Do you know that you can configure Google to filter garbage? Take these steps For the best search results, including the addition of my work in Lifehacker as a preferred sourceField
Last year, I wrote that we should all be afraidOpenai's Ai Video Generator. The initial deployment of Sora promised hyperrealistic videos, which, although fascinatingly, frightened me. While fans of AI see the future of films and show generated AI, I see a future where no one can say that it is real or falseThe field for me is the only destination for this technology is mass misinformation.
Since a year and a half, these videos generated by AI have not only become more realistic; They have also become more affordable as companies such as Google Make your tools easily accessible to everyone who wants to payThe field is the situation in which we are with the latest announcements of Openai: Sora 2A new model of artificial intelligence for creating a video with audio, as well as a new Sora application for creating and exchange of products created by AI.
Sora 2
Openai is the marketing of Sora 2 as a massive update by Sora-usual with GPT-3.5 and GPT-1, respectively. The company says that the new model can generate complex videos that could not. This includes, in particular, Olympic gymnastic routine; A person who performs a reverse flag on the Paddleboard, which “accurately” models the physics of water; And also the skater performing a triple axel with a cat on his shoulder.
This tweet is currently not available. It can be loading or removed.
One common drawback with video models of AI Their lack of understanding of real physics. Visual may look realistic, but the elements can accidentally transform together, while others can disappear and appear without rhyme or mind. Openai says that Sora 2 does not make these mistakes so often. Basketball, which misses the hoop, cannot appear there magically there; Instead, he will bounce off the sworn establishment, as expected. The company warns that the model is still imperfect, but improved. Based on this, the model is better in continuity in different pictures: taking Openai in your own word, your videos should maintain a sequence between the takes, and you should be able to dictate various types of styles, including “realistic”, “cinematic” and “anime”.
Perhaps the largest jump with Sora 2 is the ability to add elements of the real world to the model, the Openai function calls the “Kame”. You can place real people in the Sora 2 model and ask AI to create them in any video you want. Openai shows a number of examples of how their employees add themselves to various videos, and although the quality is inconsistent, this is a giant leap from Jibjab daysField
Like the Google Veo 3, Sora 2 model can generate a realistic audio video. Video ads Shows this: the roar of the elephant; skater on ice; Water spray on earth. But more impressive (and in relation to people speakThe field Sam Altman, generated by AI, explains the new model and the application in this video, and although this is quite obvious to those of us who know that this is AI, I can imagine many People would not know that this was not a real altman in the video.
Nursing application
Openai says that the Sora application has become the “natural evolution of communication.” The company considers this a way for people to create remix of other users, especially with the ability to load their own face and similarity with the model.
At the moment, the application is only invited, although You can download it for free in the app store todayYou can get an idea of experience, however, from both Demo -Video Openai fell on Tuesday, as well as posts from people who already have access.
This first example of Demos Openai has a double camera of the Openai scientist Bill Pinches and Sam Altman. The video contains the established picture of two men who lead a conversation, which approaches a close plan, when the pipe quickly speaks of the income of the application, and then to the close plan of the altman, which takes the ranting before closing the original shot. At first glance, this is the type of video that you can expect to scroll by on tiktok or coils, but this video is completely generated by AI.
Openai employees demonstrate a series of other pre -generated examples, including a cameo that turns into a cartoon, the other that shifts the effect on the anime, and the other that generates the news report on one of the drug addicts for ketchup. (The latter, I could add, this is rather rudely.) They also demonstrate a remix video that you will find in the channel, as you can ask the Soru to configure the video as you want. One video shows the piblose in the “announcement” for the Sora 2 cologne, but others remicked it so that it was from a toothpaste or completely in Korean.
These videos are quite realistic: in one thing, you think that you just watch a clip of a tennis match, but it turns out a camera with Rohan Sahai from Openai. After the “Sakhai” wins the match, the video moves to his “interview” in which he thanks the haters. Others are more obvious that AI – although, again, it is not enough for most people to scroll, perhaps notes.
Safety, according to Openai
Cameurs sound like confidentiality and a nightmare, although Openai has some protection. You can’t just use whose face is a face for any video, and you can download your own face to the platform. The setting of the Cameo function in the application is simple, if it does not push it out extremely. The application will scan your face, like setting up the face identifier on the iPhone, and then send data to the Openai system for “tons of validation” for blocking imitators, or users who may want to create cameras without your consent. After approval, you choose who can create camera, including all users, friends, users whom you specially approve of, or only you.
What do you think so far?
As for the video themselves, the Sora application applies a visible watermark to any clip exported from the application. If you have already seen which one from these videos on the Internet, you will notice a small brand “Sora” on each, similar to the water sign that you see on the Tiktok clips exported to other platforms. There are also models of reasoning under the hood to block users from creating “harmful” content, especially in relation to cameras.
If you are a teenager using the Sora application, you cannot scroll forever. After a time of scrolling, there will be a reload time to prevent you from spending hours of scrolling these videos with artificial intelligence. While adult accounts will not have this restriction, the application “pushes” you to take a break.
Who asked for this?
With all due respect to Openai and his security team, this application seems to be a disaster for many reasons.
On the one hand, Openai made it as easily to generate hyperrealistic videos with a short form, as well as ask Siri about the weather. I appreciate that all these videos are supplied with watermarks, but they do not need a lot of skills to edit them – at least, as most people will not notice. As soon as this is widely available, all our channels on social networks will suffer from this content. And see how most of this goes with the video And Audio that are quite realistic, many people will be fooled many content.
This is bad enough when it is connected with stupid videos, like rabbits jumping on a trampolineThe field but what happens when the “politicians” say something yelling or “celebrity”, steals something from the store? One viral video Shows Sam Altman, trying to escape with a graphic processor in TargetBefore stopping the guard. How many more videos Sora will show Sam Altman, and someone else, who approves their cameras, is committed, committing crimes or just doing something embarrassing? Those who have sufficient strength or glory can expose the video, but by that time it will be too late: most people who have seen this will perceive it as a fact.
This tweet is currently not available. It can be loading or removed.
At this point, it is remarkable that there are safety measures to prevent people from making remixes on cameras of other users without permission, but the risk of abuse here supreme: What will happen if someone finds out how to “scan” someone’s face with a video, or hack the settings that block others from using their initial face scan? If they can bypass Openai safety measures, they can then redo this person’s face in any video approved by the platform. At this moment, the cat leaves the bag.
Look, I'm chronically online. I am not going to pretend, as if I do not like a good meme generated with AI when he encounters my food. But I'm not going to spend my free time scroll Nothing But the generated AI brain is rot. I am sure that people will find creative ways to make funny videos using Sora or have a good time, making cameras with their friends, but this is the point: in addition to the simple novelties of technology, there is nothing to do from this.
It's time to stop believing in something You see on the Internet: someone could just cook it in the application.
