OpenAI has released the DALLE 3 neural network with ChatGPT integration
OpenAI has introduced its next-generation neural network, DALLE 3, capable of analyzing and understanding longer, more detailed text messages than its predecessors, and generating incredibly detailed images that accurately reproduce all elements of the request. Moreover, the creators intend to combine it with ChatGPT to help users formulate messages that are as clear as possible to the image generator, as well as correct them if the result does not ultimately coincide with expectations.
DALLE 3 features:
DALLE 3 will not:
Everything that will be created by the user in the process of interaction with DALLE 3, he will be able to dispose at will. In other words, post on social networks, send, donate, sell and print. You do not need to obtain permission for any of these actions.
It is reported that DALLE 3 is already significantly ahead of DALLE 2 in accuracy and efficiency. Simultaneously with testing, the origin classifier is being polished, which will be able to distinguish the creativity of a neural network from a human one, if necessary.
At the moment, access to the new generation image generator is open only to the research group, but in October they promise to allow owners of paid ChatGPT, Plus and Enterprise subscriptions via API.
DALLE 3 features:
- Breaks the received complex instructions into small components, clearly picking out every significant detail from the description.
- Makes compositions taking into account the location of objects relative to each other.
- Has higher detail, providing realism to the generated image – for example, when depicting human hands, with which DALLE 2 had problems.
DALLE 3 will not:
- Copy the writing style of living famous artists in order to protect creators from attacks on their creativity. However, this ban does not apply to the masters of the past, such as Monet, Rembrandt or Vermeer.
- Generate so-called adult content or images related to cruelty.
- Draw media personalities to prevent the spread of fake news. Since the works turn out to be extremely believable, they could be passed off as photographs, causing a blow to the reputation of this or that person.
Everything that will be created by the user in the process of interaction with DALLE 3, he will be able to dispose at will. In other words, post on social networks, send, donate, sell and print. You do not need to obtain permission for any of these actions.
It is reported that DALLE 3 is already significantly ahead of DALLE 2 in accuracy and efficiency. Simultaneously with testing, the origin classifier is being polished, which will be able to distinguish the creativity of a neural network from a human one, if necessary.
At the moment, access to the new generation image generator is open only to the research group, but in October they promise to allow owners of paid ChatGPT, Plus and Enterprise subscriptions via API.
Comments: 0