New Delhi: OpenAI, the San Francisco-based artificial intelligence startup on Wednesday, unveiled an upgraded iteration of its DALL-E image generator, referred to as DALL-E 3, to a select group of testers. This advanced version has been seamlessly integrated into ChatGPT, OpenAI’s widely used online chatbot.
DALL-E 3 boasts an enhanced capability to generate images with increased realism compared to its predecessors. Notably, it exhibits a remarkable proficiency in generating images featuring letters, numbers, and human hands, as stated by OpenAI.
“It is significantly more proficient at comprehending and accurately representing user requests. Our aim was to develop technology with a more precise command of the English language,” Aditya Ramesh, a researcher at OpenAI, said.
With the integration of the latest DALL-E version into ChatGPT, OpenAI is cementing its chatbot’s position as a central hub for generative A.I. This encompasses the generation of text, images, audio, software, and various other digital media independently. Since ChatGPT gained widespread attention last year, it has sparked a competitive race among leading Silicon Valley tech companies to lead in A.I. advancements.
Meanwhile, Google on Tuesday launched an updated version of its chatbot, Bard, enhancing its integration with popular services like Gmail, YouTube, and Docs. Additionally, other image generators like Midjourney and Stable Diffusion updated their models earlier this summer.
OpenAI, on the other hand, has been connecting its chatbot to various online services such as Expedia, OpenTable, and Wikipedia for a while. However, this recent development marks the first time OpenAI has combined a chatbot with an image generator.
Previously, DALL-E and ChatGPT were separate applications. With the latest update, users can employ ChatGPT’s capabilities to create digital images simply by describing their preferences. Alternatively, they can generate images using descriptions generated by the chatbot, streamlining the production of graphics, art, and other forms of media.
In a demonstration, Gabriel Goh, an OpenAI researcher, showcased ChatGPT’s ability to generate detailed textual descriptions that can subsequently be used to create images. For example, the bot quickly generated multiple images based on textual descriptions of a logo for a restaurant called Mountain Ramen.
The new iteration of DALL-E is proficient at generating images based on multi-paragraph descriptions and meticulously following specific instructions. However, like all image generators and A.I. systems, it is not immune to errors.
OpenAI is currently fine-tuning the technology and plans to release DALL-E 3 to the wider public next month. It will be accessible through ChatGPT Plus, a subscription service priced at $20 per month.
Experts have raised concerns about the potential misuse of image-generating technology for spreading disinformation online. To address this, OpenAI has incorporated tools into DALL-E 3 designed to prevent problematic content, such as sexually explicit images and depictions of public figures. The company is also working on restricting DALL-E’s ability to mimic specific artistic styles.
In recent months, A.I. has been a source of visual misinformation, with instances like a synthetic, albeit unsophisticated, simulation of an explosion at the Pentagon briefly affecting the stock market in May. Voting experts also express concerns that this technology could be maliciously employed during major elections.