be solar. Maximum performance and longevity at a mere 40 watts of power. Just in case I need a reliable media fix off-grid. ;)
My next SAN project is going to be solar. Maximum performance and longevity at a mere 40 watts of power. Just in case I need a reliable media fix off-grid. ;)
I think your link might be wrong? At first I got really excited because a solar-powered gaming PC sounds awesome, but there wasn’t anything about solar on the site.
I think your link might be wrong?
At first I got really excited because a solar-powered gaming PC sounds awesome, but there wasn’t anything about solar on the site.
It's not for gaming at all, hasn't even got a monitor. It's a link to an Asus device that allows you to stack 48TB of space on 12 M.2 drives, compatible with TrueNAS, and the entire rig runs fine...
It's not for gaming at all, hasn't even got a monitor. It's a link to an Asus device that allows you to stack 48TB of space on 12 M.2 drives, compatible with TrueNAS, and the entire rig runs fine on half a cheap solar panel's worth of power generation in direct sunlight. Or, it'll run for about 7 hours on a solar-charging battery pack like this. No moving parts, and as long as it is powered on once every three years or so to refresh the M.2 cells it should last a couple decades. It's for storing Netflix in the palm of your hand and with room to spare.
Ah, that’s really cool! Shame on me for not actually reading the link AND having the nerve to confidently call it out! To my credit, it looks very similar to my MSI Trident 3, so I saw it and just...
Ah, that’s really cool! Shame on me for not actually reading the link AND having the nerve to confidently call it out!
To my credit, it looks very similar to my MSI Trident 3, so I saw it and just assumed it was a direct analogue.
Oh. My God. You just unlocked a new horror for me. Like I like using chat bots for trying to understand the current scientific understanding of their harms, and they also came in clutch this...
I don't want to be pessimistic, but when I think about how powerful this tool is, and then I pair it with information like this, it feels like we're in an informational free-fall right now. The term "post-truth" has been around for a while, but it's never felt as urgent as it does in our present moment.
Oh. My God. You just unlocked a new horror for me. Like I like using chat bots for trying to understand the current scientific understanding of their harms, and they also came in clutch this weekend finding cat feeders for my family and reducing my parent's cable bill. But this thought just makes me want to move to a cabin in the woods and live without electricity.
I noticed that the voices still sound off. But the video itself is hard to tell apart from real footage depending on the scene (well, unless it has subtitles). Then again, I remember only a few...
I noticed that the voices still sound off. But the video itself is hard to tell apart from real footage depending on the scene (well, unless it has subtitles).
Then again, I remember only a few years ago language models didn't seem like they would produce text as well as they do now either. So I wouldn't be surprised if soon it will be indistinguishable from real footage.
I can't help but wonder how well it deals with a larger narrative though, and how feasible it is to make it do so. It is rather challenging for AI to remember details of previous prompts. So you might just end up with video in which the main character randomly looks like a different person depending on the scene. So you'd not be able to replace a film crew entirely unless you resolve that.
For videogames I think these days we make them inside the game engine. It used to be that games would use pre-rendered video for cutscenes but it was always weird and immersion breaking if the art style suddenly changes when going to a cutscene. It looks more natural now. Going back to pre-rendered video would seem like a step backwards, whether AI is involved or not.
The only easy tell left is words - it still consistently messes up text rendered on signs for example. I wonder why it has such a hard time with lettering. I'd say the skin tone is also a bit of a...
The only easy tell left is words - it still consistently messes up text rendered on signs for example. I wonder why it has such a hard time with lettering. I'd say the skin tone is also a bit of a giveaway, because real humans rarely have skin that's so consistently smooth and featureless unless they've just come out of a studio makeup room.
My instinct is because it’s not trying to write words, it’s trying to paint word-shapes. I imagine if big road signs were common further back in time, impressionist paintings might give us a...
it still consistently messes up text rendered on signs for example. I wonder why it has such a hard time with lettering.
My instinct is because it’s not trying to write words, it’s trying to paint word-shapes. I imagine if big road signs were common further back in time, impressionist paintings might give us a similar vibe, with road markings being messy and unintelligible and yet perfectly give the impression of a road sign from a distance
Every time I see this type of generative video, I can't help but notice how they still aren't good enough. Maybe for a low effort YouTube video, but we are a long way off from something like this...
Every time I see this type of generative video, I can't help but notice how they still aren't good enough. Maybe for a low effort YouTube video, but we are a long way off from something like this replacing a VFX artist on a major Hollywood production. It's good but it's clearly not real. It still hasn't gotten a lot of the details right and doesn't really hold up to any sort of scrutiny.
I do think it's just a matter of time before we get over that hump. It could be 5 years, it could be 10, but we will get there eventually. There's too much money for corporations to make by just using AI instead of paying real artists. That sucks.
It's also just a matter of time before these are truly weaponized and used for political purposes. I mean, if he could, don't you think Trump would just fabricate videos of his enemies to use as evidence to lock them up? I could easily see his regime doing something like generating a video of Kilmar Abrego Garcia saying he's in MS-13 and he hates the U.S.?
I expect lots of slop, but the flip side of that might be indy movie or TV show makers that use it to make something interesting. For example, South Park animation is pretty crude, but they did...
I expect lots of slop, but the flip side of that might be indy movie or TV show makers that use it to make something interesting. For example, South Park animation is pretty crude, but they did manage to connect with people.
I wouldn’t be surprised it if happens by the end of the year. Run Veo 3 for the rough animation, then have an actual artist do the finishing. In theory this could mean the ability to make more...
I wouldn’t be surprised it if happens by the end of the year. Run Veo 3 for the rough animation, then have an actual artist do the finishing. In theory this could mean the ability to make more movies (and thus, no job losses), but in practice there’s an upper limit to the number of movies, and especially the number of good movies.
It doesn't have to get to a point that it's good enough for a major Hollywood production, although I still wouldn't bet against it. I'm confident we'll see this adopted in the mainstream for...
It doesn't have to get to a point that it's good enough for a major Hollywood production, although I still wouldn't bet against it. I'm confident we'll see this adopted in the mainstream for little tweaks to real media. Imagine if the adult film industry gave their consumers the ability to tweak videos a little bit. Extend this scene for a little longer. What if the cameraman didn't cut away to this angle and instead went to this angle? Augmenting existing media with realistic enough additions seems attainable within the next couple of years.
Eventually as these tools are adopted some people will not pursue their passion into a career... because they will just think "why bother?". Like a self fulfilling prophecy these tools will be...
Eventually as these tools are adopted some people will not pursue their passion into a career... because they will just think "why bother?".
Like a self fulfilling prophecy these tools will be used because they are good enough, cheap... and because some people will have just give up deleting the human alternative.
This isn't new, my dream career is making physical props and effects, with my inspiration being Jim Henson when he does the darker stuff. However I graduated high school in 2013, CGI already took...
Eventually as these tools are adopted some people will not pursue their passion into a career... because they will just think "why bother?".
This isn't new, my dream career is making physical props and effects, with my inspiration being Jim Henson when he does the darker stuff. However I graduated high school in 2013, CGI already took over so I was in this scenario of "why bother?"
I have a friend that is 5+ years older and he got a college degree for this kind of stuff, but all its good for now is to make him feel more legit as an local artist and it got him a job making toys for Hot Topic 10+ years ago when they sold that stuff.
(This paragraph is a poorly phrased rant, feel free to ignore)
I really wish we could find a better balance between taking advantage of new technology and still using the old styles when they are appropriate. Unfortunately the bottom line is its cheaper and easier to use AI/CGI, so thats the decision that gets made. If we could look past the bottom line to actually making something creative, and letting their vision get made then I think there would be a lot more of the old styles would still be done. But instead original ideas seem like a thing of the past and its all live action remakes or an nth level sequel for a series that really should have died long ago.
I’m expecting most commercials to use this within a few years. Movies will use this as the new green screen in places. But I can’t see audiences being quick to tolerate AI actors.
I’m expecting most commercials to use this within a few years. Movies will use this as the new green screen in places. But I can’t see audiences being quick to tolerate AI actors.
They won't know. It'll start with mostly live performance, fixed up/edited with AI. Over time it'll shift towards less performance, more AI. I predict an article in x years that says "did you know...
But I can’t see audiences being quick to tolerate AI actors.
They won't know. It'll start with mostly live performance, fixed up/edited with AI. Over time it'll shift towards less performance, more AI. I predict an article in x years that says "did you know this award winning scene was actually made entirely with AI?" And people will be shocked, even though by that point it'll be standard.
I can definitely see a lot of video games making use of it along with other generative art models, and coding agents as well. They all substantially lower the barrier to entry (and cost) for...
I can definitely see a lot of video games making use of it along with other generative art models, and coding agents as well. They all substantially lower the barrier to entry (and cost) for making games. But yeah, I don't see serious actors in films and TV shows being totally replaced by AI generated ones any time soon either.
Agreed. I wonder what new kinds of movies enthusiasts will make. Remember Red versus Blue? Machinima (making movies with video game engines) was an interesting trend for a while. Or maybe people...
Agreed. I wonder what new kinds of movies enthusiasts will make. Remember Red versus Blue? Machinima (making movies with video game engines) was an interesting trend for a while.
Or maybe people will make something shorter and less coherent like music videos.
Music videos with AI visuals are already out there. It's a medium where a series of short cuts are appropriate, and variety of imagery can be accepted as long as it's giving the right vibe. It's a...
Music videos with AI visuals are already out there. It's a medium where a series of short cuts are appropriate, and variety of imagery can be accepted as long as it's giving the right vibe. It's a great fit even before it can give long term narrative coherence.
I'm doubtful this will hit professional games anytime soon. Unlike movies, video games need very specific assets both to let them animate properly and not bog performance down to a crawl. Most...
I'm doubtful this will hit professional games anytime soon. Unlike movies, video games need very specific assets both to let them animate properly and not bog performance down to a crawl. Most optimization problems in modern games comes from the art pipeline: specifically, lighting and model complexity. This is a big part of why all the AI slop flooding asset stores is basically useless for any serious game.
Sure, maybe some cutscenes can be made with AI. But it's hard to take shortcuts on anything you'll control or interact with as of now. It'll probably linger for 5-10 years on what's currently out there: generative art workflows to spam out skins, or for environmental textures.
I think AI's going to make larger changes to the gameplay than the art assets at first. I'm looking forward to The Wayward Realms. It's the old Daggerfall / Morrowind / Oblivion developers that...
I think AI's going to make larger changes to the gameplay than the art assets at first. I'm looking forward to The Wayward Realms. It's the old Daggerfall / Morrowind / Oblivion developers that left Bethesda, so it's basically going to be Elder Scrolls done right minus Todd Howard's creeping crud. What really interests me is the level of AI baked in - like the old 'radiant' engine from Bethesda games but this time as an active game master managing 40 (yes 40) factions across the game world, all reacting to the player, generating quests, dialog, and NPC daily routines. That should take RPGs to a whole new level of replayability.
I've been seeing these awful ads on social media that are 100% AI. Mostly testimonials from the "person" speaking. One was even for a over-the-counter medication, which should probably be illegal,...
I've been seeing these awful ads on social media that are 100% AI. Mostly testimonials from the "person" speaking. One was even for a over-the-counter medication, which should probably be illegal, right?
I don't know that much about image generation AI, but the other day I was trying to think through how you would confirm the authenticity of video if AI video tools like this become widespread......
I don't know that much about image generation AI, but the other day I was trying to think through how you would confirm the authenticity of video if AI video tools like this become widespread...
One idea I had for recording confirmably authentic video of something would be to set up multiple cameras at different angles. I think it would be pretty immediately apparent if two videos of, say, a person talking and gesturing, didn't match each other - spatial perception of human bodies and faces is something that humans are particularly good at. And it strikes me as a difficult task to not just generate a realistic video but multiple realistic videos of the same event that are all spatially coherent with each other.
But like I said, I'm not an expert on the topic and maybe someone who is could weigh in on this.
In the videos I've seen (such as the water bottle person video), it's not even consistent between shots. If you watch the conference room scene that starts around 1:15, it can cut away and back to...
In the videos I've seen (such as the water bottle person video), it's not even consistent between shots. If you watch the conference room scene that starts around 1:15, it can cut away and back to a consistent person, but each shot seems to be a different cast of characters. It's particularly noticeable if you try and figure out who's at the head of the table or look for any of the people from up-close shots in the wider shots. So I suspect your plan would work even better and reveal discrepancies much larger than just minor spatial irregularities.
I wouldn't be surprised if merely doing a slow 360 would still be enough to make AI video come up short. In earlier AI videos, it seemed to me like it only mattered that the new stuff panning into view matched what was on-screen. So doing a slow 360 at a public event would give you a whole new crowd of onlookers. I really hope these "problems" don't get solved soon.
I agree these "AI actors" do look way better than ones we've seen before. I'd like to see a blind video to be honest because I feel like I'm seeing weirdness watching the clips in this thread but...
I agree these "AI actors" do look way better than ones we've seen before. I'd like to see a blind video to be honest because I feel like I'm seeing weirdness watching the clips in this thread but I'm not sure if it's because I know they are AI and I'm looking for flaws or not.
I'm also curious how easy these clips are to direct honestly. Like for any creative trying to direct, is the expectation you are going to sit there for hours and hours regenerating a clip with slightly different prompts to try and get the exact next sequence you want?
Also if the rest of my small experience of AI image generation has anything to go by, their controls like the camera controls (jib up) are going to ruin a perfect scene. You might get a clip you like except the "camera" is in the wrong place, try to move it and the whole scene completely changes. Or changes enough that it's invalidated what you're looking for.
Sounds incredibly tedious IMO.
It certainly was tedious when I tried to get Midjourney to create halfway decent images a couple years ago, but my understanding is that AI-based image editing tools are improving. OpenAI brags...
It certainly was tedious when I tried to get Midjourney to create halfway decent images a couple years ago, but my understanding is that AI-based image editing tools are improving.
OpenAI brags about their latest image generator here. I tried it myself and it did seem much better, but it’s also very slow at generating each image, so in the end it’s still tedious.
Mark my words, our grandchildren will be subscribing to movie and TV AI services where instead of Netflix they'll be sharing prompts and seeds of TV shows amongst themselves. People will be able...
Mark my words, our grandchildren will be subscribing to movie and TV AI services where instead of Netflix they'll be sharing prompts and seeds of TV shows amongst themselves. People will be able to tailor the humor to their own sense of humor and tailor the acting and actors to their own personal favorites.
If they don't like an ending, they can generate a different one.
People won’t just search for a title, they’ll input a prompt. They’ll share “seeds” of shows with each other; starting points that generate unique variations. Humor, tone, pacing, themes, settings, and character archetypes will all be adjustable. A viewer who prefers dry, sarcastic humor can tweak a show accordingly. Someone else might want the same plot but with more emotional weight or different music. Acting styles and even the visual appearance of characters could be changed to suit personal taste.
There will be completely AI generated content, and content that is guided and directed by humans but generated by AI.
People will be generating Season 2 of Firefly, as well as re-generating a better season of Season 7/8 of Game of Thrones. There will still be collaborative or standout “runs” that gain traction, like a particularly well-regarded 10-season continuation of Firefly or a widely shared reimagining of Game of Thrones. But the core model is individual consumption and customization. It’s not about contributing to a central canon, it’s about creating your own.
No more canceled shows. If a show's not good enough people can fix it themselves. People will be sharing their tweaks and adjustments, which will be rated.
Infinite sequels, evolving worlds, infinite spinoffs. There will be no limits.
You'll have a basic plot, but then variations depending on taste. Think a "Spy Thriller" and the variations are "Set in the Star Wars Universe" or "Set in the Flintstones Stone age" or "Animated in the style of Rick and Morty".
Yeah I'm not describing it as a good thing, but just where we're headed. However I'm sure there are many things in today's world that many in the early 1900s would have considered a nightmare...
Yeah I'm not describing it as a good thing, but just where we're headed.
However I'm sure there are many things in today's world that many in the early 1900s would have considered a nightmare today. Haha Probably why we have so many boomers refusing to give up positions in governments and going crazy about a world they'll probably never get to see, because they too think this world's a nightmare.
I'm not convinced the entire genre would survive that. It seems like something that could actually kill the idea of watching tv by fast forwarding to the end where it's all been done. I could see...
I'm not convinced the entire genre would survive that. It seems like something that could actually kill the idea of watching tv by fast forwarding to the end where it's all been done. I could see some people becoming infinitely addicted and ruined by such a thing. But I could see a lot of people burning out on it in a few months.
I could very easily be wrong. I just think sitting back an observing without directing will always be a necessary part of it. I think it's more likely that you'll see AI settings created by "show runners", and systems of delivery that respond to the user live. Using biometrics or similar to decide if the viewer is responding positively or negatively to humor, politics, sci fi, etc.
So less hands on than sharing prompts that you have visibility and control of, and more lining up to the content feed that serves automatically adjusting content designed for maximum addiction. Just as dystopian, though. I don't want to see me kids sticking electrodes to their temples so they can activate AI mode on their TV because it's the only thing they enjoy any more.
I mean I think what you said is what it will be like at first, but eventually. I mean it's not much different an idea as the computers behind Holodecks in Star Trek.
I mean I think what you said is what it will be like at first, but eventually.
I mean it's not much different an idea as the computers behind Holodecks in Star Trek.
At this point I'm really tempted to just use a generative model to create video content, novels, a webcomic, a website, etc. If AI has taken over to the point where some shitty algorithm is...
At this point I'm really tempted to just use a generative model to create video content, novels, a webcomic, a website, etc. If AI has taken over to the point where some shitty algorithm is preventing me from going into another accountancy role (whether through some bullshit automated screening process, or because ChatGPT can crunch numbers better than me), then it's only a matter of time before more people are displaced.
This is the nail in the coffin for me, I think. I can't see myself ever truly engaging with new digital media without some kind of proof of authenticity, or trusting any digital communication without a way to vet it in real life.
We are so fucked.
we really do need to heavily regulate (if not, outright ban) all deepfakes and invoke heavy punishments on anything that falls short of good faith. Deception in this day and age is every bit as dangerous as murder.
Do we need to ban Photoshopped images as well? I fail to see a fundamental difference. If you know that any form of media, whether it's written text, still images or video is easily manipulated or faked, you simply discount the validity of anything in that form of media that doesn't come with some independent verification.
"anything that falls short of good faith" Is the determining factor here. But yes, I do think you should be criminally liable for maliciously making a bad Photoshop of MS-13 on someone's hands.
Parody would be good faith, as well as anything explicitly labeling itself as fake.
As seen with modern discoree, society has clearly lost its ability to properly verify, and it's only getting harder to distinguish betwene real and fake. When man falls to their own devices, that's when government steps in to regulate. So this seems appropriate.
I would indeed put disinformation on the same platform as factors like gambling, alcohol, and sexual intimacy. Natural factors of life that government chooses to regulate.
This is from ChatGPT and I didn't verify its accuracy because I'm just illustrating what kind of ballpark our exposure to various forms of media is in:
Your mind is not capable of discounting the validity of this kind of avalanche of content. You can say to yourself "it may not be real" but your lizard brain will still process it as if it were, to an extent. The less attention you pay to what you absorb, the less scrutiny you will be able to apply and the more directly the marketing messages get embedded. If you try to scrutinise everything you see, all the time, you will tire out really fast. Each of us only has so much bandwidth for such high level cognitive processing a day.
Did you know that Netflix is now deliberately producing slop content? They call it "optimising for second screenability", meaning it should be brainless enough for you to be able to follow along while you're actually engaged with a handheld device (your "first screen"). That's from some likely searchable online source (I forgot where, might remember later). My own interpretation is that they're investing in it due to its massive subliminal marketing potential.
Subliminal messages in marketing are banned due to their stunning effectiveness. A subliminal message is something you don't consciously pay attention to. When a Netflix customer is "second screening" some slop content, they are exposing themselves to such messaging in a way that is currently completely legal.
What counts as independent verification? If the government says so? The president? Who is actually independent and can verify the thing?
Official announcement:
https://blog.google/technology/ai/google-flow-veo-ai-filmmaking-tool/
I'm seeing a bunch of Veo 3 stuff being posted to various Slack and Discord channels I'm part of, and it's honestly crazy how realistic they are. The lip syncing feels a tiny bit off on most of them, but unlike a lot of other AI generated human imagery, I am not experiencing the uncanny valley effect when looking at them... which is creepy in its own way.
I honestly was still seeing some uncanny valley in a lot of it but I thought it looked great for videogame cutscene stuff. Then I saw this other one on twitter this morning and it’s the first time it feels real to me that some movies and media we connect with will (perhaps soon) be generated with AI:
https://x.com/HashemGhaili/status/1925616536791760987
It's like 95% there. Weird teeth, exaggerated emotions like we're playing LA Noire, and text telling me to "keep rell". I suspect the last 5% are the keys to the kingdom and exceedingly difficult to attain.
I've seen a couple commercial-likes with someone running and that's when it looks indistinguishable. As soon as they start talking it's... off.
There will be commercials completely built with AI very soon. The ones that do not really need in depth acting, but a person walking by or showing off their shoes.
For actual deep fake stuff we need to go deeper. Although, your video will absolutely kill on Facebook. FB is already cooked when it comes to AI.
I was actually just thinking about this myself not long after Google I/O. I apologize; I'm going to jynx this but what I came up with is what if Google updated the YouTube ad platform? Brands could pay for a targeted campaign with maybe a general theme and then just let Google run with it. Google could take that, plus their user targeting/personalization infrastructure and dump some kind of "make an ad script" prompt into Gemini for a user or groups of users. Then take that output and dump it into Veo. Then take that output and put it into your YouTube videos, per their last "put ads into the best locations" development. Boom. Hyper-personalized ads at the perfect time for maximum click-through..
Cost-wise I don't know how this pans out, but I would be shocked (SHOCKED!) if this wasn't being worked on.
That is almost precisely what Zuck has said he intends to do with Facebook. Though that also includes fake AI accounts 'posting' submarine advertising for products and other astroturfing tools.
Someone already made a proof of concept that to my eyes is already there.
The lip sync is a bit off and there are a few other things not quite right. But lip sync being off is something I have already seen with commercials for decades (here in the Netherlands anyway) as a lot of them are made in one country and then just dubbed over for other countries. In some ways it made it more of a commercial to me.
This is an extension of the 80/20 rule. You can get 80% of the result for 20% of the cost. It applies to so many things in so many ways and is another way of saying you can have good or cheap. Applies in metrology too - cheap to measure an centimeter, expensive to measure a nanometer. Applied to generative AI: (currently) easy to have a generated video, hard/expensive to have a believable generated video. Except it's getting tech-generationally better on the timescale of months.
With all that said, the speed of progression of AI is.... beyond breakneck. We've had people working on AI SINCE ALAN TURING. We've had computers + people working on AI since we've had computers. But these past few years? It's as if we're seeing decades in days.
Utterly convincing.
Like, yes, the subtitles and text on the signs are obviously wrong, but I had a moment where I wondered whether those were added in to actual footage to make it look AI-generated. Was the actual dupe that this was real footage, and someone just edited it to masquerade as something AI-generated? Did they direct some of the actors to exaggerate just a little bit to come across as a little bit less than real?
I've since gone through the video multiple times and can spot several small "tells" throughout, but if you're not looking for them then this is, at a casual glance, a professionally made video of actual people. Also the fact that I genuinely can't tell what's real anymore, even in a simple video like this, is incredibly disorienting.
With how rapidly AI is advancing, we're likely only months away from the obvious "tells" being iterated out, making future videos absolutely unclockable.
I don't want to be pessimistic, but when I think about how powerful this tool is, and then I pair it with information like this, it feels like we're in an informational free-fall right now. The term "post-truth" has been around for a while, but it's never felt as urgent as it does in our present moment.
The last barrier to fall will be the speed. There's still a delay after a prompt before the response in text, and one can't have an AI render a live stream in real time reacting to the chat. That delay is the only hurdle keeping this from feeling 'alive' and fooling everyone. When it falls...
I'll just keep my fingers crossed for a mega solar flare that scores a direct hit on Earth. That'll end everything electronic forever with extreme prejudice. The internet needs a total do-over and that's the only way I can imagine it happening. :P
I plan on taking the Murderbot approach to the near future:
(no, not killing people)
Turning down my emotional response to things and holing up to consume the massive archive of media I've accumulated.
My next SAN project is going to be solar. Maximum performance and longevity at a mere 40 watts of power. Just in case I need a reliable media fix off-grid. ;)
I think your link might be wrong?
At first I got really excited because a solar-powered gaming PC sounds awesome, but there wasn’t anything about solar on the site.
It's not for gaming at all, hasn't even got a monitor. It's a link to an Asus device that allows you to stack 48TB of space on 12 M.2 drives, compatible with TrueNAS, and the entire rig runs fine on half a cheap solar panel's worth of power generation in direct sunlight. Or, it'll run for about 7 hours on a solar-charging battery pack like this. No moving parts, and as long as it is powered on once every three years or so to refresh the M.2 cells it should last a couple decades. It's for storing Netflix in the palm of your hand and with room to spare.
Ah, that’s really cool! Shame on me for not actually reading the link AND having the nerve to confidently call it out!
To my credit, it looks very similar to my MSI Trident 3, so I saw it and just assumed it was a direct analogue.
Oh. My God. You just unlocked a new horror for me. Like I like using chat bots for trying to understand the current scientific understanding of their harms, and they also came in clutch this weekend finding cat feeders for my family and reducing my parent's cable bill. But this thought just makes me want to move to a cabin in the woods and live without electricity.
I noticed that the voices still sound off. But the video itself is hard to tell apart from real footage depending on the scene (well, unless it has subtitles).
Then again, I remember only a few years ago language models didn't seem like they would produce text as well as they do now either. So I wouldn't be surprised if soon it will be indistinguishable from real footage.
I can't help but wonder how well it deals with a larger narrative though, and how feasible it is to make it do so. It is rather challenging for AI to remember details of previous prompts. So you might just end up with video in which the main character randomly looks like a different person depending on the scene. So you'd not be able to replace a film crew entirely unless you resolve that.
For videogames I think these days we make them inside the game engine. It used to be that games would use pre-rendered video for cutscenes but it was always weird and immersion breaking if the art style suddenly changes when going to a cutscene. It looks more natural now. Going back to pre-rendered video would seem like a step backwards, whether AI is involved or not.
The only easy tell left is words - it still consistently messes up text rendered on signs for example. I wonder why it has such a hard time with lettering. I'd say the skin tone is also a bit of a giveaway, because real humans rarely have skin that's so consistently smooth and featureless unless they've just come out of a studio makeup room.
My instinct is because it’s not trying to write words, it’s trying to paint word-shapes. I imagine if big road signs were common further back in time, impressionist paintings might give us a similar vibe, with road markings being messy and unintelligible and yet perfectly give the impression of a road sign from a distance
Every time I see this type of generative video, I can't help but notice how they still aren't good enough. Maybe for a low effort YouTube video, but we are a long way off from something like this replacing a VFX artist on a major Hollywood production. It's good but it's clearly not real. It still hasn't gotten a lot of the details right and doesn't really hold up to any sort of scrutiny.
I do think it's just a matter of time before we get over that hump. It could be 5 years, it could be 10, but we will get there eventually. There's too much money for corporations to make by just using AI instead of paying real artists. That sucks.
It's also just a matter of time before these are truly weaponized and used for political purposes. I mean, if he could, don't you think Trump would just fabricate videos of his enemies to use as evidence to lock them up? I could easily see his regime doing something like generating a video of Kilmar Abrego Garcia saying he's in MS-13 and he hates the U.S.?
I expect lots of slop, but the flip side of that might be indy movie or TV show makers that use it to make something interesting. For example, South Park animation is pretty crude, but they did manage to connect with people.
I wouldn’t be surprised it if happens by the end of the year. Run Veo 3 for the rough animation, then have an actual artist do the finishing. In theory this could mean the ability to make more movies (and thus, no job losses), but in practice there’s an upper limit to the number of movies, and especially the number of good movies.
It doesn't have to get to a point that it's good enough for a major Hollywood production, although I still wouldn't bet against it. I'm confident we'll see this adopted in the mainstream for little tweaks to real media. Imagine if the adult film industry gave their consumers the ability to tweak videos a little bit. Extend this scene for a little longer. What if the cameraman didn't cut away to this angle and instead went to this angle? Augmenting existing media with realistic enough additions seems attainable within the next couple of years.
Eventually as these tools are adopted some people will not pursue their passion into a career... because they will just think "why bother?".
Like a self fulfilling prophecy these tools will be used because they are good enough, cheap... and because some people will have just give up deleting the human alternative.
This isn't new, my dream career is making physical props and effects, with my inspiration being Jim Henson when he does the darker stuff. However I graduated high school in 2013, CGI already took over so I was in this scenario of "why bother?"
I have a friend that is 5+ years older and he got a college degree for this kind of stuff, but all its good for now is to make him feel more legit as an local artist and it got him a job making toys for Hot Topic 10+ years ago when they sold that stuff.
(This paragraph is a poorly phrased rant, feel free to ignore)
I really wish we could find a better balance between taking advantage of new technology and still using the old styles when they are appropriate. Unfortunately the bottom line is its cheaper and easier to use AI/CGI, so thats the decision that gets made. If we could look past the bottom line to actually making something creative, and letting their vision get made then I think there would be a lot more of the old styles would still be done. But instead original ideas seem like a thing of the past and its all live action remakes or an nth level sequel for a series that really should have died long ago.
I’m expecting most commercials to use this within a few years. Movies will use this as the new green screen in places. But I can’t see audiences being quick to tolerate AI actors.
They won't know. It'll start with mostly live performance, fixed up/edited with AI. Over time it'll shift towards less performance, more AI. I predict an article in x years that says "did you know this award winning scene was actually made entirely with AI?" And people will be shocked, even though by that point it'll be standard.
I can definitely see a lot of video games making use of it along with other generative art models, and coding agents as well. They all substantially lower the barrier to entry (and cost) for making games. But yeah, I don't see serious actors in films and TV shows being totally replaced by AI generated ones any time soon either.
Agreed. I wonder what new kinds of movies enthusiasts will make. Remember Red versus Blue? Machinima (making movies with video game engines) was an interesting trend for a while.
Or maybe people will make something shorter and less coherent like music videos.
Music videos with AI visuals are already out there. It's a medium where a series of short cuts are appropriate, and variety of imagery can be accepted as long as it's giving the right vibe. It's a great fit even before it can give long term narrative coherence.
I'm doubtful this will hit professional games anytime soon. Unlike movies, video games need very specific assets both to let them animate properly and not bog performance down to a crawl. Most optimization problems in modern games comes from the art pipeline: specifically, lighting and model complexity. This is a big part of why all the AI slop flooding asset stores is basically useless for any serious game.
Sure, maybe some cutscenes can be made with AI. But it's hard to take shortcuts on anything you'll control or interact with as of now. It'll probably linger for 5-10 years on what's currently out there: generative art workflows to spam out skins, or for environmental textures.
I think AI's going to make larger changes to the gameplay than the art assets at first. I'm looking forward to The Wayward Realms. It's the old Daggerfall / Morrowind / Oblivion developers that left Bethesda, so it's basically going to be Elder Scrolls done right minus Todd Howard's creeping crud. What really interests me is the level of AI baked in - like the old 'radiant' engine from Bethesda games but this time as an active game master managing 40 (yes 40) factions across the game world, all reacting to the player, generating quests, dialog, and NPC daily routines. That should take RPGs to a whole new level of replayability.
I've been seeing these awful ads on social media that are 100% AI. Mostly testimonials from the "person" speaking. One was even for a over-the-counter medication, which should probably be illegal, right?
Can they create clips long enough (or separate clips with consistency of content) that you can make a full commercial or longer media yet?
I don't know that much about image generation AI, but the other day I was trying to think through how you would confirm the authenticity of video if AI video tools like this become widespread...
One idea I had for recording confirmably authentic video of something would be to set up multiple cameras at different angles. I think it would be pretty immediately apparent if two videos of, say, a person talking and gesturing, didn't match each other - spatial perception of human bodies and faces is something that humans are particularly good at. And it strikes me as a difficult task to not just generate a realistic video but multiple realistic videos of the same event that are all spatially coherent with each other.
But like I said, I'm not an expert on the topic and maybe someone who is could weigh in on this.
In the videos I've seen (such as the water bottle person video), it's not even consistent between shots. If you watch the conference room scene that starts around 1:15, it can cut away and back to a consistent person, but each shot seems to be a different cast of characters. It's particularly noticeable if you try and figure out who's at the head of the table or look for any of the people from up-close shots in the wider shots. So I suspect your plan would work even better and reveal discrepancies much larger than just minor spatial irregularities.
I wouldn't be surprised if merely doing a slow 360 would still be enough to make AI video come up short. In earlier AI videos, it seemed to me like it only mattered that the new stuff panning into view matched what was on-screen. So doing a slow 360 at a public event would give you a whole new crowd of onlookers. I really hope these "problems" don't get solved soon.
I agree these "AI actors" do look way better than ones we've seen before. I'd like to see a blind video to be honest because I feel like I'm seeing weirdness watching the clips in this thread but I'm not sure if it's because I know they are AI and I'm looking for flaws or not.
I'm also curious how easy these clips are to direct honestly. Like for any creative trying to direct, is the expectation you are going to sit there for hours and hours regenerating a clip with slightly different prompts to try and get the exact next sequence you want?
Also if the rest of my small experience of AI image generation has anything to go by, their controls like the camera controls (jib up) are going to ruin a perfect scene. You might get a clip you like except the "camera" is in the wrong place, try to move it and the whole scene completely changes. Or changes enough that it's invalidated what you're looking for.
Sounds incredibly tedious IMO.
It certainly was tedious when I tried to get Midjourney to create halfway decent images a couple years ago, but my understanding is that AI-based image editing tools are improving.
OpenAI brags about their latest image generator here. I tried it myself and it did seem much better, but it’s also very slow at generating each image, so in the end it’s still tedious.
Mark my words, our grandchildren will be subscribing to movie and TV AI services where instead of Netflix they'll be sharing prompts and seeds of TV shows amongst themselves. People will be able to tailor the humor to their own sense of humor and tailor the acting and actors to their own personal favorites.
If they don't like an ending, they can generate a different one.
People won’t just search for a title, they’ll input a prompt. They’ll share “seeds” of shows with each other; starting points that generate unique variations. Humor, tone, pacing, themes, settings, and character archetypes will all be adjustable. A viewer who prefers dry, sarcastic humor can tweak a show accordingly. Someone else might want the same plot but with more emotional weight or different music. Acting styles and even the visual appearance of characters could be changed to suit personal taste.
There will be completely AI generated content, and content that is guided and directed by humans but generated by AI.
People will be generating Season 2 of Firefly, as well as re-generating a better season of Season 7/8 of Game of Thrones. There will still be collaborative or standout “runs” that gain traction, like a particularly well-regarded 10-season continuation of Firefly or a widely shared reimagining of Game of Thrones. But the core model is individual consumption and customization. It’s not about contributing to a central canon, it’s about creating your own.
No more canceled shows. If a show's not good enough people can fix it themselves. People will be sharing their tweaks and adjustments, which will be rated.
Infinite sequels, evolving worlds, infinite spinoffs. There will be no limits.
You'll have a basic plot, but then variations depending on taste. Think a "Spy Thriller" and the variations are "Set in the Star Wars Universe" or "Set in the Flintstones Stone age" or "Animated in the style of Rick and Morty".
Going to be honest: what you've just described sounds like a nightmare to me. I'm not sure I could live in such a world.
Yeah I'm not describing it as a good thing, but just where we're headed.
However I'm sure there are many things in today's world that many in the early 1900s would have considered a nightmare today. Haha Probably why we have so many boomers refusing to give up positions in governments and going crazy about a world they'll probably never get to see, because they too think this world's a nightmare.
I'm not convinced the entire genre would survive that. It seems like something that could actually kill the idea of watching tv by fast forwarding to the end where it's all been done. I could see some people becoming infinitely addicted and ruined by such a thing. But I could see a lot of people burning out on it in a few months.
I could very easily be wrong. I just think sitting back an observing without directing will always be a necessary part of it. I think it's more likely that you'll see AI settings created by "show runners", and systems of delivery that respond to the user live. Using biometrics or similar to decide if the viewer is responding positively or negatively to humor, politics, sci fi, etc.
So less hands on than sharing prompts that you have visibility and control of, and more lining up to the content feed that serves automatically adjusting content designed for maximum addiction. Just as dystopian, though. I don't want to see me kids sticking electrodes to their temples so they can activate AI mode on their TV because it's the only thing they enjoy any more.
I mean I think what you said is what it will be like at first, but eventually.
I mean it's not much different an idea as the computers behind Holodecks in Star Trek.
At this point I'm really tempted to just use a generative model to create video content, novels, a webcomic, a website, etc. If AI has taken over to the point where some shitty algorithm is preventing me from going into another accountancy role (whether through some bullshit automated screening process, or because ChatGPT can crunch numbers better than me), then it's only a matter of time before more people are displaced.
Is this why humans will get the brain chip, to keep up with Ai?