The AI research business OpenAI has launched DALLE 2, an upgrade to its text-to-image generator that appears to be a significant step forward.
DALLE 2 can, in essence, make art from natural language input, such as "a picture of a fox sitting in a field at daybreak in the manner of Claude Monet." The objective, according to OpenAI, is to generate "original, realistic visuals and art" that can "combine concepts, qualities, and styles."
It's a massive accomplishment for a business working on numerous AI projects, like GPT-3, which can write some amazing text.
The following research article from OpenAI goes into more information about how DALLE 2 works. DALLE 2 operates in two steps, using the CLIP contrastive models: the first makes a CLIP image embedded with a text caption, and the second generates an image based on it.
The results are stunning, and they might have far-reaching implications for the art and graphic design industries, particularly video game firms, which now hire designers to develop worlds and concept concepts meticulously.
DALLE 2 expands on the work of OpenAI's DALLE 1, which was revealed in January 2021. The name is a mash-up of Salvador Dal and WALLE.
OpenAI is one of the fascinating businesses in the AI sector right now.
Elon Musk, Sam Altman, Ilya Sutskever, and others founded the company in 2015 to create a "friendly AI" that helps humanity. The non-profit startup competes with Alphabet's DeepMind.
DALLE 2 is a fantastic example of this technology in action, pushing the frontiers of what is possible for picture generation using AI. The service may even alter existing photographs to meet the customer's needs.
Of course, there is still a long way to go. The researchers discovered that labeling one thing as another can readily deceive its CLIP basis. However, these issues will be sorted out and improved upon over time.
OpenAI intends to make DALLE 2 available to a restricted group of academics soon.