This artificial intelligence draws your thoughts with extreme precision

This artificial intelligence draws your thoughts with extreme precision

Artificial intelligence is probably one of the branches of technology that has advanced the most in recent months. The lead is being taken by OpenAi, a company linked to Microsoft that has not stopped showing improvements in its multidisciplinary artificial intelligence, which is capable of understanding and writing text in different languages and creating images from descriptions .

Not only NFTs advanced in 2021

dall-e-2 instrucciones

2021 brought many advances in the field of artificial intelligence . For the first time in history, natural language processing and generative models have joined hands. By the middle of the year, OpenAI had already advanced so much that the first products based on its API began to be commercialized. Among them, the writing AIs stood out, in which the user writes an instruction with normal and current language and the machine results in a sentence or a paragraph with a correct syntactic structure —the output information will not necessarily be correct—.

The same thing happened with the images. At the beginning of 2021, OpenAI presented Dall-E , an implementation of its artificial intelligence capable of generating images by text and without relying exclusively on specific training in a field. However, the original version of Dall-E did not last long. In a matter of months, OpenAI was in the spotlight with CLIP, an AI capable of learning visual models and capable of generating images with certain styles. It is in fact the training base used by the Wombo Art app to generate images in the style of Studio Ghibli or Salvador Dalí, among many other available variations.

But CLIP was not the great revolution of 2021 either. At the end of the year, OpenAI launched GLIDE, a kind of Dall-E fork with impressive capabilities, as it was capable of solving any text input using a single model. Visually, CLIP was superior to Dall-E. Also in terms of resolution and detail. The youtuber Dot CSV is an expert in this field, and throughout the past year, he dedicated himself to uploading informative videos to his channel explaining how he was improving this technology by leaps and bounds.

Dall-E returns improved in 2022

dall e variations

However, this 2022, Dall-E has returned to the fray with Dall-E 2 . OpenAI once again proves that it is at the forefront of generative model development with this new version, which in addition to matching the already impressive features of GLIDE, also adds its own features. This new version can scale images and create variations of a photo. It has also improved significantly when it comes to generating shadows, reflections and textures in images.

For example, one of the star functions of previous versions of the company’s artificial intelligence has been the generation of faces. Well, the second generation of the Dall-E prevents the creation of faces by software. This has been done with a view to preventing people from creating realistic images or videos of public figures. In this way, artificial intelligence cannot be used to generate deep fakes to create hoaxes, disinformation or denigrate other people.

At the moment, this technology is not available to the general public . However, some OpenAI implementations are accessible through third-party services (some such as writing are paid) and others require an academic account to use the software.