Following numerous requests from DALL-E 2 users, OpenAI has authorized the modification of photos featuring real people.
Whether SLAB allowed this type of manipulationthis was not the case with DALL-E 2. This measure was taken with the aim of restrict misuse artificial intelligence (AI). However, many users have asked OpenAI to reconsider this restriction. Thus, the company responded favorably to this request to offer them the possibility of expressing their imagination more freely. This is possible thanks to the development of techniques to limit abuse.
A response to DALL-E 2 beta testers
One of the biggest problems with text-image generators like DALL-E 2 is that users can abuse of. The deepfakes are, among other things, one of the most widespread manifestations of these abuses.
Therefore, if OpenAI decided to change its usage policiesit is because he has improved its security system. In other words, the possibility of modifying the faces is accompanied by an initiative “minimizing the potential for harm” of the deepfakes. Of course, the company will also monitor attempts to generate violent pictures, sexual, etc Additionally, OpenAI communicated this major change to its customers via email. The content includes sample queries from them.
In particular, customers have reported on their use of SLAB for imagine yourself in some outfits or hairstyles. With this tool, other change the background of photos of their family. A reconstructive surgeon even communicated to OpenAI that he used DALL-E in his work. The text-image generator helped his patients to imagine the result of reconstructive surgery. In addition, film directors also expressed the desire to exploit DALL-E 2 within the framework of their profession.
A perpetual fight against abuse
This change in the DALL-E 2 usage policy will put OpenAI’s filtering technology to the test. It should be noted that customers have already reported complaints about the inaccuracy of the system as well as his overzealousness.
In any case, the fight against misuse will continue to be a major challenge for text-to-image generators. It is a perpetual battle whose the stakes are sometimes extremely high.
This is all the more true for OpenAI, which takes advantage of the Microsoft support and of large venture capital firms. He wishes to avoid controversies which affect in particular Stable Spread of Stability AI. This artificial intelligence also generates images from texts. However, the tool does not impose no restrictions to its users unlike DALL-E 2.
Just like OpenAI’s text-to-image generator, Stability AI’s allows you to edit photos of real people. Currently, a good number of individuals use Stable Diffusion in order to create pornographic content and not consensual of celebrities. Moreover, Emma Watson has already been a victim.