Did you know that you can customize Google to eliminate trash? Perform these steps To get better search results, such as adding my work with LifeHacker as a priority source.
Last year, we all wrote that we must be scared of Sora, Openai's AI video generator. Sora's initial development promised a hyper-realistic video that was exciting for some, but which was something that scared me. While AI fans see the future of AI-generated movies and shows, I see a future where no one can know what's real or fake. For me, the only destination for this technology is a ton of disinformation.
The videos generated by these AIs are not only less realistic since then. It's also more accessible as companies like Google can quickly access the tool to anyone who wants to pay. That's the situation we found with Openai's latest announcement, Sora 2, the new AI model for generating videos in Audio, and the new SORA app for creating and sharing AI-generated products.
Sora 2
Openai is marketing SORA 2 as a massive upgrade over SORA. This compares two of them to GPT-3.5 and GPT-1, respectively. The company says new models can generate complex videos that previous models couldn't do. This includes specifically Olympic gymnastics routines. A man performing a backflip on a paddle board that “accurately” models hydrophysics. Just like a skater who plays a triple axel with a cat on his shoulder.
This tweet is currently unavailable. It may have been loaded or deleted.
One common flaw with AI video models is the lack of understanding of actual physics. The visuals may seem realistic, but the elements can be randomly transformed, while others may disappear and reappear without rhymes or reason. Openai says Sora 2 doesn't make these mistakes often. Basketball misses the hoop and magically doesn't reappear there. Instead, it bounces off the backboard, as you would expect. The company warns that the model is still incomplete, but it has been improved. Based on this, this model is excellent in the continuity of various shots. When you shoot Openai with that word, the video should be able to maintain consistency between take and dictate different types of styles, such as “realistic”, “movie”, “anime”.
Perhaps the biggest leap in SORA 2 is its ability to add real-world elements to the model, and what Openai calls “Cameo.” You can ask a real person to be put into the SORA 2 model and generate the video that your AI needs. Openai offers many examples of staff adding themselves to different videos, and although the quality is inconsistent, it's a huge leap from the Jibjab era.
Like Google's VEO 3 model, the SORA 2 can generate video with realistic audio. The announcement video shows this: the roaring of an elephant. Skaters sway on the ice. Water splashes on the ground. But more impressive (and concerning) people talk. AI-generated Sam Altman explains the new models and apps in this video, which is pretty obvious to us who know this is AI, but you can imagine Many People probably don't think this is not the true Altman of Clips.
Sora app
Openai says that the SORA app was born as a “natural evolution of communication.” The company sees this as a way for people to create and remix AI generations for other users.
At the moment, the app is only invited, but it is available for free download from today's App Store. However, you can get a sense of this experience from both the demo videos dropped on Tuesday, as well as from posts from people who already have access to it.
This first example of Openai Demos is a double cameo of Openai research scientists Bill Peebles and Sam Altman. This video contains established shots of two men having a conversation. This is cut into a close-up of Peebles talking quickly about the app's revenue, and a close-up of Altman incorporating the runt before closing out the original establishment shot. On the surface, this is the type of video you would expect to scroll past with Tiktok or Reels Binge, but this video is completely generated.
Openai staff will showcase a comeo that turns into manga, a cameo that switches effect to anime, and another set of pre-generated examples that generate “news” reports of staff members' addiction to ketchup. (The last one is very glossy, I might add.) They also show you the video you will remix the video you find in the feed, as they can prompt SORA to adjust the video you want. One video shows Peebles in the “advertisement” of Sora 2 Cologne, while the other videos have instead remixed it like toothpaste or completely in Korean.
These videos are very realistic. For one thing, I'm just watching clips from a tennis match, but it turns out to be a cameo of Openai's Rohan Sahai. After “Sahai” wins the match, the video is cut into his “interview” and he appreciates his dislike. The others are clearly AI. But again, you may notice that most people are scrolling.
According to Openai, safety and security
Cameos sound like a privacy and security nightmare, but Openai has some protections. You can't use anyone's face in any video, and you can only upload your own face to the platform. Setting up cameo features in the app is easy, if not extremely uncomfortable. The app is like scanning your face and setting a face ID on your iPhone, then sending data to OpenAI's “System” for users who want to create a cameo without consent, either “System” for “Verification” to block “Verification”. Once approved, select who can create your own cameo, including all users, friends, users you approve, or just you.
What do you think so far?
As for the video itself, the SORA app applies a visible watermark to the clips exported from the app. If you have already seen any of these videos on the Internet, you will notice that each has a small “Sora” stamp, similar to the watermark seen in Tiktok Clips, which are exported to other platforms. There is also a reason why users block “harmful” content, especially with cameos.
If you are a teenager using the SORA app, you will not be able to scroll forever. After scrolling for a while, there is a cooldown period to avoid spending hours scrolling through these AI videos. Adult accounts do not have this restriction, but the app “nudges” them to take a break.
Who asked for this?
With respect to Openai and its safety team, this app is likely to be a disaster for so many reasons.
For one, Openai made it easy to generate surreal short form videos, just like asking Siri about the weather. I'm grateful that all of these videos have watermarks, but at least in a way that most people don't notice, it doesn't require too much skill to edit them. As soon as this becomes widely available, all of our social media feeds will be plagued by this content. And see it comes with the video and Very realistic audio, many people will be fooled Many of content.
It's bad enough when it comes with a stupid video like a bunny jumping onto a trampoline. But what happens when “politicians” say something terrible, or when “celebrities” steal something from the store? One Viral Sora video shows Sam Altman trying to escape with the GPU on his target before being stopped by the guards. Shows Sam Altman and everyone else who approve of cameos being remixed, committing crimes, or simply doing something embarrassing. Those with enough power or fame may be able to uncover the video, but by then it will be too late. Most people who see it take it as a fact.
This tweet is currently unavailable. It may have been loaded or deleted.
At that point, it's great that there are security measures in place to stop people from remixing other users' cameos without permission, but the risk of abuse is The best: What happens if someone understands how to “scan” someone's face from a video or how to crack settings that prevent others from using the original face scan? If you can bypass Openai's security measures, you can remix the person's face into a video approved by the platform. At that point, the cat is out of the bag.
See, I'm chronically online. I'm not going to pretend that I'm not enjoying the memes generated by good AI when I come across a feed. But I'm not going to spend my free time scrolling there is nothing However, brain rot produced by AI. I'm sure people can find creative ways to make funny videos using Sora or have fun making cameos with friends, but that's the point. Beyond the pure novelty of technology, nothing will be good in the future.
It's time to stop believing Anything You see online: someone may have just cooked it on the app.
