DALL·E 2 is an artificial intelligence program that creates artistic and lifelike images by using descriptions in natural language. It is built on GPT-3, which has 175 billion parameters, and pairs of text and images.
The beta version of DALL·E 2 has been released by OpenAI recently and its accuracy, realism, and resolution are greatly improved. One million people will receive invitations to leave their queue during the coming weeks.
In the first month, invited users will receive 50 free credits, which can be used to finish one original DALL·E 2 prompt and generate four pictures or get an edited change tip, which generates three pictures. Invited users will continue to get 15 free credits each month after that.
The beta version of DALL·E 2 is launched
Since the preview version of the system was published three months ago, more than 3,000 artists from more than 118 nations have participated in testing it. These creatives include illustrators, playwrights, filmmakers, landscape designers, tattoo artists, costume designers, and even augmented reality (AR) designers.
DALL·E 2, which enables users to work quickly and effortlessly, is used by artists and creative professionals to inspire and speed up their creative processes. Already, DALL·E 2 has been utilized to create magazine covers, music videos for children with cancer, and the fulfillment of innovative concepts.
Users can upload photographs they’ve taken with DALL·E 2 or created themselves and realistically and contextually change them using Edit. A DALL·E 2-generated or user-uploaded image can be modified to create fresh iterations that take inspiration from the original. The DALL·E 2 platform’s My Collection feature allows users to directly retain generations.
At the same time, authorities are taking precautions in order to regulate the future of artificial intelligence technologies. For instance, the UK prepared the AI rulebook two months after EU AI Act.
OpenAI team is taking measures against risks
Before releasing DALL·E 2 in beta, the OpenAI team worked with researchers, artists, engineers, and other users to learn about the risks. The team has implemented several actions as a result of their discussions.
To prevent the tool from being used to spread false information, their system forbids image uploads with realistic faces and tries to impersonate well-known people and powerful politicians. Modern techniques were also used to prevent the photorealistic production of real people’s faces.
They have enhanced the precision of content filters while protecting free expression in order to better prevent images that violate their content guideline. This prohibits people from producing content that is violent, adult-oriented, or political, among other types. To minimize DALL·E 2’s exposure to these concepts, the most graphic training data was taken.
The team has utilized an innovative technique to make the generated photos of humans in DALL·E 2 more reflective of the diversity of the global populace. This technique is applied at the system level when a prompt is given that does not indicate race or gender, such as “CEO.”
The team will continue to employ automated and human monitoring measures to stop the abuse.
“Expanding access is an important part of our deploying AI systems responsibly because it allows us to learn more about real-world use and continue to iterate on our safety systems. We are continuing to research how AI systems, like DALL·E 2, might reflect biases in its training data and different ways we can address them,” OpenAI wrote in its blog post. What do you think, will AI replace designers in the future?