a tool based artificial intelligence (AI), possible convert any text to an image is the newest creation Google. The technology, presented on Monday 23, is not yet expected to be made public as it needs some improvements, but has already caught the attention of experts.

judicialpicture”, the tool works similarly DALL-EIt is a program that creates images from texts developed by OpenAI. It’s pretty simple, just open the software and type in any random description that comes to mind, such as “two robots sitting at a table with the Eiffel Tower in the background” and wait for the magic to happen.

The platform’s artificial intelligence can rely on a range of styles to render the image, such as oil paintings, photographs, and CGI renderings. In some cases, the result is not what the user expects, as the scene drawn by the search giant’s robot looks unfinished, blurred or smeared, but in others the drawing is quite realistic.

A team of human evaluators hired by the Mountain View company, Photos created with Imagen technologyBy comparing with drawings created by competing software. For some, Google’s system delivered better results than the longer-developed DALL-E.

Problem found

Although it has demonstrated that it can produce photographs that are closer to reality, Google program that creates images from text It has some limitations and problems. Big tech said, without giving further details, AI often encodes social biases as well. come across as racist, sexist, or toxic another way.

AI showed surprising results.

how do you remember BoundaryThe problem may lie in the web data used to train the algorithm, possibly without any improvement, including those that reflect social stereotypes and derogatory associations with certain marginalized groups. This allowed the tool to learn and copy some of the types of behavior found online.

Only images of animals, food and objects were shown by the Mountain View company.

The company itself acknowledged that the show may prioritize “image generation of people with lighter skin tones and a trend towards images depicting different occupations to accommodate Western gender stereotypes.” Interestingly, the photos of the people generated by the engine were not published by Google.

More images automatically generated by technology.

Due to these issues, the tech still needs some improvements and testing before it’s rolled out to users in general, its usability hasn’t been confirmed by the tech giant either.

Source: Tec Mundo

Previous articleThe big winner of the 17th stage of the Giro d’Italia 2022 is Santiago Buitrago!
Next articleWhy the Star Wars prequels aren’t as bad as we thought

LEAVE A REPLY

Please enter your comment!
Please enter your name here