▲圖片來源：dataconomy

The list of features for DALL-E 2 now includes a brand-new feature called Outpainting. Like inpainting, a well-known component of various picture creation models, Outpainting modifies specific areas of an existing image while keeping its aesthetic.

Unlike inpainting, which makes changes just inside the constraints of the source image, this feature continuously extrapolates data to generate a complete scene from the source. The new tool expands source images to any final image size.

More than 3,000 artists from more than 118 countries have tested the system since its preview version was introduced months ago. Among these creatives are playwrights, screenwriters, landscape designers, tattoo artists, costume designers, and even creators of augmented reality (AR).

Artists and creative professionals use DALL-E 2 to inspire and improve their creative processes, enabling users to work fast and effortlessly. DALL-E 2 has already been used to produce magazine covers, music videos for kids with cancer, and the realization of creative ideas.

▲圖片來源：dataconomy

The OpenAI team collaborated with researchers, artists, engineers, and other users to learn about the dangers before releasing DALL-E 2 in beta. As a result of their conversations, the team has taken several initiatives.

Their technology forbids image submissions with realistic faces and tries to imitate famous people and influential leaders to prevent the tool from being used to distribute misleading information. Modern methods were also employed to avoid realistically producing genuine people’s faces.

DALL-E 2 LAUNCHES THE OUTPAINTING FEATURE

OpenAI has been displaying artists who successfully use the new functionality in the Twitter thread for the feature announcement. Check out the thread if you want to see what the new tool is capable.

WHAT IS OUTPAINTING?

Thanks to a technique known as “Inpainting,” changes within an uploaded or generated image are already feasible utilizing DALL-E 2’s Edit function. With Outpainting feature, users can now enlarge the original image and create large-scale images in any aspect ratio. This application considers the image’s existing visual elements, such as shadows, reflections, and textures, to preserve the context of the original image.

▲圖片來源：dataconomy

DALL-E 2, an AI system that generates original images and artwork from a description in natural language, is used by more than a million people today. Artists have already created stunning images using the new Outpainting tool, contributing to our understanding of its powers in the process.

The Outpainting feature is now available to all DALL-E 2 users on desktop. To explore new creative possibilities, visit labs.openai.com and get on the waiting list.

CONCLUSION

DALL-E 2 is a strong, adaptable instrument for creativity. Although the samples we’ve seen are wonderful and remarkable, they may have been selected primarily by OpenAI’s personnel. We don’t believe their intentions are evil, given the extensive problems they revealed in the system card document. Even so, we should, at the very least, exercise caution if they won’t permit unbiased researchers to examine DALL-E 2’s outputs.

When considering and examining models like the DALL-E 2, there is a position we prefer to adopt. It’s incredibly simple to stop exercising critical thinking and accept DALL-E 2’s stunning outputs. That is precisely what enables businesses like OpenAI to operate freely in a non-accountability sector that is all too typical.

▲圖片來源：dataconomy

Another question is whether building DALL-E 2 was indeed essential. Because it seems they wouldn’t be willing to halt deployment whether the risks can be effectively handled or not, we might ultimately experience a net negative (the tone from the system card document is clear: they don’t know how to address most potential worries).

That argument has a lot to say, so we will elaborate on it in a subsequent piece. DALL-E 2 has effects outside of the field of AI. For better or worse, DALL-E 2 will change the future.

轉貼自： dataconomy

若喜歡本文，請關注我們的臉書 Please Like our Facebook Page： Big Data In Finance