Change of course for the DALL-E 2 artificial intelligence (AI) image generator software: its owner, OpenAI, is relaxing its rules of use and making it possible to upload images to the system.
DALL-E 2 makes it possible to generate high quality images from text. The addition of photorealistic faces was previously prohibited, for fear of seeing false creations – but which could pass for real – of people in compromising positions (pornography, for example).
But OpenAI says it has made the necessary improvements to its security system by
minimizing the potential for harm of the
deep fakes, notably. For example, the persons concerned will have to give their consent before their face is found in the software.
Respond to the request
OpenAI justified this reversal of the situation by the desire of Internet users to have a tool
to imagine outfits and hairstyles on oneself and modify the backgrounds of family photosamong others.
A reconstructive surgeon tells us he uses DALL-E to help his patients visualize results, and filmmakers tell us they want to edit footage of scenes with people to speed up their creative process.the company said in a letter to users.
Competition is also increasingly strong for this AI software. Others, like Stability AI’s Stable Diffusion, already allow photos of real people to be uploaded into their systems. The British company judges for its part that it is up to Internet users to decide whether their use is ethical or not.
The beta version of DALL-E 2 is only offered to a select group of people (approximately one million at the end of August), by invitation only.