Robots creating images? It looks like something out of a movie, but it’s just Google’s newest service that uses artificial intelligence (AI) to generate any image from text. The technology can mix different styles, be it oil paintings, photographs and CGI renderings.
Named as Imagen, the technology giant’s new service has already become a pioneer in the field of image creation due to the quality in which the product is displayed. Until then, the leader in this market was DALL-E, a program from the OpenAI laboratory.
However, it is important to point out that the service has not been made available to the public and that the stunning images created from AI don’t come out perfect every time. Images generated by text templates can appear unfinished or blurry, common problems with the technology that were also found in the competitor DALL-E.
The results presented by Google show the best images created from text, that is, some filters were made before publication.
In addition to generating imperfect figures, the product of images from AI was also not released to the public due to the legal problems it can generate. Although most people use creativity for good, it is possible that some apply the system to create fake news, for example. Google itself points out that the tool can encode social biases, that’s because it analyzes large amounts of data and they are based on patterns to learn and replicate.
The show “codes various social prejudices and stereotypes, including a general bias to generate images of people with lighter skin tones and a tendency for images depicting different professions to align with Western gender stereotypes.”
This problem can also be found in DALL-E, for example, one of the reasons that did not allow the tool to be made available to the public. But both companies point to the use of filters that try to prevent the creation of racist, violent and pornographic images.
Have you watched the new videos on YouTube of the Digital Look? Subscribe to the channel!