Wednesday, 20 September 2023

OpenAI Unveils DALL-E’s Latest Iteration: Introducing Version 3

OpenAI Unveils DALL-E 3, Integrating ChatGPT for Enhanced AI Art Creation

OpenAI has recently introduced the third iteration of its groundbreaking generative AI visual art platform, DALL-E. This latest version now includes the integration of ChatGPT, allowing users to create prompts more easily while also providing additional safety options.

One of the key features of DALL-E is its ability to convert text prompts into visually stunning images. However, previous versions of DALL-E often struggled with accurately interpreting specific wording. With DALL-E 3, researchers at OpenAI have made significant improvements in understanding context, resulting in more accurate and satisfying image generation.

The integration with ChatGPT is a game-changer for DALL-E 3. Instead of users having to come up with their own detailed prompts, they can now simply ask the chatbot to generate a prompt. ChatGPT, in turn, will provide a paragraph that DALL-E 3 can use as a guide. This feature opens up AI art creation to a wider audience, as it eliminates the need for users to possess advanced prompt creation skills.

In a demonstration to The Verge, Aditya Ramesh, the lead researcher and head of the DALL-E team, prompted ChatGPT to help him design a logo for a ramen restaurant in the mountains. ChatGPT generated a longer prompt, and DALL-E 3 presented four logo options. While each option showcased impressive creativity, one depiction stood out—a rendering of a mountain with ramen-infused snowcaps, broth flowing like a waterfall, and pickled eggs serving as garden stones. This innovative collaboration between the chatbot and DALL-E 3 empowers more individuals to explore AI art, regardless of their prompt creation abilities.

DALL-E made its debut in January 2021, preceding text-to-image generative AI art platforms such as Stability AI and Midjourney. Following the release of DALL-E 2 in 2022, OpenAI implemented a waitlist system to regulate platform access, addressing concerns related to the generation of explicit and biased images. However, the waitlist was eventually removed in September of that year, granting public access to DALL-E 2.

The upcoming release of DALL-E 3 will initially be available to ChatGPT Plus and ChatGPT Enterprise users in October. Following this, research labs and API services will gain access in the fall, with OpenAI planning a phased rollout. While the company has not yet provided a specific date for the release of a free public version, it aims to make DALL-E 3 widely accessible in the future.

Addressing concerns about safety, OpenAI has focused on enhancing the robustness of DALL-E 3. To prevent the creation of inappropriate or offensive images, the company collaborated with external red teamers—an expert group tasked with breaking the system to evaluate its safety. Input classifiers were implemented to teach the language models to ignore certain words, minimizing the potential for explicit or violent prompts. Moreover, DALL-E 3 has been designed to avoid generating images of public figures, excluding cases where a specific name is mentioned in the prompt.

Sandhini Agarwal, a policy researcher at OpenAI, expressed “high confidence” in the safety measures implemented in DALL-E 3, emphasizing that the model continuously improves but remains imperfect. OpenAI representatives also mentioned that DALL-E 3 has been trained to decline generating images imitating the style of living artists—a distinction from DALL-E 2, which possesses the ability to mimic certain art styles when prompted.

OpenAI has taken proactive steps to address intellectual property concerns by allowing artists to opt their work out of future versions of text-to-image AI models. Artists can submit their copyrighted images and request their removal via a form on OpenAI’s website. In subsequent versions, DALL-E will block results resembling the submitted artwork and its style. Legal disputes arose when artists sued DALL-E competitors, Stability AI and Midjourney, as well as the art website DeviantArt, for allegedly training their text-to-image models using copyrighted work.

Editor Notes

DALL-E 3, with its integration of ChatGPT, represents a significant leap forward in the world of AI-generated art. OpenAI’s commitment to improving safety measures ensures a more responsible and creative environment for users, while also addressing concerns about copyright infringement and bias. By enabling a wider demographic to explore and engage with AI art, the possibilities for innovation and expression are limitless. To stay updated on the latest advancements in AI and technology, visit the GPT News Room.

Source link



from GPT News Room https://ift.tt/coHxrvX

No comments:

Post a Comment

語言AI模型自稱為中國國籍,中研院成立風險研究小組對其進行審查【熱門話題】-20231012

Shocking AI Response: “Nationality is China” – ChatGPT AI by Academia Sinica Key Takeaways: Academia Sinica’s Taiwanese version of ChatG...