Earlier this year, DALL-E 2 showed us its enormous ability to generate images from any text and a notable improvement over the first generation. However, during all these months, this OpenAI artificial intelligence has had an important limitation: the conditions of use prevented researchers who have access to it from sharing auto generated face picturesbut this just changed.
As our colleagues from Genbeta collect, DALL-E 2 has been able to generate photorealistic faces from the beginning, but those responsible for the project did not allow its circulation outside doors due to the rrisk of creating deepfakes and false content. So what has changed so researchers can share auto-generated faces? OpenAI has implemented a number of changes to improve the security of this technology.
DALLE-2 can generate faces, but with these limitations
To reduce risks, OpenAI ensures that DALL-E 2 now prevents researchers from generating photorealistic faces from any “realistic face”. In addition, the platform rejects any attempt to generate faces of celebrities and public figures. It should be noted that previously it was not allowed to generate faces of prominent political figures, but after this update, the limitations include any famous personalities.
We always talk about researchers because, unlike DALL-E Mini, DALL-E 2 is currently in a closed testing phase, so it is not yet available to the general public. Even those who have access to this technology still have certain limitations. However, numerous examples of its operation can be found on social networks. And some publications let us see what the photorealistic faces generated by this technology look like.
— Danielle Computer Images 💿 (@djbaskin_images) June 21, 2022
Danielle Baskin, via Twitter account @djbaskin_images describes how you have used DALL-E 2 to generate photorealistic faces with different makeup designs. “I added the phrase ‘triangular geometric neon yellow’ to my directions and got these amazing hair and makeup ideas,” she says, adding “when you look into the eyes [a las imágenes generadas]you know it’s about artificial intelligence.”
— Ken Kocienda (@kocienda) June 19, 2022
An example of how DALL-E 2 can be used to generate images from another image can be seen in a Tweet by Ken Kocienda. Starting from Johannes Vermeer’s famous painting ‘Girl with a Pearl Earring’, this user asked to generate a studio scene, with photographers, cameras on tripods, and lots of lights, yes, in “extremely detailed oil painting”. The result, which is surprising, highlights the potential of this type of constantly evolving tool.
Behind DALLE-2 is GPT-3, a powerful language model introduced in 2020 capable of generating various types of content through written prompts, and a “broadcast” model, which start with a pattern of random dots and gradually alter that pattern to generate an image when it recognizes specific aspects. In this field we also find other promising proposals, such as the recently presented Google Parti, which uses an autoregressive model and allows longer text entries.
In Xataka | The first judge made by artificial intelligence is quite impartial. Bad (and good) news for justice