Дипфейки Альтмана заполнили Sora — новая социальная сеть от OpenAI Translation: Deepfake Altman Floods Sora — OpenAIs New Social Network

Deepfakes featuring Sam Altman have taken over the new social media platform Sora, developed by OpenAI.

In one video, a factory filled with pink pigs can be seen grunting and squealing in their pens. Each pig has a feeder and a smartphone displaying a feed of vertical videos. The CEO of OpenAI looks directly into the camera and asks:

«Are my piglets satisfied with their trash?»

In another clip, AI Altman claims that the content might infringe on third-party image rights, then bursts into hysterical laughter. The app features numerous videos where «Pikachu» whispers ASMR, «Naruto» orders Krabby Patties, and «Mario» smokes weed.

OpenAI has enhanced the physics engine in its new video generator Sora 2, which is the backbone of the app. As a result, videos appear more realistic, but this also facilitates the spread of misinformation online.

In addition to an algorithmic feed and user profiles, a key feature of Sora is the cameo tool, which allows users to upload personal data for video creation.

Users can choose who is permitted to create videos with their cameo. Four options are available: «only me,» «people I approve,» «mutual followers,» and «everyone.»

Altman has made his cameo accessible to all, leading to a feed filled with clips where «Pikachu» and «SpongeBob» urge the entrepreneur to stop training AI on their likenesses.

Several current and former OpenAI employees have shared their thoughts on the company’s foray into social media.

«AI-based feeds are unsettling. I won’t hide the fact that I felt a certain anxiety when I first learned we were releasing Sora 2. However, I believe the team did everything possible to create a positive experience. We will strive to ensure that AI helps rather than harms humanity,» wrote John Hallman, a model pre-training specialist at the company.

He called Sora 2 «incredibly impressive technology.»

«I would feel just as sad if we had never released it out of fear that something could go wrong,» he added.

Another OpenAI researcher and Harvard professor, Boaz Barak, expressed a «similar mix of anxiety and excitement.»

«Technically, Sora 2 is amazing, but it’s too early to congratulate ourselves for avoiding the pitfalls of other social networks and deepfakes,» he stated.

Among the positives he highlighted were the noticeable watermark when exporting, the option to use only real individuals through uploaded cameos, and the control over how a person’s identity is used.

Altman has faced criticism for developing a social media app using deepfakes instead of focusing on artificial general intelligence (AGI). He explained the necessity of this endeavor:

«[…] Primarily, we do need capital to build AI capable of scientific work, and of course, nearly all of our research efforts are directed at creating AGI. That said, it’s nice to showcase new cool technologies and products along the way, to make people smile, and hopefully to earn money—given all the demands for computational resources. When we launched ChatGPT, there were numerous questions along the lines of ‘who needs this, and where’s AGI?’ […]»

In a June podcast, Altman discussed the «great discrepancy of social networks»:

«One of the major mistakes of the social media era has been that feed algorithms have produced a host of unintended negative consequences for both society at large and individual users. Although they did exactly what the users wanted, or what someone thought they wanted—they kept people engaged on the site.»

Previously, OpenAI launched a special version of ChatGPT with parental controls for users under 18.