OpenAI has expanded access to its AI text-to-image generator, DALL-E 3. ChatGPT Plus and Enterprise users yesterday got access to the DALL-E 3 model within the ChatGPT app. DALL-E 3 allows users to employ ChatGPT for creating extended and more visually descriptive prompts to guide the image generator.
This wider release comes after OpenAI announced DALL-E 3 last month, giving The Verge a demo of its enhanced capabilities.
OpenAI’s DALL-E 3 addresses controversies
Interestingly, before this integration with ChatGPT, Microsoft’s Bing Chat and Bing Image Generator were the first platforms to offer broader public access to DALL-E 3. This collaboration enabled users to generate a variety of images, though not without controversy (see below). Despite Microsoft’s efforts to block certain prompts, users found ways to bypass these restrictions.
Critics have had concerns about text-to-image generators like DALL-E, Midjourney, and Stable Diffusion ever since they were released. Beyond the copyright issues of how AI generates images, there’s what users are generating, such as explicit content or misleading images of public figures.
Responding to these concerns, OpenAI has taken measures to ensure the safety and appropriateness of images generated by DALL-E 3. The company has published a paper detailing the research behind DALL-E 3. Additionally, the company claims it can identify images generated by DALL-E 3 with a 99% accuracy rate using its internal “provenance classifier” tool.
