I've got two main thoughts here (which I'm debating writing more about):
1. It's interesting how OpenAI and Meta are targeting different parts of social media's 90/10/1 funnel (90% of users consume, 10% engage, 1% create). The Vibes app is after the 90% who will be consuming AI generated content, while the Sora app is closer to the 10% who will edit/remix friends' content.
2. Hey, maybe don't voluntarily put your likeness into an AI-cloning app. Maybe I'm being paranoid about this, but it reminds me of early 23andMe users - in the end, their genetic data was nearly sold to the highest bidder as the company went bankrupt.
I love AI as a technology. So many amazing use cases… Video generation isnt one of them. I honestly do not see any social utility here with unlimited downsides. LLMs can help improve code security, theyre an LSP on steroids, aid in translation. Other models are great for translating languages. China uses machine vision to create agriculture drones so that humans do not have to work in dangerous heat.
The video generators just seem to be very good at making racist propaganda and deepfake pornography and identity theft
I mean, I can see at least several viable "non-crappy" use cases for AI video:
1. As a tool for indie filmmakers to tell stories and make short films. I've seen several great examples.
2. Video models can eventually become "world models" that can simulate different visual environments and scenarios, which could in turn be used to train robots and other hardware that needs to interact with the real world.
But you're right in that the most mainstream use cases seem to be memes and slop, which is why I feel it's such a shame that OpenAI went with the "social app" positioning.
In the end, it all comes down to the user. LLMs can also be used to code harmful malware, provide people with instructions for doing something illegal, pornographic written content, or simply generate low-effort filler slop for content farms.
Every technology is as good as the applications we find for it!
1. Yeah, for sure. OpenAI is trying to make it all about prioritizing creation vs. consumption. Still, I'm not sure "obsessively making memes" vs. "obsessively consuming memes" is enough of a distinction to matter in terms of the impact endless short-form content has on us.
2. True. I mean, I'm definitely tempted to try it out, and I have previously used myself as a reference/character image in image models and image-to-video models, but that's still fundamentally different from creating a "Cameo" anyone can reuse. Yes, there are controls in place to make your "Cameo" private or just for friends, etc., but that doesn't at all negate the "selling data to third parties" aspect you're talking about.
Generally agree with your take- not sure how all the $$ and the brilliant minds feel this is the great path forward for humanity and AI. Heck , even just for generative AI, is it just another tool for doomscrolling dopamine hits?
I came to you just after posting my little reaction on this, looking for what others might have to say! :)
Nice video, and I enjoyed your tongue-in-cheek sprinkling of humor here and there.
It sounds like we very much agree on this. I love the promise of what AI might do for the individual creative person, provided we can solve the legal and ethical guidelines around it.
But I absolutely don't love that "press a button, get a recycled meme" is the direction we're explicitly pushing this.
And yes, we have way too many short-form dopamine scrolling machines already.
Yes I feel like the inevitability of genAI brushes away the nuance we need talking about it at this stage. I’m pretty conflicted about it and generally have stayed away from using it after some initial dalliances.
As for the gen AI social media or Slop Social, hard time convincing me of its value. The distraction machine and the attention wars have enough going on already.
Totally agree with this take. The videos themselves are cool and represent the next phase in video generation - but the free launch is also a Trojan Horse. Get folks used to making crazy realistic videos of themselves with their friends doing wild things and then the bait and switch. My question is whether this will eventually be moved over to Plus and Pro users or an extra upcharge. What's so clear, and someone else on Substack (I think it was Nate) pointed this out, is how much OpenAI is just looking for a way to bring ads into their models. The financials are too daunting - subscriptions alone are not going to make the ledger balance.
Even without the bait-and-switch direct monetization (whether via ads or premium features or access to paying users only), getting mainstream audiences sucked into yet another social app feels iffy to me. OpenAI is pretty explicit about the "free" period being time-limited, so I do expect this to eventually roll over to Plus and Pro accounts.
You could make the argument that at least chatbots could present ads only when relevant and seamless in a chat rather than interrupting the experience like current Facebook feeds and Google search pages.
Then again, that's potentially even more insidious. At least with clearly demarcated ads on Facebook and Google, we can develop banner blindness and just learn to ignore the ad. But what happens when an ad is subtly baked in as a gentle recommendation inside a seemingly neutral answer from a chatbot?
One can hope there'll be clear laws and regulations in place for how ads are marked, but I'm not overly optimistic in the short term.
Remember Qik? Of course you don’t, it was Skype’s attempt to be relevant in social and compete with the likes of Snapchat. We were trying to shine Skype up and OpenAI is at the vanguard, but I’ve heard (rumor?) that Sam wants a social media platform and is jealous of influence Musk has with X. There was a similar sounding app created by a guy who left substack. I installed it for a bit and was entertained by the novelty — specialty was AI cameos and you could add in friends. Its name was verb-y and so you could effectively re-stack or re-tweet or re-something. Novelty wore off and I can’t even find it on my phone anymore. So.
Never heard of any of those, but I absolutely see why companies make this play: Virality is the fastest way to adoption. I just wish this didn't simultaneously undermine the "we're fundamentally changing the world with smart and capable models" message.
Yeah the social app thing is a bit weird. But they said it will soon be available on the API so then anyone can use it I guess - outside of the social network.
Not sure the VPN thing works on the iPhone - I'm French so I have access only to the French Appstore, regardless of how I access internet with the phone. If you have a way to do it, I'm interested!
Yeah, for sure. There will be an API and perhaps even a standalone platform down the line. But my concern is about the short-term launch, how it frames the model, and what this tells us about OpenAI's priorities.
I don't have an iPhone either, and I'm in Denmark, so I've been using Sora 2 directly via the browser: sora.com. See if that works for you!
I've got two main thoughts here (which I'm debating writing more about):
1. It's interesting how OpenAI and Meta are targeting different parts of social media's 90/10/1 funnel (90% of users consume, 10% engage, 1% create). The Vibes app is after the 90% who will be consuming AI generated content, while the Sora app is closer to the 10% who will edit/remix friends' content.
2. Hey, maybe don't voluntarily put your likeness into an AI-cloning app. Maybe I'm being paranoid about this, but it reminds me of early 23andMe users - in the end, their genetic data was nearly sold to the highest bidder as the company went bankrupt.
I love AI as a technology. So many amazing use cases… Video generation isnt one of them. I honestly do not see any social utility here with unlimited downsides. LLMs can help improve code security, theyre an LSP on steroids, aid in translation. Other models are great for translating languages. China uses machine vision to create agriculture drones so that humans do not have to work in dangerous heat.
The video generators just seem to be very good at making racist propaganda and deepfake pornography and identity theft
I mean, I can see at least several viable "non-crappy" use cases for AI video:
1. As a tool for indie filmmakers to tell stories and make short films. I've seen several great examples.
2. Video models can eventually become "world models" that can simulate different visual environments and scenarios, which could in turn be used to train robots and other hardware that needs to interact with the real world.
But you're right in that the most mainstream use cases seem to be memes and slop, which is why I feel it's such a shame that OpenAI went with the "social app" positioning.
In the end, it all comes down to the user. LLMs can also be used to code harmful malware, provide people with instructions for doing something illegal, pornographic written content, or simply generate low-effort filler slop for content farms.
Every technology is as good as the applications we find for it!
1. Yeah, for sure. OpenAI is trying to make it all about prioritizing creation vs. consumption. Still, I'm not sure "obsessively making memes" vs. "obsessively consuming memes" is enough of a distinction to matter in terms of the impact endless short-form content has on us.
2. True. I mean, I'm definitely tempted to try it out, and I have previously used myself as a reference/character image in image models and image-to-video models, but that's still fundamentally different from creating a "Cameo" anyone can reuse. Yes, there are controls in place to make your "Cameo" private or just for friends, etc., but that doesn't at all negate the "selling data to third parties" aspect you're talking about.
Generally agree with your take- not sure how all the $$ and the brilliant minds feel this is the great path forward for humanity and AI. Heck , even just for generative AI, is it just another tool for doomscrolling dopamine hits?
I came to you just after posting my little reaction on this, looking for what others might have to say! :)
https://substack.com/profile/87890550-abhishek-rao-shakey/note/c-163813180
Nice video, and I enjoyed your tongue-in-cheek sprinkling of humor here and there.
It sounds like we very much agree on this. I love the promise of what AI might do for the individual creative person, provided we can solve the legal and ethical guidelines around it.
But I absolutely don't love that "press a button, get a recycled meme" is the direction we're explicitly pushing this.
And yes, we have way too many short-form dopamine scrolling machines already.
Thanks Daniel.
Yes I feel like the inevitability of genAI brushes away the nuance we need talking about it at this stage. I’m pretty conflicted about it and generally have stayed away from using it after some initial dalliances.
As for the gen AI social media or Slop Social, hard time convincing me of its value. The distraction machine and the attention wars have enough going on already.
Totally agree with this take. The videos themselves are cool and represent the next phase in video generation - but the free launch is also a Trojan Horse. Get folks used to making crazy realistic videos of themselves with their friends doing wild things and then the bait and switch. My question is whether this will eventually be moved over to Plus and Pro users or an extra upcharge. What's so clear, and someone else on Substack (I think it was Nate) pointed this out, is how much OpenAI is just looking for a way to bring ads into their models. The financials are too daunting - subscriptions alone are not going to make the ledger balance.
Spot on!
Even without the bait-and-switch direct monetization (whether via ads or premium features or access to paying users only), getting mainstream audiences sucked into yet another social app feels iffy to me. OpenAI is pretty explicit about the "free" period being time-limited, so I do expect this to eventually roll over to Plus and Pro accounts.
And yes, I absolutely think that "ads" are coming. To both these visual SoMe apps but also the chatbots themselves. Perplexity was already talking about ad experiments in late 2024 (https://www.perplexity.ai/hub/blog/why-we-re-experimenting-with-advertising).
You could make the argument that at least chatbots could present ads only when relevant and seamless in a chat rather than interrupting the experience like current Facebook feeds and Google search pages.
Then again, that's potentially even more insidious. At least with clearly demarcated ads on Facebook and Google, we can develop banner blindness and just learn to ignore the ad. But what happens when an ad is subtly baked in as a gentle recommendation inside a seemingly neutral answer from a chatbot?
One can hope there'll be clear laws and regulations in place for how ads are marked, but I'm not overly optimistic in the short term.
Remember Qik? Of course you don’t, it was Skype’s attempt to be relevant in social and compete with the likes of Snapchat. We were trying to shine Skype up and OpenAI is at the vanguard, but I’ve heard (rumor?) that Sam wants a social media platform and is jealous of influence Musk has with X. There was a similar sounding app created by a guy who left substack. I installed it for a bit and was entertained by the novelty — specialty was AI cameos and you could add in friends. Its name was verb-y and so you could effectively re-stack or re-tweet or re-something. Novelty wore off and I can’t even find it on my phone anymore. So.
Perhaps it’s a play for GenZ?
Never heard of any of those, but I absolutely see why companies make this play: Virality is the fastest way to adoption. I just wish this didn't simultaneously undermine the "we're fundamentally changing the world with smart and capable models" message.
I believe ChatGPT still holds the virality record for fastest adoption by a long shot
OpenAI is calling Sora 2 the "GPT3.5 moment for AI video," so let's see.
Yeah the social app thing is a bit weird. But they said it will soon be available on the API so then anyone can use it I guess - outside of the social network.
Not sure the VPN thing works on the iPhone - I'm French so I have access only to the French Appstore, regardless of how I access internet with the phone. If you have a way to do it, I'm interested!
Yeah, for sure. There will be an API and perhaps even a standalone platform down the line. But my concern is about the short-term launch, how it frames the model, and what this tells us about OpenAI's priorities.
I don't have an iPhone either, and I'm in Denmark, so I've been using Sora 2 directly via the browser: sora.com. See if that works for you!